information to users

INFORMATION TO USERS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted.

Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Photographs included in the original manuscript have been reproduced xerographically in this copy.

Higher quality 6” x 9" black and white

photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

Bell & Howell Information and Learning 300 North Zeeb Road, Ann Arbor, Ml 48106-1346 USA 800-521-0600

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.


DECISION-MAKING FOR ALLOCATION OF PUBLIC RESOURCES IN DECENTRALIZED DISTRICT HEALTH SYSTEMS IN UGANDA

by George W illia m Pariyo

A dissertation subm itted to the Johns H opkins U n ive rsity in c o n fo rm ity w ith the requirem ents fo r the degree o f D octor o f P hilosophy

B altim ore, M aryland Novem ber, 1999

©G eorge W illia m Pariyo 1999 A ll rights reserved


UMI Number: 9964185

Copyright 1999 by Pariyo, George William

All rights reserved.

UMI UMI Microform9964185 Copyright 2000 by Bell & Howell Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code.

Bell & Howell Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346


ABSTRACT

The main aim o f this study was a better understanding and documentation o f how decisions are being made to allocate resources within the public health sector at district level in Uganda and how this process could be improved e.g., through use o f composite indicators like the D A L Y or HEALY. The setting is in the context o f decentralization which began in 1993. It was an exploratory case study and involved 10 districts. A ll 68 district health managers, administrators, and top political leaders who were available and 30 key informants were interviewed. Both structured questions with use o f Likert type scales as well as open-ended questions were used. Information was obtained about their perceptions and the district health system and support structures. Focus Group Discussions were conducted with 21 politicians o f the District Council. Allocations to PHC and hospitals were analyzed from health plans and budgets. Record review and observations were done to complement the interviews. Content analysis was done manually. Quantitative analysis was done to produce frequencies, proportions and per capita financial allocations. Relationships were tested with correlation and multiple linear regression analysis. STATA (Version 6) and SPSS for Windows (Version 8.0) were used. Decision modeling was done with the aid o f flow-charts and If/Then statements. The main findings include identification o f 7 main decision rules that districts follow in allocation decisions based on availability o f funds, historical categories, presence o f national guidelines, number o f people affected, potential to cause death, popular local demand, and decision makers expected benefits. An average o f 59.1 % (standard deviation o f 14.7%) o f total health budgets were found to be allocated to PHC. There was no evidence o f link o f resource allocation to health services coverage or health status measures in all districts. National disbursements to districts were

ii


largely based on population and existing infrastructure which does little to address existing inequities although there were indications that relatively more donor funds were going to the disadvantaged districts. The main challenges in improving resource allocation decision-making are the relative lack o f data for use o f burden o f disease and cost-effectiveness approaches. However, the study shows that a lot o f information potentially available is not being used to guide allocation decisions. Challenges are a lack o f confidence and skills in information handling, continued perception o f lack o f decision-making power at the district level, as well as an over-stretched capacity to deal with the many tasks o f a decentralized health system. The Rockefeller Foundation provided financial support for the study.

Thesis Advisory Committee:

Dr. Richard Morrow (Professor) - Advisor Dr. Gilbert Burnham (Associate Professor) Dr. David Bishai (Assistant Professor)

Supervision o f Uganda Fieldwork: Dr. Fred Wabwire-Mangen (Senior Lecturer)

Committee o f Final Readers:

Dr. Kenneth H ill (Professor) - Chairman Dr. Alan Lyles (Assistant Professor) Dr. Richard M orrow (Professor) - Advisor Dr. Gilbert Bumham (Associate Professor) Dr. Helen Abbey (Professor) - Alternate Dr. Timothy Baker (Professor) - Alternate

iii


Ad maiorem Dei gloriam

iv


PREFACE

This doctoral dissertation is the culmination o f research work conceived with the aim o f documenting the many current processes, challenges, and opportunities in district level public health resource allocation associated with decentralization in Uganda. It's focus is on decision-making. It should be pointed out from the beginning that there have been many positive developments in Ugandan districts since the advent o f decentralization. There are also many problems. This work was concerned with those aspects and issues more directly relevant to resource allocation decision making and does not attempt to do a formal evaluation o f the decentralization process in Uganda as such. The manuscript is divided into four main parts. Part I contains the chapters that cover the overall methodological approach to the study relevant to the three thematic papers presented under Part II. Part I therefore covers statement o f the problem, reasons why the study was deemed relevant, objectives and guiding questions for the study in Chapter 1, literature review in Chapter 2, conceptual framework in Chapter 3 and methods in Chapter 4. Part II is made up o f the three papers which are the core o f this work with each addressing an aspect o f resource allocation decision-making. Paper 1 (Chapter 5) examines the current processes, influences, stake-holders, and decision-maker preferences. Paper 2 (Chapter 6) looks at the information currently available or potentially obtainable. It identifies the gaps that could be filled and suggests some basic steps and resource allocation indicators in order to move to a more objective, needs-based and cost-effectiveness approach to resource allocation such as the use o f a composite indicator like the Disability Adjusted Life Year (D A L Y ) or the Healthy Life Year (H E A LY ). Paper 3 (Chapter 7) looks at the functioning and capacity o f the district health system. It examines those contextual issues like management structures, staff perceptions, logistics, health

v


status and health services utilization. Part III contains a general summary discussion o f the study findings and concluding remarks (Chapter 8). It highlights the need and areas for future research on the subject. Finally Part IV is a collection o f appendices and bibliography. The appendices include a map that shows the study areas, organizational charts o f district management structures, technical notes, and the study instruments that were used. The notes cover some o f the more technical aspects relevant to decision-making. A brief curriculum vitae o f the investigator is appended at the end o f the manuscript.

vi


ACKNOWLEDGMENTS

I would like to thank all those persons who contributed to make this work possible. These persons are too numerous to mention exhaustively. I mention here those names that come to my mind o f persons whose support in seeing me through was particularly useful. I apologize for those whom I might have inadvertently omitted. First and foremost, my Advisor. Dr. Richard Morrow, who together with Dr. David Bishai and Dr. Gilbert Bumham formed my thesis advisory committee. They all have busy work schedules but availed themselves to provide suggestions, encouragements, and critique various stages and drafts during this work which were crucial in enabling me get through to this point. Dr. Fred Wabwire-Mangen o f the Institute o f Public Health, Makerere University, provided overall local supervision o f the research process in Uganda and offered suggestions to improve the readability of the manuscript. Valuable advice and suggestions also were received from Drs. Baker and Winch of the Department o f International Health. I am most grateful to the members o f the committee o f Final readers for having accepted to read my thesis and form the panel for the final oral examination. They are Dr. Kenneth H ill (Professor). Dr. Timothy Baker (Professor), Dr. Helen Abbey (Professor). Dr. Richard Morrow (Professor), Dr. Alan Lyles (Assistant Professor) and Dr. Gilbert Bumham (Associate Professor). M y choice o f topic for the final dissertation research was motivated and helped along the way by the many faculty at the School whose courses or advice shaped my thinking and provided the impetus to study the problem o f decision-making fo r resource allocation. They are Richard Morrow, Gilbert Bumham, David Bishai, W illiam Reinke, Jonathan W e in e r, Timothy Baker, Carl Taylor, Alan Lyles, Peter Winch, Henry Mosley, Christopher Cassirer and Alan Sorkin. I would also like to thank those whose management courses provided the background on which I was able to

v ii


build upon coming to Hopkins. They are Ranieri Guerra and his team at ICHM in Rome. Management Sciences for Health (M SH) o f Boston, and Christopher Murray o f Harvard University. Collaborators such as Dr. Bart Criel o f the Institute o f Tropical Medicine (Antwerp), and Dr. Daniele Giusti o f the Uganda Catholic Medical Bureau, IPH colleagues Dr. Olico-Okui, and Dr. Maurizio Murru, deserve mention for their past discussions with me on some o f the issues addressed in this study. I would like to thank Prof. Gilbert Bukenya and Dr. Mark White both previously at the IPH for their support in my decision to pursue further education. Other colleagues at Makerere and in the districts with whom I have interacted over the years definitely contributed to my thinking and I thank all o f them. Colleagues and senior officers at my parent employer, the M inistry o f Health, provided useful advice and insights. In particular, I would like to mention Mr. Nathan Obore, former Permanent Secretary in the M inistry o f Health, Prof. Francis Omaswa, Director General o f Health Services, Drs. Henry Mwebesa, Prosper Tumusiime, Sam Okuonzi, Ambrose Talisuna, and Mr. Chris Mugarura, who availed themselves to meet me and provided useful supplementary information. Mr. Francis Luwangwa and Dr. Kiyaga o f the M O LG provided valuable insights and background documentation for which I'm grateful. Research assistance was provided by Ms. Edna Jurugo. and Drs. Robinah Najjemba and Margaret Lamunu. Together with drivers David Kiwanuka. Dan Sempebwa, and Fred Ibanda. they braved many road hazards with me in the trips to the districts. To all these people I ’ m very grateful. Thanks to other IPH staff for moral support (Francis Ngabirano, Jane Frances, Nabatanzi, Alice. Aida and others). M y most sincere thanks to all the respondents drawn from among the citizens, health managers, administrators, and district political leaders from the study districts o f Arua, Bushenyi, Hoima. Kapchorwa, Kibaale, Kiboga. Masaka, Mbale, Mbarara and Moroto.

v iii


I thank in a special way the Honorable M inister o f Health o f Uganda, Dr. Chrispus Kiyonga. who found time from his many commitments while at Hopkins to discuss with me and Dr. WabwireMangen some o f the issues o f improving health o f Ugandans, and provide insight into the health policies. W ithout financial support from the Rockefeller Foundation neither my studies at Hopkins nor the field-work for the dissertation would have been possible. The fieldwork support was provided under the Africa Dissertation Internship Awards Program. I w ill always be grateful to the Foundation and it's officers for their support. I would like to make a special mention o f those persons I dealt with; Omeata Prawl. An Trotter, Charlanne Burke and Florence M uli-M usiim e o f the New York and Nairobi offices respectively, as well as Seth Berkeley and Doris Mugrditchian formerly at the Foundation for their tim ely actions. I have benefitted over the years from an education to life in a friendship spanning many countries and localities, too numerous to list here. These friends have always reminded me o f the reasons to study and do the work that I do. I thank especially Fr. Luigi Giussani, through whom I met all o f them. I thank all those who offered a special friendship during the time o f my studies and supported me and my fam ily in various ways to carry on. They are; Anne and Michael Nganda. Fr. Edo M orlin, Cristina and Giuliano Gargioni, my brother Stanley and cousin Florence. Cristina Ravera and family, neighbours the Okwero, Ocen, and Tibaijuka families, Fr. Peter Tiboni, Carol and Julius Masiga, Charles Ombanya, Philip Aribo, Stefano Cerutti, Marisa and Alberto Scotti. Massimo and Giuliana Robberto, Margie and John McCarthy, Sara and Carlo Lancelloti, Rachel and Chin Pham, Melanie and Mark Danner, Grainne and Tom Tobin, and all other friends in Kampala and Washington, thank you! I shared many happy moments o f fun and relaxation in between the long hours at study with

ix


some o f these people who at one time or another were my house-mates (Zahid A li. Matthias Ringkamp, Bernard Hayek, Ozan Tutunoglu, and Elie Salameh), or Saturday breakfast buddies at 2913 G uilford (Nader and Sereen, Mandana. Andrei, Richard, R alf and Karina), or friends and fellow struggling doctoral students (Che-Ming Yang, M aria Lin. Claire Lucas, Paurvi Bhatt. Chris Coles, Chris Mast, R o lf Klemm, Paulo Baroso, N iloufer Randeria, Mandana Hajj, Matt Lynch, Sonia Singh, Gabrielle Breugelman, Melisse Murray, Rebecca Malouin, Isis Pluut, Juliana Yartey, Nafis and ‘ Kuor Kum oji), or one-time classmates and Hopkins friends (Mani Sheik, Barbara Kerstiens, Elizabeth Childs. Hany Abdalla, Irene Kuo, Priya Duggal, T iki Firdu, Olugbenga Obasanjo. W ilfred Emonyi, Yaika Jeng, Todd Koppenhaver). or fellow compatriots in the struggle for better health in A frica (members and friends o f the African Public Health Network), Ugandan countrymen and women at Hopkins in my time (Jackie Luyimbazi, Cissy Kityo-Mutuluuza. Noah Kiwanuka. Fred Nalugooda, Fred Makumbi. Godfrey Kigozi, the Hon. Chrispus Kiyonga, Linda Barlow), I thank them all. Their companionship helped to lighten the mood and refresh my mind at critical moments. M y appreciation goes to the Dean, Associate Deans, all the faculty and staff in the school, particularly the Department o f International Health, the International Student Office, libraries, security. Information Systems and Registrar's Office, for making my stay at Hopkins a happy and memorable one. For their readiness to assist or offer useful advice I thank especially Teresa Callison. Velma Pack, Nancy Stephens, Carol Buckley, Murray Welsh. Betty Addison, Lenora Davis, Bruce M ille r and Katrina Alston. Finally. I would never have persevered without the love, care, understanding, and support o f those most dear to me. my family; my wife Annette, children Giovanna, Peter and Edward who endured my long absence from home (I could hear them fighting for the phone each time I called!). M y parents and others who sacrificed so much to keep me alive and in school, especially my father Joseph Odradria, my beloved mother Petronella Komurwemo Ateenyi (RIP), uncles Raphael A liyo,

x


Peter Adaku, John Nyakua and Henry Mbarwa, aunt Perpetua Atriezo (RIP), grandmother Lady Bridgit Mateeka Nyakabiito Amooti (RIP), aunts Kevina and E m ily for the encouragement, grandfather Andrew Ejidra (RIP), and grandmother Rosa Odraa (RIP), together with all other family members and neighbors who collectively raised me. As the saying goes, it takes a village to raise a child! I should thank all my fellow siblings and cousins, with whom I shared a loving extended fam ily environment. I remember especially the companionship o f my late brother Francis Xavier Jurua Mugisa Atwooki (RIP). I started my struggles for education follow ing in his footsteps and together we endured many challenges, most memorable o f which was escaping being eaten by a leopard on our way home from school!

xi


DECLARATION

This manuscript is being presented for purposes o f fu lfillin g the dissertation requirements for the award o f a Doctor o f Philosophy (PhD) degree o f the Johns Hopkins University. It has not been submitted anywhere else for any academic award nor has it been published before. A ll sources o f materials quoted in this manuscript that are not my own original work have been cited and indicated in the list o f references. Errors, i f found, should not be attributed to the people acknowledged here unless so stated. I remain responsible for the contents and opinions expressed in this work and any comments and feedback on it w ill be appreciated.

George W illia m Byaruhanga A kiiki Pariyo Baltimore, Maryland (USA) November 1999

x ii


TABLE OF CONTENTS

Abstract ..................................................................................................................................................ii P re fa ce .................................................................................................................................................... v Acknowledgments ...............................................................................................................................vii D eclaration........................................................................................................................................... xii Table o f Contents............................................................................................................................. xiii List o f T a b le s ...................................................................................................................................

xxi

List o f Figures and Boxes ............................................................................................................ xxiii List o f A bbreviations..................................................................................................................... xxiv Operational Definition o f T e rm s ............................................................................................... xxviii

PART I: M ETHO DO LO G ICAL A P P R O A C H ................................................................................ 1 CHAPTER 1: IN T R O D U C T IO N ........................................................................................... 2 1.1 Statement o f the P roblem .....................................................................................2 1.2 Study Ju stifica tio n ............................................................................................... 3 1.3 Background Health S ituation.............................................................................. 5 1.4 Scope o f the S tu d y ............................................................................................... 8 1.5 Research Objectives and Q uestions.................................................................. 10 1.5.1 Goal:

.................................................................................................. 10

1.5.2 O bjectives:......................................................................................... 10 1.5.3 Guiding Questions: ...........................................................................11 CHAPTER 2: LITERATURE REVIEW .............................................................................14 2.1 Health Sector R eform s..................................................................................... 14

x iii


2.2 Allocations for H e a lth ........................................................................................ 16 2.3 Patient and Community Perspectives ..............................................................21 2.4 D ecisio n -M a kin g ................................................................................................ 22 2.5 Organizational and Management Theories ..................................................... 26 2.5.1 Bureaucratic Approach .................................................................... 26 2.5.2 Political Approaches ........................................................................ 29 CHAPTER 3: CONCEPTUAL FRAM EW O RK ................................................................ 32 3.1 General Health Systems M o d e l........................................................................ 32 3.2 Decision-Making Model ................................................................................... 34

3.3 Rational Decision-Making F ra m e w o rk............................................................36 CHAPTER 4: M ETHO DO LO GY ....................................................................................... 39 4.1 Study Setting ...................................................................................................... 39 4.2 Study Design ......................................................................................................42 4.3 Variables ............................................................................................................ 42 4.4 Intervening V ariables......................................................................................... 43 4.5 Data Collection ..................................................................................................43 4.6 Sample Size D eterm in atio n...............................................................................45 4.7. Sampling Procedure ......................................................................................... 46 4.8 Quality Assurance ..............................................................................................47 4.9 Analysis and Interpretation ...............................................................................47 4.10 Model S election................................................................................................48 4.11 Threats to V a lid ity ............................................................................................49 4.11.1

Internal V a lid ity .............................................................................49

x iv


4.11.2

External V a lid it y ..........................................................................51

4.12 Protection o f Human Subjects and Ethical Considerations......................... 51 4.13 Utilization and Dissemination o f Results ..................................................... 52

PART II: TH EM E PAPERS .............................................................................................................. 55 CHAPTER 5: PROCESS A N D REASONS FOR CURRENT DECISIO N-M AKING PRACTICES ............................................................................................................ 56 5.1 Introduction ........................................................................................................56 5.2 Background and O bjectives.............................................................................. 57 5.2.1 Overall Goal o f the Study ............................................................... 57 5.2.2 Objectives ......................................................................................... 58 5.3 Subjects and Methods .......................................................................................59 5.4 R e s u lts ................................................................................................................ 59 5.4.1 Respondents.......................................................................................59 5.4.2 District Health Planning and Budgeting ........................................ 60 5.4.3 General Decision R u le s ....................................................................63 5.4.3.1 Decision Rule 1: Historical C ategories........................ 63 5.4.3.2 Decision Rule 2: A vailability o f M o n e y ...................... 64 5.4.3.3 Decision Rule 3: A vailability o f G u id e lin e s................65 5.4.3.4 Decision Rule 4: Number o f People A ffe c te d

68

5.4.3.5 Decision Rule 5: Problem Causes D e a th ...................... 68 5.4.3.6 Decision Rule 6: Demand by Influential Persons . . . . 72 5.4.3.7 Decision Rule 7: Expected Personal G a in .................... 73 5.4.3.8 Other Influences on Allocations .................................... 75

xv


5.5 Discussion .......................................................................................................... 85 5.6 Summary and Conclusions ...............................................................................90 5.7 Policy Implications ........................................................................................... 91

CHAPTER 6: DEVELOPING INFO RM ATIO N FOR RESOURCE A LLO C A TIO N D E C IS IO N -M A K IN G ..............................................................................................92 6.1 Introduction ........................................................................................................ 92 6.2 Background and O bjective s...............................................................................93 6.2.1 Overall Goal o f the S tu d y :................................................................93 6.2.2 Study O bjectives:...............................................................................94 6.3 Subjects and Methods ....................................................................................... 94 6.4 Results ................................................................................................................ 97 6.4.1 Main Findings ...................................................................................97 6.4.2 Use o f Data for Allocations and Payments .................................... 99 6.4.3 Financial A llo ca tio n s .................................................................... 100 6.4.3.1 What is Money Allocated fo r? .................................... 100 6.4.3.2 Influences on Financial Allocations ......................... 107 6.4.4 Health Management Information System .................................. 116 6.4.4.1 Criteria Suggested for Resource A llo ca tio n s

117

6.4.4.2 Requests for In fo rm a tio n ............................................. 118 6.4.4.3 Estimating Disease Burden ........................................... 120 6.4.4.4 Effectiveness o f Current Interventions ..................... 120 6.4.4.5 Cost o f Inte rve n tion s......................................................122 6.5 Discussion ...................................................................................................... 127

xvi


6.5.1 Allocations for PHC .......................................................................127 6.5.2 Use o f Data ......................................................................................128 6.5.3 Allocations for Service D e liv e ry ....................................................129 6.5.4 Infrastructure E ffe c t.........................................................................131 6.5.5 E q u ity ................................................................................................ 132 6.6

Summary and Conclusions .......................................................................... 134

CHAPTER 7: TOWARDS BU ILD IN G C A P AC ITY FOR EFFECTIVE DECISION M A K IN G IN DISTRICT LE V E L RESOURCE A L L O C A T IO N .................... 135 7.1 Introduction ...................................................................................................... 135 7.2 Background ...................................................................................................... 137 7.2.1 Overall Goal o f S tu d y :.....................................................................137 7.2.2 Study O bjectives:............................................................................ 137 7.3 Subjects and Methods ..................................................................................... 138 7.4 R e s u lts ...............................................................................................................141 7.4.1 District Profiles ............................................................................... 141 7.4.2 Life Expectancy............................................................................... 142 7.4.3 Health Services Utilization ............................................................ 144 7.4.3.1 OPD Attendance ............................................................ 144 7.4.3.2 Ante-Natal C linic A ttendance.......................................146 7.4.3.3 Infant Measles Immunization Coverage ......................148 7.4.4 Ownership and Team w ork.............................................................. 150 7.4.5 Capacity for Financial M anagem ent............................................. 150 7.4.6 Bureaucracy..................................................................................... 152

x v ii


7.4.7 O ffice M anagem ent........................................................................ 152 7.4.8 Information Management............................................................. 153 7.4.9 Policy Interpretation ...................................................................... 154 7.4.10 Planning and B udgeting................................................................ 154 7.4.11 Corruption in Health Units ..........................................................155 7.4.12 Management S tructures................................................................155 7.4.13 Staff W e lfa re .................................................................................157 7.4.14 L og istics......................................................................................... 160 7.4.15 Supervision o f PHC units ............................................................161

7.5 Discussion ........................................................................................................ 164 7.5.1 Life Expectancy...............................................................................164 7.5.2 OPD Attendance .............................................................................166 7.5.3 ANC Attendance ............................................................................ 167 7.5.4 Infant Immunization Coverage ..................................................... 168 7.5.5 Getting Value-for-Money ..............................................................169 7.6 Summary and Conclusions ............................................................................ 171 7.7 Policy Implications ......................................................................................... 172

PART III: INTEGRATING R E M A R K S ....................................................................................... 173 CHAPTER 8: GENERAL DISCUSSION A N D CONCLUSIONS ................................ 174 8.1 General Comments......................................................................................... 174 8.1.1 Re-orienting to PHC and Equity ................................................... 175 8.1.2 Public/Private M i x .......................................................................... 177

x v iii


8.1.3 Compliance with Standards............................................................ 178 8.1.4 Use o f Available Resources............................................................ 179 8.1.5 Ownership and Control ................................................................180 8.1.6 Information Costs ...........................................................................182 8.1.7 Improving Allocation Decision-Making .......................................182 8.1.8 Rational Allocation Framework ................................................. 185 8.1.9 Decentralization and Dem ocracy................................................... 187 8.2 General Conclusions ....................................................................................... 190 8.3 Future Research............................................................................................. 191

PART IV: REFERENCES.................................................................................................................193 APPENDICES

.................................................................................................................. 194

Appendix A: Structure o f Community Representation and Governance in Uganda . . 194 Appendix B: Hierarchy o f Groups Involved in Health Planning and B u d g e tin g

195

Appendix C: Basic Structure o f District Health Management Teams in Uganda . . . . 196 Appendix D: Hierarchy o f O fficials Managing District Health Finances ......................197 Appendix E: Steps in Building a Rational Decision-Making Model ........................... 198 Appendix F: Technical Notes on Principal Components and Factor Analysis .............203 Appendix G: Comprehensive District Health Services Budget Items ........................... 208 Appendix H: Selected District Profile and Resource Allocation Indicators .................212 Appendix I: Detailed Financial Allocation Indicators by StudyD is tr ic t.........................222 Appendix J: Casewise Diagnostics - SelectedFinancial Indicators ...............................232 Appendix K:Casewise Diagnostics - Selected Health Outcomes and Utilization Indicators

.............................................................................................................................235

x ix


Appendix L:

CHECK-LIST OF QUESTIONS FOR K E Y IN FO R M A N T INTERVIEW S ......................................................................................... 237

Appendix M:

FOCUS GROUP DISCUSSION GUIDE FOR H E A LT H CO M M ITTEE M E M B E R S ................................................................................................238

Appendix N:

DISTRICT PROFILE A N D GENERAL IN F O R M A T IO N .................. 239

Appendix O:

Q UESTIONNAIRE FOR DISTRICT P O LITIC A L LEADERS/ADM INISTRATORS ......................................................... 250

Appendix P:

GENERAL INFO RM ATIO N ON FINANCES FOR H E A LTH IN THE DISTRICT

................................................................................................258

Appendix Q:

Q UESTIONNAIRE FOR DISTRICT F IN A N C IA L MANAGERS .. 262

Appendix R:

QUESTIONNAIRE FOR HOSPITAL F IN A N C IA L MANAGERS . 279

Appendix S:

QUESTIONNAIRE FOR D H M T MEMBERS

.....................................284

B IB L IO G R A P H Y ................................................................................................................ 311 CURRICULUM V IT A E ......................................................................................................322

xx


LIST OF TABLES

Table 5.4.1: Perceived Disease Problems and Priority Categories .................................................69 Table 5.4.2: Reasons Why Respondents Assigned a Problem to First Priority C ategory

71

Table 5.4.3: Health Managers Perceptions o f Public Concern on Selected Items ....................... 76 Table 5.4.4: Respondents Perception o f Important Causes o f their Job Satisfaction ................... 77 Table 5.4.5: Health Support Services Perceived as P rio ritie s ..........................................................81 Table 5.4.6: Respondents Priority Categorization o f Selected Health Services A ttribu te s Table 5.4.7: Main Decisions for Resource Allocation

82

.................................................................... 84

Table 5.4.8: Issues Influencing Allocation Decisions and Possible Decisions/Actions to Correct T h e m ........................................................................................................................................91 Table 6.4.1: Summary Statistics - Per Capita Allocations for Non-salary District PHC Recurrent Budget for 1997/98 and 1998/99 Financial Years in 10 Ugandan Districts by Broad Budget C a teg o ry................................................................................................................... 102 Table 6.4.2: Summary Statistics - Per Capita Allocation for Personnel Remuneration in 1997/98 District Health Budgets.........................................................................................................103 Table 6.4.3: Summary Statistics - Total Per Capita Allocations for PHC and Hospital Services in 1997/98 District Health Budgets ........................................................................................104 Table 6.4.4: Summary Statistics - Per Capita Allocation for Selected Budget Items for 1997/98 and 1998/99 ......................................................................................................................... 106 Table 6.4.5: Summary Statistics for Allocations for Routine Immunizations per Child < 1 year for 1997/98 and 1998/99 Financial Years .........................................................................107 Table 6.4.6: Summary Statistics for Allocations for National Immunization Days/Polio Campaign per Child < 5 years for 1997/98 and 1998/99 Financial Years .......................................107 Table 6.4.7: Proportion o f Total Non-salary Health Budget Allocated for PHC and Selected District Ind ica tors................................................................................................................. 108 1. Summary S ta tistics...........................................................................................................108 2. Correlation Coefficients and Significance Levels ........................................................108 Table 6.4.8: Proportion o f Non-salary Health Recurrent Budget Allocated for PHC and Selected District In d ica to rs................................................................................................................. 109 1. Summary S ta tistics ........................................................................................................ 109

xxi


2. Correlation Coefficients and Significance Levels .......................................................110 Table 6.4.9: District per Capita Allocations for PHC Non-salary Recurrent Budget and Selected District In d ica to rs.................................................................................................................110 1. Summary Statistics ......................................................................................................110 2. Correlation Coefficients and Significance Levels ........................................................I l l Table 6.4.10: Non-salary Recurrent Budget Allocated for Integrated PHC Service Delivery and Selected District Indicators............................................................................................... 112 1. Summary S ta tistics .......................................................................................................... 112 2. Correlation Coefficients and Significance Levels ........................................................112 Table 6.4.11: - Per Capita Allocation for HIV/Reproductive Health and Selected District Indicators ............................................................................................................................. 113 1. Summary S ta tistics.......................................................................................................... 113 2. Correlation Coefficients and Significance Levels ........................................................113 Table 6.4.12: Per Capita Funds Released in 1997/98 from all Sources and Selected District Indicators ............................................................................................................................. 114 1. Summary S ta tis tic s .......................................................................................................... 114 2. Correlation Coefficients and Significance Levels ........................................................114 Table 6.4.13: National Disbursements o f Public Financial Resources by Broad Budget Centre Categories for 1997/98 Financial Y e a r...............................................................................115 Table 6.4.14: Criteria Respondents Would Use in Allocating Resources to Health Programmed 17 Table 6.4.15: Current Problems. Degree o f Concern. Efforts Being Made and Action and Information Needed ............................................................................................................ 124 Table 7.4.1: Life Expectancy and Selected District In d ica to rs......................................................143 1. Summary S ta tis tic s .......................................................................................................... 143 2. Correlation Coefficients and Significance Levels ........................................................143 3. A N O V A ........................................................................................................................... 143 Table 7.4.2: Out-patient Department New Cases and Selected District Indicators..................... 145 1. Summary S ta tistics.......................................................................................................... 145 2. Correlation Coefficients and Significance Levels ........................................................145 3. A N O V A ........................................................................................................................... 145 Table 7.4.3: Ante-natal New Cases and Selected District Indicators ...........................................147 1. Summary S ta tis tic s .......................................................................................................... 147 2. Correlation Coefficients and Significance Levels ........................................................147 3. A N O V A ........................................................................................................................... 147 Table 7.4.4: Summary o f Models .................................................................................................... 149

x x ii


Table 7.4.5: Health Managers’ Perception of Adequacy of Salaries and Wages for Staff at Three Levels o f the District Health System ............................................................................... 158 Table 7.4.6: Health Managers’ Perception o f Regularity o f Salaries and Wages for Staff at Three Levels o f the District Health System ................................................................................. 159 Table 7.4.7: Reasons Mentioned as Constraints to Doing Supervision to PHC Units and H o sp ita ls................................................................................................................................ 162

LIST OF FIGURES AND BOXES

Box 1.1: Strategies for health reforms in Uganda.................................................................................7 Figure 3.1: Conceptual Framework o f a District Health System in U g a nd a .................................. 34 Figure 3.2: Simplified Decision-Tree Model fo r Theoretical Public Health Resource Allocation Decisions in Ugandan Districts ........................................................................................... 36 Figure 4.1: Map o f Uganda Showing Study Districts ...................................................................... 40

x x iii


LIST OF ABBREVIATIONS

AG M

- Annual General Meeting (o f the M inistry o f Health).

AHA

- Area Health Authority (Britain).

AIDS

- Acquired Immune Deficiency Syndrome.

ANC

- Ante-Natal Care.

AR I

- Acute Respiratory Infection.

BPF

- Best Practice Frontier.

BOD

- Burden o f Disease.

CAO

- Chief Administrative Officer.

CBA

- Cost-Benefit Analysis.

CEA

- Cost-Effectiveness Analysis.

CFO

- Chief Financial Officer.

CPI

- Consumer Price Index.

D,

- Phase one decentralization districts (1993/94 Financial Year)..

D:

- Phase two decentralization districts (1994/95 Financial Year).

D,

- Phase three decentralization districts (1995/96 Financial Year).

D A ID

- District Assistant Inspector o f Drugs.

DALY

- Disability Adjusted Life-Year.

D A N ID A

- Danish Agency for Development Cooperation.

DDHS

- District Director o f Health Services

DEA

- Data Envelopment Analysis.

DFC

- District Finance Committee.

DGHS

- Director General o f Health Services (topmost technical head o f MOH).

DHC

- District Health Committee.

DHE

- District Health Educator.

DHI

- District Health Inspector.

D H M T - District Health Management Team (all technical heads o f health sections). DHO

- District Health Office.

DHSP

- District Health Services Project (o f the MOH).

DHT

- District Health Team .

DTBLS

- District Tuberculosis and Leprosy Supervisor. x x iv


DHP

- District Health Plan.

DHV

- District Health Visitor.

DMO

- District Medical Officer.

DMU

- Dispensary Maternity Unit

DSC

- District Service Commission.

DSSC

- District Social Services Committee.

DW D

- Department o f Water Development.

EPI

- Expanded Programme on Immunization.

FP

- Family Planning.

GDP

- Gross Domestic Product.

GNP

- Gross National Product.

GPC

- General Purpose Committee.

FGD

- Focus Group Discussion.

HDI

- Human Development Index.

H E A LY

- Healthy Life Year.

HEO

- Higher Executive Officer.

H IV

- Human Immunodeficiency Virus.

HMIS

- Health Management Information System.

HSD

- Health Sub-District.

IEC

- Information Education Communication (usually refers to health education).

IM C I

- Integrated Management o f Childhood Illness.

IM R

- Infant M ortality Rate.

IPH

- Institute o f Public Health (Makerere University - Uganda).

LC I

- Local Council One (village council).

LC II

- Local Council Two (parish council).

LC III

- Local Council Three (sub-county council).

LC IV

- Local Council Four (county council, now defunct).

LC V

- Local Council Five (district council, same as LDC).

LDC

- Local District Council.

M CH

- Maternal and Child Health.

MFEP

- M inistry o f Finance and Economic Planning.

MPH

- Master o f Public Health.

XXV


MOH

- M inistry o f Health.

M O LG - M inistry o f Local Government. MMR

- Maternal M ortality Rate.

MS

- Medical Superintendent.

NGO

- Non-Governmental Organization.

NHS

- National Health Service (Britain).

NIDS

- National Immunization Days (aimed at polio eradication).

OPD

- Out-Patient Department.

PHC

-

PFP

- Private for Profit.

PNFP

- Private not for Profit.

PPP

- Purchasing Power Parity.

PSC

-

Public Service Commission.

P Y LL

-

Potential Years o f L ife Lost.

Q A LY

-

Quality Adjusted L ife Year.

QA

- Quality Assurance.

QAC

Primary Health Care.

- Quality Assurance Committee (o f the MOH).

RAWP - Resource Allocation W orking Party (Britain). SAHRA

- South African Health Resource Allocation.

SEU

-

SMR

- Standardized M ortality Ratio.

STD

- Sexually Transmitted Disease.

STI

- Sexually Transmitted Infections.

STIP

- Sexually Transmitted Infections Project (o f the MOH).

TB

- Tuberculosis.

TBA

- Traditional Birth Attendant.

UNDP

Subjective Expected U tility.

- United Nations Development Programme.

UNEPI - Uganda National Expanded Programme on Immunization. UNFPA

- United Nations Fund for Population Activities.

UPA

- Under Priviledged Area (Britain).

UNICEF

- United Nations Children’ s Fund.

WDR

- W orld Development Report. xxvi


WES

- Water and Environmental Sanitation.

W HO

- W orld Health Organization.

W IC

- Women, Infants and Children.

x x v ii


OPERATIONAL DEFINITION OF TERMS

Access - the proportion o f people in a defined geographical area who are within reasonable reach o f a service (e.g., walking distance, or within 5 kilometer radius). Access

= those who can reach and use a service people living in an area

Allocate

- make available resources for a defined purpose, person or place.

Coverage - proportion o f people who would benefit from a service Coverage

who actually get it.

= people receiving a service____ people who would benefit from a service

Decentralization

- the transfer o f power for decision making over certain aspects o f government from central government to sub-national, semi-autonomous units o f governance (also called local authorities) e.g., districts, municipalities, sub-counties, etc depending on legal provisions.

Discretionary budget

- an item over which the local managers have authority to decide allocations, amounts, etc. Examples o f discretionary items in Ugandan budgets were IEC activities and health education, routine immunizations, etc. Non-discretionary items included salary and lunch allowance, polio eradication and capital projects.

Disposable income

- the amount o f money left-over to spend according to one’ s wishes after mandatory payments have been made. For districts these mandatory commitments include paying salaries, paying suppliers or other contractors. Disposable income is the amount they could spend on service improvement.

xxvitt


Effectiveness

- the extent to which intended results are achieved under real field conditions or according to set objectives.

Efficacy

- the extent to which an intervention works under ideal or experimental conditions.

Efficiency

- the ability to produce results from given inputs in a way that is productive e.g., minim izing waste, achieving the intended results at least cost, or achieving the most optimal output for a given unit o f resources.

Exclusive benefit

- economic term referring to benefit from a service that accrues only to the individual members participating or investing in it. The application to health is controversial.

Equity

- the distribution o f benefits o f a service in a reasonable and fair manner. There are different interpretations o f the word “ fa ir" and may include in terms o f inputs (budget), proportion covered, proportion o f need that is met. or equalization o f residual need.

Improved health

- the word improved is used in this study to refer to a better health status as objectively determined using a measure that takes into account mortality and morbidity, such as the D A L Y or HE A LY .

Public

- the term public is used to refer to those services that are funded from government or other community organized sources or through charitable activities o f NGO and donor sources. Charitable refers to a primarily non-profit intention.

Private-for-Profit

- refers to a service offered as an income generating business with the primary intention o f making profit, e.g., a private clinic.

Private-not-for-Profit

- refers to institutions that are not government owned or controlled but

x x ix


that are prim arily intended to serve the communities in which they are situated. The primary motivation is service and not profit-making. Rationing

- the term refers to the fixed amount o f a resource that may be used in a period, place or by a person. Not considered in this study.

Resources

- the means available or that can be drawn on for doing something, usually reference is made to resources such as human, financial, material and time. In this study focus was especially on finance.

Rival benefit

- again an economic term referring to services or investments in which one person's benefit lessens the benefit accruing to another person. Applications to health are controversial.

XXX


PART I: M ETHODOLOGICAL APPROACH

CHAPTER 1: INTRODUCTION

CHAPTER 2:LITERATURE REVIEW

CHAPTER 3:CONCEPTUAL FRAMEWORK

CHAPTER 4: METHODOLOGY


CHAPTER 1: INTRODUCTION

This chapter provides a general introduction to this case study o f resource allocation decision-making at district level in Uganda. It is a view o f the ever challenging economic problem o f allocation o f scarce resources in the face o f competing needs which is a problem affecting most developing nations. Uganda is one o f those countries attempting to restructure it's economy through introduction o f market reforms, decentralization and democratization. Uganda attained independence from Britain in 1962 and enjoyed relative peace and prosperity till the coup o f 1971 ushered in a m ilitary dictatorship. A fter many years o f political turmoil and economic collapse o f the 1970s and 1980s. the health services which had been o f reasonably good quality and were provided free to all citizens went into a state o f dysfunction (Okuonzi and Macrae, 1995). Due to the new economic policies o f cutting government subsidies and structural adjustment, government resources available to health became even more scarce. The article by Okuonzi and Macrae (1995) provides an excellent description o f the political and health policy development situation in pre- and post independence Uganda and that description w ill not be repeated here. It is in this context that an attempt is being made to re-build the health services. The chapter provides a description o f the problem under study, relevance o f the study, a brief background on health reforms in Uganda and study objectives.

1.1 Statement of the Problem Since the Uganda Government started a radical process o f decentralization in 1993 the provision o f all social services previously provided by central government became a responsibility o f district local governments. District health managers in Uganda have been expressing the need for guidance in their resource allocation and budgeting for health activities since decentralization o f health services started. Since then, there have been attempts at providing guidelines from the central


government ministries responsible for health, local government and finance. However, these guidelines have often been criticized by the district managers as unrealistic or as not very practical and so are sometimes not followed (M inistry o f Health [Uganda], 1994). It is generally known that after decentralization the flow o f money to the districts improved substantially. Although still inadequate to properly meet the needs o f health care provision in the districts, this marked a significant improvement in the financial situation o f health services at district level. There continued to be concern among health workers and the general public that the quality o f services was not improving as expected. Reviews by the M inistry o f Health (M O H ) indicated that most o f the money that had been sent to the districts after decentralization started was mainly going to hospitals with little additional resources going to first contact health care or promotion/prevention activities (Mwebesa, 1996). Health staff working in the predominantly primary health care oriented rural health services have often expressed their concerns that hospitals seem to be taking most o f the money and little is left for the activities in the rural areas (Guma. 1996). Even though resources are scarce, it was considered that the country was generally not getting value for money being spent on health services. The problem was largely attributable to poor decision-making in allocation o f resources (Omaswa, 1994).

1.2 Study Justification It was not known or documented how decisions about allocation o f funds to different health interventions were being made at the district level. Anecdotal observations at the beginning o f decentralization had suggested that in some districts more resources may have been going to the hospital while less was going to PHC (M inistry o f Health [Uganda], 1994). There was a need to understand the reasons for this especially in the face o f a clear policy to emphasize PHC. The central government makes allocations for health to districts based on a number o f factors such as population,

3


presence o f donor projects, etc. However, in the spirit o f decentralization, districts are supposed to allocate the health resources to different interventions and activities themselves based on their priorities (M inistry o f Health [Uganda], 1998). To date, no systematic study on the issue o f resource allocation decision-making had been conducted in Uganda. The closest approximation to a study on resource allocation was a W orld Bank initiated study conducted in Uganda. Kenya, Tanzania and Ethiopia, which estimated Disability Adjusted Life Years (D A LY s) for the top ten conditions (The W orld Bank. 1994). It seems not to have produced much guidance for district health managers as to how resources should be allocated but was mostly aimed at national level policy makers (Luwaga. 1996; Makumbi, 1996; Mudusu, 1996). The functioning o f health systems in developing countries in a decentralized setup has not been much studied and there is an urgent need to study the effects o f the widespread decentralization and health sector reforms that are going on in many developing countries ( W HO. 1988; M ills. Vaughan et al {Editors}. 1990; Collins and Green, 1994; Kutzin. 1995). It was these problems that prompted this study on resource allocation decision-making with interest in how decisions are being made and why. which decisions are being made, which criteria are being used and why. and how the process could be improved e.g., the potential for and use o f a value-for-money or burden o f disease and cost-effectiveness approach (D A LY s or HEALYs for the resources spent). This study was intended to collect information upon which policy makers and district health managers (Appendices B, C and D) could base future decisions to improve resource allocation, using currently available information and practical, not-so-sophisticated skills. It is expected that the results o f this study w ill make a contribution to the understanding o f how resource allocation decisions are being made within the health sector at district level in Uganda. Policies could be shaped upon findings o f this research affecting developmental activities like training, building new facilities, staff expansion, etc or recurrent ones like purchase o f drugs and

4


payment o f salaries and incentives. In this way, the study was a contribution to the knowledge base for improvement o f the management o f scarce health resources in Uganda.

1.3 Background Health Situation The issue o f how best to allocate health resources is made especially complex by a variety o f factors. From the perspective o f improving health status, which should ideally be the underlying objective o f health resource allocation, the country, like many others in Sub-Saharan Africa still has a long way to go (The W orld Bank. 1994). The following data from UNICEF's "The State o f the World's Children 1999" (UNICEF. 1999) illustrates how bad the situation is: In 1997, infant mortality rate (IM R ) was estimated at 86 per 1.000 live births, under five mortality rate at 137 per 1,000 live births; in the period 1980 to 1997, maternal mortality ratio (M M R ) was reported at 510 per 100.000 live births. The annual populationgrowthrate was estimated at 3.2% for the period 1990 to 1997 (UNICEF. 1999). The health needs are many and mainly due to infectious diseases like malaria, acute respiratory infections (ARI), diarrheal diseases, and HIV /AID S . Life expectancy at birth o f 41 years is one o f the lowest in the world, the situation having been aggravated by the H IV /A ID S epidemic, with Uganda having one o f the highest seroprevalence rates in the world, estimated at about 7.5% for the population in 1996 (UNICEF-Uganda, 1998). In 1990, among blood donors the infection rate had been reported for women aged 15 to 21 years as being between 25% to 32% in some surveillance sites (M inistry o f Health [Uganda], 1993). Although the W orld bank estimates that countries in Sub-Saharan A frica need to spend at least US $ 12 per capita to assure a minimum "basic health care package", Uganda in 1991 spent only S 5.65', half o f which was from government and the other half from private sources. Non-

i This is un-adjusied for Purchasing Power Parity (PPPS). which allows for comparisons based on actual purchasing power in different countries.

5


Governmental Organizations (NGOs) and donor agencies (The World Bank, 1993). The country is generally quite poor, relying mostly on export o f coffee for foreign exchange earnings. In 1996 GDP per capita was estimated at 305.825 Uganda Shillings2 (M inistry o f Planning and Economic Development [Uganda], 1997). In 1997 it was at 322,996 shillings but the dollar value had fallen to about US S269 due to appreciation o f the dollar against the shilling1(M inistry o f Finance. Planning and Economic Development [Uganda], 1998). It was estimated that in 1991 Uganda spent about 19% o f GDP on health, 11% o f which was from external aid and 8% from domestic revenues (M inistry o f Health o f Uganda, 1993). More recent estimates suggest that total expenditure on health was 6.1 c/r o f total government expenditure in 1997/98 (UNICEF - Uganda, 1998). The Government o f Uganda recognizes the need to reallocate resources in support o f PHC and has a declared intent o f re-orienting all health services to support PHC. Under the heading "major strategies", the Government White Paper on Health Update and Review (1993) listed the strategies shown in Box 1.1.

Equivalent to about US $ 278 per capita. Again this is not PPPS-adjusted.

This has serious implications since most supply inputs for health such as drugs and other pharmaceutical supplies often have to be imported using US .$.


Box 1.1: Strategies for health reforms in Uganda. (Source: White Paper on Health Policy Update and Review 1993, M inistry o f Health. Uganda)

| "A ) Resource mobilization through the central government budget.... j

i

!B) Alternative/additional financing mechanisms.... j

j C) Shift resource allocation toward the most cost-effective public health and clinical interventions.... D) Effectively implement the decentralization programme in the health sector.... |E) Restoring the functional capacity o f and improving the efficiency o f existing government! ;facilities.... IF) Facilitating greater role for NGOs, private sector, and communities.... jG) Capacity building in the health sector...."

;

Studies done as to why people seek care in private facilities rather than government ones have consistently found that people are dissatisfied with the lack o f drugs, long waiting times, inconvenient hours o f operation, poor interpersonal communications by health workers, and bribery and corruption associated with government health services (Okeilo, Konde-Lule et al. 1997). Improvement in quality o f services was one o f the main aims underpinning decentralization o f all central government services to district level. Quality, efficiency and equity are the 3 main policy goals o f the M inistry o f Health o f Uganda (M inistry o f Health [Uganda), 1993). In the new health policy o f 1999, the M inistry o f Health (M O H ) is proposing the creation o f Health Sub-Districts (HSD) to become the main level for implementation o f health programmes and activities in the spirit o f further decentralization o f health services. This is meant to free the currently over-burden district health authorities to better plan for and co-ordinate service delivery

7


rather than attempt to do actual implementation. It is also aimed at promoting equity and bringing services and personnel nearer to where people live (M OH, 1998; MOH, 1999; Kiyonga, 1999). In order to address the problem o f verticalized delivery o f health services, the M inistry o f Health is introducing what it calls the Sector Wide Approach (SWAP). In this policy, districts w ill make comprehensive health plans and budgets and all resources from government, donors and local district revenue pooled into a “ common basket” (Kiyonga, 1999).

1.4 Scope of the Study In this study, the term “ resources" was used in reference to finances since all the others e.g.. personnel, drugs and supplies, equipment, buildings, etc can ultimately be expressed in monetary terms, and finances give us a common means o f estimating value that is placed on different health interventions and priorities in real terms. Focus was on public resources rather than the private sector even though the two influence each other. This is mainly due to issues o f practicality within the scope o f this study. Public resources for health were taken to include funds from central government and donors to districts. This money goes in the form o f an un-declared share o f the block grants that districts are supposed to spend on health, as well as PHC conditional grants allocated specifically to be spent on health. In addition districts are supposed to spend some o f their own locally generated revenue on health. Also included were resources from UN agencies (UNICEF, etc) and other donors that provide funds in the context o f bilateral or multi-lateral collaboration between the Uganda Government and foreign governments or other international organizations as long as they appeared in the health plans and budgets o f the District Health O ffice (DHO). Resources from other sources such as foundations or Non-Governmental Organizations that were planned for together with public health officials and appeared in the district health plans and budgets were also considered public resources. Similarly, concern was mainly with resources for health recurrent budgets which were

8


analyzed in more detail than capital development uses. Uses for new buildings, or purchase o f new vehicles, were noted where allocations for such capital projects was determined by the district decision-makers themselves and appeared in the district health plans and budgets. This is because at the moment only recurrent budgets from central government have been decentralized and districts still do not have much control over allocation to capital projects except if they decide to divert or save part o f their recurrent funds (M inistry o f Health [Uganda]. 1998). Secondly, the main obstacle to expansion o f Primary Health Care (PHC) services and improvement o f health status is usually the availability o f recurrent budget funds (McIntyre. Bourne et al. 1991). The study was a preliminary and exploratory one and therefore health-related activities such as water and sanitation were not analyzed in detail. This was done in order to first focus attention on what is happening within the health sector itself. Another reason is that expenditures on such activities as water and sanitation are within the area o f jurisdiction o f another body, the Department o f Water Development (DW D). This exclusion is consistent with other resource allocation studies internationally (McIntyre. Bourne et al. 1991). The allocation from centre to districts and by districts between different sectors although o f interest as well, were not the focus o f this study. It is stated that the health services should be re oriented to provide quality services in an efficient and equitable manner. The focus was on how resource allocation decision-making could be improved given the present situation o f information, knowledge and skills availability or lack there-of in Uganda, and the severe shortage o f public health personnel and other resources. Allocation procedures currently being implemented were studied with a view to pave the way for better allocations (allocative efficiency, equity) in the spirit o f continuous quality improvement. The actual link between use o f resources and improvement in health status, though the desired end result o f any resource allocation (Vargas-Lagos, 1991), was beyond the scope o f this study and was not covered.

9


It should be noted that although mention w ill be made o f various M inistry o f Health policy initiatives and decentralization in general, a comprehensive review o f these policy goals and how they are affected by resource allocation decisions was beyond the scope o f this study. It should also be mentioned that although there were some questions guiding the research process and data collection and analysis, this was a case study aimed at documenting and generating hypotheses and not a hypothesis testing one.

1.5 Research Objectives and Questions The overall aim o f the study was to provide information for district and national health managers and resource allocators as to what are the main influences on resource allocation in the districts. This information could be used to design policy interventions that w ill promote resource allocation decisions that improve the health o f the people. The study aimed to achieve the following purposes by the end o f December 1998:

1.5.1 Goal:

To review the process of decision-making for allocation o f public health resources in decentralized districts in Uganda in order to provide a basis for development of allocation procedures for district health systems that address health needs o f the people and promotes rational use o f scarce resources.

1.5.2 Objectives: 1.

Identify the stakeholders in the allocation process and the role they play in allocation decisions.

2.

Identify the formal guidelines/criteria that are currently existing in the districts with regard to allocation o f public funds to health activities, i f any, and the informal influences that play a role.

10


3.

Determine the values and preferences held by the decision-makers in regard to publicly funded health services.

4.

Assess the extent to which decisions being made in allocation o f public health funds relate to the health problems perceived as priorities (e.g., as defined by the decision-makers' perceptions and based on national guidelines such as the national basic health services package) as well as reasons for any observed deviations.

5.

Based on the findings o f the study, outline a resource allocation decision-making model o f the process that is observed in Ugandan districts, and suggest ways to improve it.

1.5.3 Guiding Questions: Main Questions:

Although designed as a case study and not a hypothesis testing one, there were some guiding questions that influenced the research process. The main ones were: How are decisions about priorities for allocating public financial resources being made in decentralized district health systems in Uganda? How could the process be improved? Following from these two main ones, other questions were the follow ing:

Question 1:

Are decisions being made for resource allocation between health activities based on health needs expressed from the communities (e.g., through their leaders or through utilization o f services) or are they a result o f follow ing existing bureaucratic practices that were there from the past? A desirable situation would be to allocate resources for those problems perceived as being important by the people as well as identified on technical grounds as real needs. The study therefore looked out for evidence o f decision-making for resource allocation based on

health needs

perceived to be expressed by the people as well as recognized on the basis o f

11


objective technical criteria (use o f information). Question 2:

Is the technical health planning and resource allocation process based on health outcomes or health status measures o f the population? We would wish to see decision-making for resource allocation that is in line not only with people’ s expressed needs stated above, but also based on professionally identified health needs in the district. Such technical planning process should aim to improve health status o f the population through appropriate use o f the scarce resources available to the society. A means o f measuring health status is use o f composite indicators that combine mortality and morbidity measures. The study explored the potential for developing such measures at district level.

Question 3:

What share o f the decentralized health budget is being allocated to Primary Health Care (PHC) and how much to hospitals? Do personal factors affect the decision maker's u tility for the outcomes o f resource allocation decisions, e.g. the hospitals taking precedence over PHC activities? It was thought that this situation is a possibility because hospitals are likely to make a stronger case and their needs are more readily visible to district-level decision-makers than the predominantly rural based PHC units and activities.

PHC activities need to be allocated a reasonable portion o f the health budget that enables implementation o f population based health promotion, preventive, curative and rehabilitative interventions. In developing countries hospitals typically take between 40% to 80% o f the health recurrent budget. A cut-off point o f 40% was considered for recurrent non-salary budget similar to a level set by the M O H in Papua New Guinea (The W orld Bank, 1993). In this study an attempt was made to estimate the current levels o f resource allocation for PHC and hospitals in order to serve as

12


benchmarks for future district assessments and comparisons. The results, discussions and conclusions are reported in 3 thematic papers under Part II. Paper 1 which describes the current decision-making process presents findings and issues related to Objectives 1, 2. 3 and 5. Paper 2 presents findings relevant to Objectives 3.4 and 5. It deals with the ways in which the decision-making process for allocation o f resources could be improved, particularly what kind o f information is already available and what is still needed. Findings and issues related to Objective 5 as well as broader contextual issues in Ugandan districts are addressed in Paper 3. It addresses the issue o f what kind o f organizational, structural, and capacity improvements are needed for a better allocation decision-making process.

13


CHAPTER 2: LITERATURE REVIEW

This chapter presents a summary o f the relevant literature that was found. As w ill be seen, little literature o f relevance to the situation in Sub-Saharan A frica was found. The closest to the Ugandan situation is what was found on the South African Health Resource Allocation (SAHRA) based on the British Resource Allocation Working Party (RAW P). Hence these latter two methods are mentioned in a bit more detail than the others. The review covers health sector reforms, health resource allocation, patient and community perspectives, decision-making perspectives, and a summary o f the main relevant issues in organizational and management theory.

2.1 Health Sector Reforms There is growing concern about the influence o f resource allocations and payments on quality o f care not only in developing countries but also in the developed countries where the health share o f the national budget has been rising (Hahn and Lefkowitz, 1992). W ith the demographic transition taking place and general aging o f the population, health care expenditures are likely to continue rising for the foreseeable future (Mosley et al, 1990). There is a need for reforms in public policy, administrative and financial management in the health sector if services o f reasonable quality are to be provided and maintained. Increasing economic hardships make it d ifficu lt to continue providing "free" health services (Lewis. 1988; Leighton, 1995). In an extensive review o f health systems in Africa, a study concluded that a lot o f improvement in health can be achieved, despite tight financial constraints. The key is to reform health care systems and use available human and financial resources more productively. As an example The W orld Bank panel quotes the extent o f inefficiencies in the procurement, storage, prescribing and use o f drugs as being so bad in some countries that only about $12 worth o f drugs for every $100 is actually received by the consumers!

14


(The W orld bank, 1994). Given the increasing budget constraints, it has become even more important that resource allocations are made in such a way that they produce the most optimal health outcomes. The issue that arises is then how to quantify the effectiveness o f interventions, their costs, and link to improvements in the population's health status (Vargas-Lagos, 1991). It is possible to cut costs and improve quality at the same time through reducing inappropriate utilization o f resources and services and avoiding preventable adverse effects (Brown. Franco et al, 1991; Chassin. 1996). Considering that 60-80% o f the health budget is spent on personnel (Vargas-Lagos, 1991). such re-allocation o f resources w ill have to address the question o f staff distribution within the health system. In the Ugandan situation although the country continues to train health professionals, mainly physicians and nurses, most o f these are concentrated in urban areas, offering mainly curative based services. In Zimbabwe, after independence there was a significant expansion o f primary health care services with new health centers being built and reallocation o f resources to support them. However, personnel to staff these facilities remained a major problem since most well trained professionals tend to stay in the urban areas or jo in the private sector, serving the health needs o f a very small proportion o f the higher income population whose health needs are less (Loewenson, Sanders et al, 1991). How to allocate personnel in such a way that the objectives o f need-based services and equity as implied in the PHC approach are to be met is increasingly a matter o f major policy significance. As noted earlier in the section on the background health situation in Uganda, the country is also undertaking health sector reforms through the strategies o f further decentralization to Health Sub-Districts (HSD) and integration o f service delivery through the Sector Wide Approach (SWAP).

15


2.2 Allocations for Health One o f the earliest attempts in guiding resource allocation between competing priorities and health interventions was by the Ghana Health Assessment Project Team (Ghana Health Assessment Project Team. 1981; Morrow, 1984). They used mortality and morbidity measures to come up with a composite measure called the Healthy Life Years (H E A LY ). The number o f HEALYs lost from different diseases was calculated as well as the cost o f gaining HEALYs using different interventions. For each, disease interventions were then ranked according to the most cost-effective. The H E A LY does not use age-weighting as does the D A LY . Chang et al (1996) give the example o f a cost-effectiveness analysis (CEA) using qualityadjusted life years (QALYs). The intervention that is chosen in preference to another is then the one that yields the greatest number o f QALYs. Q ALYs have been advocated by some and criticized by others for their utilitarian approach and tendency to further disadvantage the less fortunate, producing a lively debate in the medical ethics literature (Cubbon, 1991;Harris. 1991;Black, 1991; Caiman, 1994; Singer, McKie et al, 1995; Harris, 1995; Barker and Green. 1996). The main issue seems to be centered around the tendency o f Q ALYs to produce greater inequity - favoring through preferential allocations, health programs that benefit the most "advantaged" for whom a year o f life would be considered o f more utility than someone say with a disability. In trying to address concerns that Q ALYs raised, others have used disability-adjusted life years (D A LY s) instead o f Q ALYs (The W orld Bank. 1993; Murray and Lopez {editors}, 1994). D A LY s and cost-effectiveness have in turn come under criticism for use o f discounting for life years and age-weighting and the problem of equity in a setting o f budget constraints (Ubel, DeKay et al. 1996; Barker and Green, 1996). The proponents for their part say they are just applying society's own views in valuing life (Murray and Lopez, 1994). The underlying principle behind HEALYs, Q ALYs, D ALYs, and other Burden of Disease (BO D) measures is the need to use one indicator that combines measures o f morbidity with

16


measures o f mortality using time and life expectancy as the common unit in order to allow for comparison o f disease states and interventions. The similarities and differences between HEALYs and D A LYs have recently been demonstrated (Hyder, Rotllant and Morrow, 1998). However, concepts such as DALYs, HEALYs, although appealing from a technical point o f view, are d ifficu lt to explain to benefisciaries and their representatives, the politicians. Such explanation is necessary to win the political stage o f resource allocation. The Oregon Health Plan attempted to use cost-benefit analysis in which community perspectives, expert opinions and costs were taken into consideration in constructing a list o f priority interventions to be funded under M ED IC AID . The process was put on hold because the Federal Government refused to grant a waiver due to some legal and political considerations. In particular it was thought that the Oregon proposals went against the “ Americans with Disabilities A ct” (Blumstein, 1997). Some people and states have used quantitative mathematical modeling for determining resource allocation (Tingley and Liebman, 1984; Rothbard. 1987; Eyles and Birch etal. 1991; Eyles and Birch et al, 1994; Pampalon and Saucier et al. 1996; Warburton. 1997). An example o f quantitative modeling is what was done in Indiana for state-level resource allocation for the Women. Infants and Children (W IC) program. They developed a model formulated as a linear integer goal program, using different service levels for categories o f W IC participants. In this approach the decision-maker is able to adjust the relative importance o f each goal, thus allowing for subjective judgement o f the decision-makers, which is often based on experience and intimate knowledge o f the program. The service level o f each priority group across the state was treated as a separate goal, with a specified target value and individual weights. The weights acted as penalties for deviating from the target. The allocation process was then modeled as a series o f decisions o f whether to fund or not to fund each priority group, subject to budget limitations. In this modeling sensitivity analysis

17


helps to test a range o f values decided on from the subjective opinion o f the decision-maker (Tingley and Liebman, 1984). Another example is the technique known as Data Envelopment Analysis (DEA), another name o f which is Best Practice Frontier (BPF), applied by Rothbard (1987) to assess allocations between different mental health programs in the State o f Pennsylvania. The technique essentially compares levels o f efficiency and productivity for similar units, production centers or programs, based on the achievement o f the best unit as a benchmark. In England, Hutchinson et al (1989) report an assessment o f two approaches used in identifying deprived populations in order to decide on additional Financial allocations to general practitioners under contract. The first method used an under-privileged area score scale and another used a material deprivation score scale. McIntyre, Taylor et al (1990) reviewed the Resource Allocation W orking Party (RAWP) formula-based method used in England for its applicability to South Africa, and other developing countries, with a view to effect re-allocation o f healthcare resources. This RAWP formula approach is favored because o f it ’ s non-dependance on demand or supply o f health services. Supply and demand o f health services such as use o f existing facilities or utilization levels may be reflective o f past and present inequities in allocation o f health care resources. These are derived from on the one hand costly capital intensive facilities and on the other, demand related to the supply o f facilities and personnel as well as willingness and a b ility to pay for services. In England RAW P was initially commissioned to review the arrangements for resource distribution within the National Health Service (NHS). and Area Health Authorities (A H A s) as well as districts. The aim was to achieve a pattern o f distribution that was objectively responsive, in an equitable and efficient way, to relative need in order to achieve “ equal opportunity o f access to health care for people at equal risk” . The RAWP objectives were largely met after a period o f 10 years, reducing the difference in expenditure per capita between the poorest and wealthiest regions from about 30% to less than 10% by 1990

18


(McIntyre, Taylor et al, 1990). The RAW P formula uses regional population, weighted by utilization adjusted forage and sex, and using standardized mortality ratio (SMR) as a proxy for morbidity. The SMRs were intended as a measure o f relative geographical risk or need rather than as a measure o f absolute morbidity, for which they have been criticized (McIntyre, Taylor etal, 1990). After review, the contribution o f SMRs in determining resource allocation targets was reduced and the remaining weighting made up o f an Under-privileged Area (UPA) score. In adopting the RAWP formula for use in South Africa, the lim itation o f data that would be available was recognized. The South African Health Resource Allocation (SAHRA), though based on the same philosophy as the RAWP, proposed to use only two service categories broadly described as curative and preventive whereas RAW P used six service categories to weight the population (Bourne. Pick et al, 1990). National hospital utilization rates were used to weight the population and the percentage distribution o f the weighted population for each region determined. The second component o f the SAHRA formula is a preventive one, in which the population is now weighted using Potential Years o f Life Lost (P Y LL). After appropriate adjustments the two weighted populations are combined according to historical financial distribution between the components. Whereas in the review o f RAWP a factor o f 0.44 was used to reduce the effect o f SMR on resource distribution, in the SAHRA formula, this factor was im p licitly set to 1 (Bourne, Pick et al. 1990). In the case o f South A frica it was found that the inequity between regions was so much that the formula was able to identify regions with comparable populations but different health status and redress the inequity. Accurate mortality and morbidity data would only become necessary as resource allocation reaches equity (Bourne, Pick et al, 1990). The first step in applying the SAHRA formula was to determine the total public sector

19


resources available for health care and their current allocation4. Only recurrent expenditure under the control o f the health administrators was considered, leaving out "health-related” activities and capital expenditure. The second step was the estimation o f relative need for health services on a geographical basis by means o f weighting regional populations for proxy morbidity measures5. The target allocation for each region was determined by proportionately allocating the total budget available for health care according to the respective weighted populations. The current allocation o f health care resources was then compared with the target allocations as determined by the SAHRA formula in form o f proportions and deviations were noted (McIntyre, Bourne et al, 1991). In a discussion o f the planning and political implications o f resource allocation for PHC, Segall (1983) makes the argument that the objective o f resource allocation for PHC is to make PHC policy a reality and that resources ought to be used to reduce national inequalities and imbalances in the health sector. He further argues that planning for the health service should be integrated since policies regulating private medical practice affects the health service under direct government control and vice-versa and restructuring may be necessary to conform to PHC priorities. He also states that resource allocation for PHC involves not only technical planning but political considerations. The political consideration arising out o f the fact that PHC implies the allocation of resources based on need and equity as opposed to demand. Such a choice implies political decisions. W hile recognizing the inter-linkage between the private and public sub-sectors o f the health system, this study focused on the public component to keep within scope o f what was feasible. Policies affecting the public component are likely to be easier to implement than those touching on private medical practice. The difficulties involved have recently been outlined in a study o f private medical

4

A similar approach was used in this study.

5 Sim ilar reasoning influenced the use o f Life Expectancy Index as a measure o f health status in Ugandan districts as reported in Paper 3 in this study.

20


practice in Uganda by Okello, Konde-Lule et al (1997).

2.3 Patient and Community Perspectives The health care delivery system should respond to the perceptions o f the population about quality o f health care and quality o f life as well as what is valued from the health services by the public6. In Tanzania, Gilson et al (1994) used qualitative interviewing techniques like focus group discussions to study community satisfaction with PHC services. They found perceived problems o f care available such as drug shortages, interpersonal relations with providers, etc. These are potentially within the capability o f the health sector to improve. Hadom (1991) suggests that public values could play a major role in the health care resource allocation decision. He suggests that preferences be mapped onto information concerning expected outcomes o f specific interventions. Such “ preference-weighted” outcomes could then determine the relative priority given to health services for each specific condition. He, however, points out that concerns about possible discrimination might lim it the applicability o f such an approach, a concern which was at the heart o f the US Federal Government decision not to grant a waiver for the Oregon Health Plan. It was feared that a public valuation o f outcomes might discriminate and go against the “ Americans with Disabilities A ct" (Blumstein, 1997). A common approach to gain a sense o f community preferences for services is to ask people how important they consider different health programs and/or services for either prevention, diagnosis, or treatment o f different health problems. These preferences could be quantified using willingness to pay approaches (Hadom, 1991). In willingness-to-pay a specific situation is described, stating the expected benefit o f the measure, the consequences i f action is not taken, and what alternatives exist. Individuals are then asked how much o f their personal income they would be

6Hcncc the inclusion o f key informant interviews with ordinary citizens that was done in this study.

21


w illin g to pay for the particular benefit, or how much additional tax levied by the government they would be w illin g to pay. Although the approach has had some criticisms, it has become widely used in evaluating different economic projects and in environmental assessments (Portney, 1994). Some o f the criticisms are the failure to take into account the fact that the same amount o f money is not valued equally by different people7, the d iffic u lty o f taking into account expected benefits, risks and costs*. as well as the hypothetical nature o f the questions (Hanemann and Michael. 1994; Diamond, Hausman et al, 1994). For these reasons, the preference-for-service approach has relatively little role to play in solving the problem o f how to allocate health resources according to Hadom (1991). This is the reason, says Hadom, why preference-for-outcomes approach is more favored, since people can more readily judge the outcomes based on their own previous experiences or the experiences o f others in their communities, or can imagine what these outcomes would have on their daily life. Generic measures are recommended in assessing outcome-preferences rather than illnessspecific measures e.g., a statement about "suffering caused by pain” is preferable rather than say “ suffering due to migraine headache” . The denominator o f health outcomes could be a generic descriptor such as “ daily activities" to allow for comparison across different conditions and different people. Such generic measures are especially necessary when considering the problem o f resource allocation9 (Hadom, 1991).

2.4 Decision-Making Young (1981) states that any study on decision-making has to address three main questions,

7

U tility for money is not a linear function. One hundred thousand shillings (about US $100) is not valued equally by an industrialist based in Kampala and a peasant farmer in a rural area o f Arua District in Uganda, for example! sDue to short-comings o f information availability and interpretability.

■> This consideration influenced the design and use o f the scales for eliciting decision-maker perceptions that is reported in paper 1 in this study.

22


namely; what alternatives are available to members in regard to the particular problem under study, which criteria are used in selecting among these alternatives, and what decision processes or principles are used when making a choice. These steps can be summarized in a flow-chart, decision table or decision tree. Young developed and tested such a model for predicting treatment choices among Mexican villagers in Pichareno. Lindley (1985) while not addressing the issue o f resource allocation specifically, presents a theoretical basis upon which all rational decisions are expected to be based. He suggests that in any decision-making there are associated uncertainties o f events and possible outcomes. He proposes a technique to attach probabilities to uncertain events and utilities to possible consequences. The best decision is then that which maximizes the expected u tility. The laws o f probability namely the convexity law10, addition law11 and multiplication law 12 and their consequences like Bayes' Theorem11 are used to ensure coherence. However, calculations o f overall benefit and u tility are in general limited especially in developing countries, by lack o f data. Often the information processing abilities o f decision-makers are far overwhelmed by the problems o f limited attention (since there are many problems to deal with), individual and institutional memory (records get lost, people die or move to other jobs), limited comprehension (rapid technological and scientific advancement).

10

I f you are considering the probability p o f an uncertain event £ happening, the probability you assign w ill depend on the information you have I. Mathem atically written as p(E\I). The convexity law states that this probability can be expressed as a non-negative number that lies between 0 and 1, or 0 i p(E \l) c I. 11

Addition law states that for two exclusive events, £ , and £ ,, on having information /. p (£ , or E :\f) = p (£ ,|/) + p (E :\D or for n exclusive events, p (£ , or £ , or...£„) = p (£ ,) + p {E 2) + ...+ p(£„).

i: M ultiplication law stales as follows: for any events £ , £ , and information I. p ( E , and £ ,|/) = p (£ ,|/) * p (£ ;|£| and t). If we leave out I, p ( E t and £ ,) = p ( E ])*p (E 2\Et). This is useful in calculating the probability o f two events both occurring. 13

I f £ , and E: are any two events, provided £ , is not zero. p (£ ,|£ ,) = p ( E ]\E2) ‘ p ( E 2)/p{El). Applying the multiplication law to £ , and £ ;. p (£ , and £ ,) = p (£ |)* p (£ ,|£ ,). Since (£ , and £ ,) is the same as ( £ : and £ ,). p (£ , and £ ,) = p (E ,)* p (£ ,|£ ; ) from which, p ( £ |) * p ( £ 2|£ ,) = p ( £ ,)* p ( £ i|£ :) - Bayes' Theorem states, i f p ( £ t) is not zero, p (£ ;|£ ,) = p ( £ 2)*p (£ ,[£ ;)/p (£ ,). (Reference: Dennis Lindley (1985); M aking Decisions. Second Edition. John W iley. 1985).

23


problems o f communication, etc (March, 1958; Simon, 1960; Simon, 1976; March, 1985). In a study o f the association between Subjective Expected U tility (SEU) and behavior, Bauman, Fisher etal (1984), report on their research into determinants o f smoking behavior among youth. They found significant associations between SEU and smoking behavior. SEU is said to apply equally well as a predictor o f other forms o f behavior. The technique involves asking, for each o f the listed outcomes expected o f the behavior, what value the subject attaches (how much he/she likes or dislikes an outcome) and responses scored using a Likert type scale. The subject is also asked to indicate his/her expectation o f the likelihood o f the outcome happening14. The subject's subjective expected u tility (SEU) is then the product o f his/her valuation and attached subjective probability. The present study also used SEU ideas to obtain an understanding o f decision-makers’ utility for selected health activities or service attributes as outcomes o f resource allocation decisions. Gladwin (1989) outlines a methodology to study decision-making called Ethnographic Decision-Tree Modeling. The methodology uses all available information such as obtained through observation, record reviews, and interviews, to progressively build a model depicting the reasons for and constraints to a decision-maker's choice. Given the presence o f certain reasons, and subject to certain constraints which must be overcome, the decision-maker chooses an outcome. The advantage o f this technique over the purely quantitative based ones is that it attempts to build a model based on the insider’ s “ emic” view o f the reality. It tries to consider all the significantly relevant factors o f the context in which decision-making is being done. It is thus more realistic and superior to linear modeling techniques. Decision-tree modeling is the technique that has been considered appropriate for modeling the decision-making process in this study since, even i f the above short-comings o f linear models were not there, getting the kind o f extensive good quality data needed would be

M These considerations were the basis for the questions on perception o f budget items and other service outcomes reported in Paper 1.

24


difficult in Uganda at this time. There are a number o f other approaches to decision-making which include the following: •

Expert makes the decision - here a person or group which is believed to have the necessary expertise is entrusted with the task o f making the decision.

•

Voting - a number o f alternatives are laid out and the group, commitee, or members o f the public, as the case may be, indicate their preference by casting a vote. Decisions can then be taken on the basis o f a majority vote, or through rank ordering.

•

Use o f criteria - a matrix o f options is laid out against some criteria. The criteria can be given weights, according to how important each is to the individual or group making decisions. The situation is even more complicated when, as happens at district level decision-making, there are numerous alternatives to choose from, and the decision-making is a group process. The problems involved in use o f criteria have been discussed in some recent books on decision-making (March, 1994; Lewis, 1997; Stone, 1997). These authors agree that decision-making, especially in matters o f public policy is a paradox at best, even in developed countries where information is much more readily available than in developing countries. The impossibility theorem first proposed by Nobel laureate in economics Kenneth Arrow, predicts that a group using multiple criteria and ranking different alternatives, can never make choices that are consistent unless they are over-ruled by a dictator!

The problems o f all these alternative ways o f making decisions are even more amplified in the developing country context, where the political process o f decision-making is made especially acute in the setting o f resource constraints.

25


2.5 Organizational and Management Theories We could view resource allocation in the districts from the perspective o f organizational and management theories. These theories are many and we mention here the relevance o f the organizational theories o f bureaucracy and political theory to the problem o f resource allocation in a district health system setting. These two represent different sides o f the theoretical spectrum with the bureaucracy theory representing the closed, rational and political theory representing the open, natural system (Scott, 1987).

2.5.1 Bureaucratic Approach First prominently articulated by Max Weber (1864-1920), one o f the most accomplished theorists in the study o f the sociology and economic behaviour o f groups and organizations15, bureaucratic theory attempts to explain the behaviour o f organizations in scientific terms as being the best form o f organization to achieve efficiency. If we view the district entity from Max Weber's theory o f bureaucracy perspective, in the ideal scenario we would make the follow ing observation: Districts are legal entities established by law and the constitution. The constitution states that districts are responsible for providing services o f a local nature such as health services. The District health system is headed by a District Director o f Health Services (DDHS). who is a doctor and usually also trained in public health. He/she is appointed by the District Service Commission (DSC) which acts in the district on behalf o f the Public Service Commission (PSC) after careful scrutiny o f his/her qualifications and oral interview. A ll responsibility for running health services therefore rests on him/her. Each o f the main sections in the health department is headed by someone specially trained for that jo b e.g.. a sanitarian in

15 See his work tilled "Economy and Society" by M a x Weber, published in German in 1956 by J.C.B. M o h r (Paul Siebeck) and English edition by the University o f C alifornia Press. 1978. 26


charge o f water and sanitation, a nurse in charge o f all nursing activities and a doctor in charge o f curative care, etc. Information about health problems and priorities is available through routinely collected data and surveys. The professionals are assumed to be competent enough to know what should be done based on these data. It is assumed that the local district leadership are interested in the health o f their people and when presented with the facts, w ill agree to allocation o f resources to programs that address those problems. The priorities for which resources are to be allocated are determined by the objective o f efficiency i.e., maximizing output for a given level o f resources. Guidelines have been handed down to the districts from the central ministries on the steps to follow in allocating and accounting for the money that central government sends to the districts. In order to safeguard against possible mis-allocation o f resources every request for funds is carefully scrutinized by the internal audit department in each district. The district auditor regularly reports to the A uditor General in the national capital. It seems surprising from the bureaucratic, scientific management viewpoint that even though written circulars were sent from the responsible central government ministries as to the procedures to decide on how to allocate funds, a lot o f confusion exists. The bureaucratic perspective assumes that everybody can see the priorities and agree on which programs should be funded based on objective technical criteria. It assumes that information that is collected is adequate and accurate. The DDHS who is the head o f the D H M T is assumed to have final full authority and that he/she heads a department staffed by carefully selected, competently qualified and well motivated people who accept their position and role in the district health system as the best for them. These assumptions are often not true in practice, especially in a country such as Uganda, which has undergone a traumatic political and economic history. There is often a conflict between the value system o f the individual and that o f the society (Lewis, 1997). What is good for the individual is often harmful to the society at large, and vice-

27


versa. This paradox has sometimes been called the "tragedy o f the commons". Imagine that you are one o f many who graze on the common pasture. I f everybody follows the agreed rationing plan, the grass w ill last longer for every body. However, if one pastoralist finds a way to secretly get more than his share (and fatten his animals in the process), he w ill do it, particularly if the chance o f benefit outweighs the potential to get caught. The recent drought in Maryland is one such example. Presumably, if all residents use the water sparingly, the reserves would last longer and we would all be better off. But what o f the individuals whose livelihood depends on the use o f water for business purposes? Their plight was illustrated by "The Sun” , a local newspaper with the following headline "Businesses ask state to ease water rules” (The Maryland Sun, August 17"'. 1999). Another reference to this problem is the famous "prisoner's dilemma” that is mentioned in many economics and management writings. Clearly, if the prisoners co-operated, it would turn out in the best interests o f each one o f them. However, the individual, i f he has to be consistent with his/her own self-interest, seeks to get what he/she perceives to be the best deal for him/herself (Lewis. 1997). In such a scenario all are worse o ff than if they co-operated! Another relevant concept is what is known as the "rule o f rescue” . Society feels a moral imperative to save those who are in immediate danger. To illustrate the concept, suppose we have a sum o f money that could be used to immunize a number o f children, say 100. and prevent disease. However, when faced w ith the choice o f a costly operation for one sick child whose life might be saved, and the possibility o f using the same amount o f money to prevent disease in some other children, society usually is most likely to choose to save the one sick child. From the point o f view o f cost-effectiveness, they should choose to immunize the children, but who is going to let the child die in favour o f other as yet un-known “ faceless” children?

28


2.5.2 Political Approaches According to Bolman and Deal (1991) basic conditions that lead to a political situation are the following: 1.

Differences - people interpret things differently and have d ifficu lty agreeing what is true or important. The health workers, managers and district politicians may see priorities differently.

2.

Scarce resources - there is not enough to give everyone what he/she wants so decisions have to be made about which resources each one gets. Resources tend to be allocated according to who is able to exert most influence (i.e., the most powerful).

3.

Interdependence - people are not individually self-sufficient and so they need each other's support and resources.

Alexander and Morlock (1994) point out that there is widespread agreement among the experts as to why systems of power and politics arise. Power and politics are most often used in organizations when goals are in conflict - as they often are in health care organizations, power is decentralized/diffused throughout the organization, and information is ambiguous - e.g.. there are many unknown influences on health. According to them in most health care organizations ambiguity and uncertainty surround both the establishment o f organizational goals and the means to achieve these goals. The other systems o f influence in organizations (authority, ideology, expertise) are not sufficient. Based on political theory, districts would be viewed as above all political entities. The leaders o f districts w ill want to provide for politically popular programs that w ill meet the approval o f their constituents, so that they can get re-elected. In Uganda, district health systems are headed by a District Director o f Health Services (DDHS) who may be well trained but there is no way this

29


individual can know everything that needs to be known in order to effectively manage the health system. As predicted by the theory o f bounded rationality, he/she has to divide his/her time among the many problems competing for attention (March, 1958; Simon, 1960; Simon, 1976; March, 1985). He/she must therefore rely on others. The power to make decisions and improve health are not centralized in the DDHS alone but is diffused throughout the district health system i.e., in the nurses, the physicians, the clinical officers, auxiliary workers and support staff. To be effective technical knowledge alone is insufficient. One needs to form alliances and coalitions in order to influence allocation o f resources in a way that is favorable to one’s section or department. Even though there is an organizational hierarchy and formal procedures, in practice these cannot be rigidly followed. Information collected about health problems and priorities is often inaccurate and confusing and the professionals may not always know what should be done. The professionals w ill often not be able to present issues and priorities in a way that the local district leadership can understand and accept. The programs for which resources are allocated may turn out to be those politically popular and not necessarily those which are technically most efficient. Since information is often not clear or adequate, decisions on how resources are to be allocated depend a lot on bargaining and balancing among interests, leading to some struggle, conflict, a sense o f winners and losers. Such interpersonal difficulties were found to be a major hindrance to the normal functioning o f the district health management team during a UNICEF operations research study in one Ugandan district (Criel and Pariyo, 1997). The guidelines supplied from the center are likely to be ambiguous, selectively available, and some people may restrict their availability as a means o f gaining power over others. In summary, based on the review o f literature and practical considerations, this study attempted to identify the stake-holders, the different decisions being made, the estimated monetary costs o f those decisions, as well as the expected benefits. This was in line w ith the first principle o f

30


effective decision-making, namely that you should know the costs and benefits o f a decision, and that you cannot make good decisions unless you have clear objectives o f what good you want to achieve, and the evil that you want to avoid (Lewis, 1997). We should aim at knowing our objectives, and the probabilities associated with the uncertain events (like whether money w ill be available or not, whether it w ill be released in time or not, etc). Like in gambling when the wheel is not fair (it favours the house!), the gambler is guaranteed to lose in the long run. I f the gambler is clear about desired objectives (e.g., the amount o f money to make) and knows the odds, he/she can know how much resources he/she is w illing to risk, and when to quit. The difference here is that a public policy decision-maker does not have the luxury o f quitting, even if he/she leaves office, somebody else has to take the mantle o f decision-making!

31


C HAPTER 3: C O N CEPTUA L FR A M EW O R K

3.1 General Health Systems Model A basic model o f a district health system is here described. The district health system is made up o f all institutions, resources and structure, whether public or private, that offer health care or support it (W HO, 1988). For purposes o f this study, focus was on the public component o f the system. A given population has a particular health status which generates health needs and demands which are based on real or perceived health problems. The purpose o f the health system is to respond to this need through provision o f information and services which may be promotive, preventive, curative or rehabilitative. In order to provide this service, the health system needs human, material and financial resources as well as time to deliver appropriate care. These resources should be appropriately mobilized and geared towards delivering high quality services. In other words, personnel should not only be competent enough to deliver the services, they should also be motivated to do the job and should be provided with the support they need to do it e.g., through regular supervision and training. Accordingly resources need to be allocated to maintain delivery o f high quality services. The process o f decision-making for resource allocation involves a technical planning phase and a political phase. In the technical phase, health professionals and managers decide what needs to be done based on their knowledge o f the epidemiological profile o f the district, their observations and expressions o f need from the communities, and their knowledge and belief o f what is an appropriate response. The D H M T then comes up with a health plan and budget, usually covering one financial year. This plan and budget are presented to a group o f politicians. In the Ugandan case, this group is appointed by the Local District Council (LD C ) and forms the District Health Committee

32


(DHC). The committee reviews what the technocrats have prepared and suggests changes accordingly before submission to the full political assembly o f the d istrict16. The decision-making process ideally comes up with a list o f objectives for the health services and a set o f priorities for the period being planned for. The administrative structure o f the district is shown in Appendix A. Resources are allocated according to interventions

that have been chosen

for

implementation. These interventions are presumably chosen as a result o f the discussion and dialogue among the different stake-holders, as well as the information available about the health status and needs o f the population. There are thus different decisions that could be made regarding resource allocation and each possible decision is associated with uncertainties and outcomes. Some o f the activities that have been planned are funded and they get implemented while others may not get funded. Similarly, activities that were not provided for could get funded thus diverting resources from those which were planned. Even if activities are funded, the entire budgeted funds may not be available or released fora number o f different reasons that may occur. Finally even when the activities get implemented, their effectiveness and output w ill vary leading to differences in outcomes. Some districts are thus likely to be more productive or better than others in delivering a high volume o f service given resource expenditure whereas others may be less productive, expending resources with less output. The quality o f outputs and outcomes w ill also vary with district. The comprehensive picture o f the relationship between health status, expressed needs, decision-making, service capacity and activity implementation and health outcomes is depicted in Figure 3.1.

16 Under decentralization the highest decision-making body at district level in Uganda is made up o f elected officials called the Local District Council (L D C ). Further description o f the structures at district level is presented under the chapter on methods. See also Appendices B, C, D and E. 33


Figure 3.1: Conceptual Framework of a District Health System in Uganda

Uncertain Events

Community Expressed Needs

Community Health Status

- funds not available - funds availab le not released - funds released not adequate - adequate funds but poor im plem entation

i i

DecisionIntervention Making/ Planning/ < Prioritization Budgeting

Y

Uncertain Events

A ctivivity Implementa tion

HEALTH OUT COMES ► (Lower Morbidity

& Mortality)

Professionally Determined Needs

^

* '

. • --------------- —Service Capacity of Health System

Legend:

----------- > ►

Indirect Relationships Direct Relationships

3.2 Decision-Making Model I f resources are available or are expected to be available, the decision-maker has to decide whether to allocate/release them for a set o f interventions X, where X represents health services interventions or activities addressing needs or problems that are perceived to be a priority and/or that w ill promote service delivery, better health outcomes, improved health status, etc. Alternatively, the decision-maker could allocate/release them for another set Y. Here. Y is taken to include any

34


other health services intervention or activity that is not very useful in promoting a better health status for the population in general but is seen as important by the decision-maker from a personal pragmatic/political point o f view. Examples o f X would include allocating more resources towards PHC service delivery e.g., increasing immunization coverage, doing an IEC campaign on prevalent health problems, promoting antenatal attendance by expectant mothers, etc. Examples o f Y would include allocating money to build new facilities in an area already covered, buying a new vehicle (that is not going to be used to extend service coverage but rather for general administrative and personal duties), or allocating for activities that w ill increase personal gain and the bureaucratic convenience o f the District Health O ffice staff. One could also take X to be allocations for invisible but beneficial recurrent expenditure to promote service delivery, and service Y to be capital expenditure. X could be PHC services and Y could be additional (non-core) hospital based services. These two option categories represent the possible use o f the available resources and in set notation are enclosed in curly brackets. The decision-maker subjects X to a series o f criteria and has to decide if the conditions are favorable or not. Such include the constraints o f effectiveness of intervention, knowledge o f the health personnel, adequacy o f the available resources and whether allocating resources to X would be acceptable or popular with the people whose approval the decision-maker has to obtain, whether the money is earmarked or not, whether guidelines exist or not. decision-maker’ s understanding o f those guidelines or not. etc. Each o f the decision points is represented in decision-tree notation by a diamond while each possible outcome is shown by a rectangle. Each o f the constraints has to be overcome before the decision-maker can choose any of the possible outcomes. A generalized theoretical model o f resource allocation decision-making, which is drawn from the wider district health system framework described before is shown in Figure 3.2.

35


Figure 3.2: Simplified Decision-Trec Model for Theoretical Public Health Resource Allocation Decisions in Ugandan Districts

Given Resources ire Available/Expected

T (Allocate Resources to Intervention/Activity X : Allocate Resources to Intervention/Activity Y) Effectiveness of Intervention/Activity X „

»

Effectiveness of Intervention/Activity Y

Y« ' 'N o

*

*.

Do personnel know how to do X? ■ ' No ^ , y c3 Can they be trained to do X ?,.1 A ' ----------------------------------- ----------------Are Resources adequate for X or are you able to get adequate,. resources for X?, . • • .. Yes

'N o 'A. Don’t Allocate to X __ ____ N o _

Would doing X be acceptable/popular choice? \ Yea •- No A . A____ Allocate to X

Don’t Allocate to X

.

^

....

.

*-*°n * Allocate to X ________________

Allocate to Y

3.3 Rational Decision-Making Framework A rational framework for decision-making is based on the need to be consistent with one's choices and avoiding incoherence. A rational decision-maker would determine the probabilities attached to uncertain events that would obtain if various decisions are taken. He/she would attach a u tility value to each o f the possible consequences for all possible decisions. The rational decision maker is expected to choose that decision or those decisions which maximize his/her expected u tility (Lindley, 1985; Lewis, 1997).We can derive a mathematical relationship for a decision-maker's expectation o f u tility and come up with a quantitative based model, using the steps outlined in Appendix E.

36


We can see from the discussion in Appendix E that a rational decision-making framework is an ideal situation that is hard to find in practice. However, i f decision-makers set objectives or make policy choices an attempt should be made to make decisions that are consistent with those choices or objectives. The data requirements for a purely rational framework are enormous. In practice, the data needs are potentially reducible to a manageable size that at least covers the most important aspects o f decision-making. Such is what is attempted by using composite measures o f mortality and morbidity such as use o f measures o f health outcomes in terms o f DALYs or HEALYs and cost-effectiveness o f interventions. Some researchers (Young 1981; Gladwin. 1989) have shown that taking a behavioral rather than a quantitative mathematical approach can give a fairly high success rate in predicting choice in real life. More-over, decision-makers do not have the time or resources to go through the elaborate steps required by a purely rational approach. What this study attempted to do was describe the process in general terms. In that regard, the approach taken was more geared to collecting preliminary information and laying a basis for future work on the subject in the attempt to narrow the gap between the real life decision-making and what we would theoretically consider rational. We do not pretend in this study that we shall be able to achieve pure rationality in resource allocation decision-making. Rather, the view is taken that it is possible to combine the use o f objective measures with more subjective individual and societal views and still improve on the current practices. This was based on the premise that before we can suggest any improvements to the process, we need to know what is going on and why it is happening. It was assumed throughout that decision-makers want to make the most optimal use o f resources for the greater good o f society. More good can be obtained, we can indeed have a better health status than what we have currently because other societies have achieved it. In technical terms we expect that allocative efficiency can be increased in the use o f the available resources. The World

37


Development Report 1993, “ Investing in Health” shows Uganda as one o f those countries with higher expenditure o f GDP spent on health yet with worse outcome, five years life expectancy lower than expected for their income (The World Bank, 1993). In contrast countries that are in similar income bracket or poorer have better outcomes at the same expenditure levels (e.g.. Mozambique) or have the same low level o f health outcome but at a lower cost (such as Zambia and Ghana).

38


CHAPTER 4: M ETH ODOLOG Y

4.1 Study Setting T ill July 1997 there were 39 districts in Uganda. 6 new ones having been created out o f the old ones with effect from July 1997 bringing the total now to 45. For purposes o f this study, 37 of the 39 old districts were eligible to be included in the study. The new ones were not likely to be fully functional by the time o f field work for the study and were not considered separately. Two others were excluded a p rio ri. These are Kampala, because it is the capital o f Uganda and is entirely urban, it's budget and management being quite different from the others. The other one excluded was Kalangala District because it is made up entirely o f small islands in Lake Victoria and was relatively newly created and not entirely functional as an autonomous health district by the time o f the study. A map o f Uganda with the study districts is shown in Figure 4.1. A typical district in Uganda has a population ranging from 100,000 to 500.000 inhabitants. A few atypical ones have populations o f up to 1,000.000 people (e.g.. Kampala. Iganga) and the one that is made up o f a group o f islands in Lake Victoria (Kalangala) has only 30.000 people. However, the district political, administrative and health management structures are identical in all districts and are as shown in Appendix A.

39


Figure 4.1: Map of Uganda Showing Study Districts

UGANDA Sudan

Kenya

.MOVOi

KITG UM

KOTIDO

GULU

°

of

R e p u b lic

♦

D em ocratic

H im

Congo

k a t a k w i

H A S IN O l

‘HO IM A

K1B0GA K lO A A lE

Kenya

MPIGI

RAKAl ) NTUHGAMO

40


Health services are generally organized along a two-tier system with a network o f primary care units and a district referral hospital. There is a District Health Management Team (D H M T) which is made up o f all section heads including the hospital and representatives from those NGOs or donor projects with significant inputs into health. The D H M T manages the technical aspects o f health in the district. The C hief Administrative O fficer (CAO) who is a civ il servant, is the head and overall accounting officer o f all government departments in the district. The CAO supervises all the administrative aspects o f technical government work in the districts and all technical section heads are directly accountable to him/her. The structure o f the D H M T and relations with the administrative and political structures are illustrated in Appendix C. The Local District Council (LDC) has statutory committees through which it functions. Health matters are handled by the Social Services Committee which is made up o f sub-committees for health and education. Other committees include District Finance Committee (D FC ), Works and Production. The political planning/budgeting relations are illustrated in Appendix B. The procedure for planning/budgeting laid down under decentralization is supposed to function as follows: Matters are discussed and a district health plan/budget is made by technical staff (see Appendix C), who then send it to M O H for “ comments” . They then present it to the District Social Services Committee for consideration. The committee may recommend changes or approve it in which case it moves to the Local District Council (LDC). In the LDC, it is the Chairperson o f the respective committee who presents plans and budgets for approval by the full council. Once approved, the technical people and administrators can then move ahead with implementation. The concern o f this study was how decisions to allocate resources within the public health sector itself are made. In particular the study examined the technical planning as well as the political process involved, as well as the reasons and constraints affecting decisions.

41


4.2 Study Design This was an exploratory, descriptive case-study design. The existing situation in districts was studied in order to identify the current practices being implemented in deciding on resource allocation as well as factors that influence the process and likely intermediate outcomes o f those decisions.

4.3 Variables The main factors o f interest were; 1.

The decisions being made at district level in regard to resource allocation; - what decisions are being made, - who makes these decisions, - why these persons are the ones who make these decisions (e.g.. legal/administrative provision), local council decision, - the basis on which these decisions are made (e.g., epidemiological information, availability o f resources, efficiency/effectiveness considerations, political concerns, etc). - possible outcomes o f decisions made, - values and preferences attached by different stake-holders to outcomes o f decisions.

2.

Relation between decisions made and resources allocated to actual budget expenditures looking at the same questions as in 1 above.

3.

A practical, feasible approach to public health resource allocation in Uganda that takes into account the current reasons and constraints affecting decisions being made; - what components should be included, - how should it be implemented, - what training needs to be done to implement it,

42


- what problems are likely to arise, - how could potential problems be solved, - what information is needed, - relevant information that is available or can be generated.

4.4 Intervening Variables Some o f the factors that might alter or influence the expected results include: - presence o f external donor assisted projects in the district, - distance from the national capital (in kilometers) and travel time. - ethnic composition o f district population, - socioeconomic development level o f the district (as indicated by e.g.. the UNDP education index, life expectancy index and income index), - duration o f decentralization experience (in years).

4.5 Data Collection Data was collected for each district from financial and administrative records as well as through interviews. Key-players in the resource allocation and health management decisions like the CAO. chairperson o f the LDC, chairpersons o f the District Health Committee (DHC) and District Finance Committee (DFC), District Treasurer, Hospital Superintendent, and members o f the D H M T were interviewed when available. These persons are usually known and attempts were made to interview all o f them in each study district. Where the substantive office holder was found to be absent from the district during the study or could not be located and interviewed, the person found acting during the visit was interviewed in lieu o f the official title-holder (indeed such deputies are usually

43


knowledgeable about what is actually going on). The interviews focused on their experience with the public health services in terms o f which activities are considered important, what decisions are made and why. why some activities are funded and not others, reasons for the procedures that are followed (or not followed), likely outcomes o f decisions and the likelihood o f outcomes, how valuable the outcomes are considered in terms o f meeting the needs o f the people, etc. Focus Group Discussions (FGDs) were conducted with elected members o f the District Social Services Committee (DSSC), the committee which does a preliminary approval o f plans and budgets before presenting to the full Local District Council (LD C ) for final authority. Issues discussed included the experiences and expectations o f these members in regard to resource allocation between different health programs, problems faced and expectations o f what the health sector should be doing, what aspects o f the health services they consider important, which activities should be given priority in resource allocation and why, etc.

Data was collected using the follow ing approaches: 1.

Key informant interviews were conducted using a check-list o f questions (Appendix L) for an in-depth view o f the issues. Such key informants included prominent community leaders such as retired civil servants, teachers, school headmasters, political leaders, community advocacy group members, religious leaders, etc. In total 24 such individual interviews were conducted with members o f the public.

2.

Focus Group Discussions were conducted with the elected district councillors who are members o f the Social Services Committee as outlined before (5 to 8 persons per FGD in the 3 districts where it was possible to conduct them)17. The main focus was on issues

17

Other planned FGDs aborted because the D HCs had either been depleted o f members due to their opting to go back to full lime civil service jobs in compliance with a new law, or simply because the D D H S decided not to call them as requested for fear o f the financial demands that would imply. The study only had resources for transport re-

44


influencing decisions for allocation o f resources between different health interventions and how the process can be improved. The proceedings o f the FGD were tape-recorded. The researcher acted as moderator and a research assistant helped take notes as well. A FGD guide was used (Appendix M). 3.

Structured in-depth interviews were done with district level decision-makers using Likert type scaled responses (DeVellis, 1991; Henerson, Lyons Morris et al, 1987). The questionnaires used were constructed for each o f the three main categories o f ditrict level decision-makers relevant to health, i.e., members o f DH M T, district Financial management staff, and district administrators/politicians. Some o f the questions for the D H M T were modified versions o f the type used for assessing district health systems (Kielman. Janosky et al. 1991). (See Appendices N, O, P, Q, R, S). These decision-makers who play a key decision-role in allocating or managing health resources are shown in Appendices C and D.

4.

Document and record reviews included district health plans and budgets, financial and administrative reports, as well as national guidelines and reports.

A ll the interviews were conducted in English which is the main official language in Uganda. Translation o f questionnaires and conducting interviews in the local languages was considered (especially for interviews with the members o f the public) but found to be impractical since there were 7 local languages in the 10 districts covered by the study!

4.6 Sample Size Determination This being a case study, the number o f 10 districts that were covered was arrived at based mainly on the concern for regional balance, the time the district was decentralized, and logistic

imbursement yet they would demand for night allowances as is the usual practice.

45


feasibility for conducting the study.

4.7. Sampling Procedure The old 39 districts o f Uganda were grouped into 4 regions as follows; - Northern, - Eastern, - Western, - Central. These groups reflect ethnicity, geographical and to some extent socioeconomic levels of development. During the process o f decentralization districts had been categorized by the central government according to general levels o f development into 3 groups called D,. D:. and D-, o f 13. 14 and 12 districts respectively. The first batch or D, districts started decentralization in 1993/94, followed by D2 in 1994/95 and D3 in 1995/96 financial years respectively. D, districts were considered the most developed and D3the least developed with D : judged to be in between the two. The 37 districts eligible to be included in the study were stratified according to the regions identified above. W ithin each o f the 4 regions. 2 districts were selected using simple random sampling with the aid o f a scientific calculator's random number function. This was done in such a way that there was one each to represent D, and D3 districts in the region respectively. This provided 8 case study districts with 4 from among the most developed (D ,) and 4 from among the least developed (D,). The remaining 2 case study districts were purposively selected from D : districts in the same region and were used to provide information on some o f the study aspects e.g.. why one might have a better health status than the other. The leaders and key persons identified previously as well as members o f the DHMTs o f the

46


respective districts were automatically included in each district.

4.8 Quality Assurance A ll interviews were conducted by the researcher with the help o f a research assistant (one o f two). The top district leadership people were interviewed by the researcher himself. The research assistants were trained and supervised by the researcher. Field editing o f the filled questionnaires and where possible going back to respondents for missing information was done. In addition, after a preliminary analysis, the researcher had the opportunity to go back to 5 o f the 10 study districts to fill gaps in financial and health services profile data, as well as check data that appeared unreasonable. During the design phase o f the study, pilot testing o f the main study instruments was done in two districts located in two o f the four regions.

4.9 Analysis and Interpretation The main aim o f the study was to generate information upon which decision-making for resource allocation can be improved. It was aimed at a better understanding o f factors that affect decision-making for resource allocation within the districts. It was also expected to obtain information that can form a basis for development o f procedures for resource allocation that promotes the improvement o f health status for the population. The unit o f analysis was thus the district as such in regard to specific allocation patterns and the whole country for factors influencing allocations. Taped records from the FGDs were transcribed and together with notes from the individual Key Informant interviews were typed into a word processor. The resulting notes were analyzed manually. Impressions that were formed during the study were checked against different information sources.

47


Structured responses from the questionnaires were analyzed by generating frequencies and percentages using Epilnfo Version 6.0. Financial data was analyzed with the help o f spreadsheets that were developed using QuattroPro Version 8.0 and information generated in form o f simple proportions, percentages, means, medians, and standard deviations. A ll financial information from the two financial years analyzed (1997/98 and 1998/99) were converted into constant 1998 Uganda shillings using the Consumer Price Index (CPI). The information was made comparable across districts by adjusting for population through the use o f per capita figures. Finally, for purposes o f exploration only, the data obtained from the rating scales was analyzed using both principal components and factor analysis techniques with the help o f computer software (STATA Version 6 and SPSS for Windows Version 8.0).

4.10 Model Selection Independent variables expected to have a linear association with the response variables were correlated in a pair-wise manner using both Pearson’ s and Spearman's correlation methods. Since the sample size was small Spearman's correlation was used in order to provide more robust results since it uses rank order rather than the actual data, and is therefore less reliant on assumptions on the nature o f the data. In general, both Pearson's and Spearman's methods yielded similar results. However, independent variables which were found to have correlations significant at the 5% level by either method were selected for the multivariate exploration using multiple linear regression. Due to the problem o f multi-collinearity, the final models were chosen after eliminating those variables which no longer caused significant change in variance on inclusion o f another correlated variable. The study generated 4 main forms o f information as summarized in the results sections o f the 3 theme papers. These are, (a) stake-holders, and main allocation items, (b) decision-maker

48


preferences, decision rules and priority categorization o f problems, (c) itemized estimates o f allocations and broad disbursements by source and district, and (d) district profile and resource allocation indicators which can be used in monitoring and assessment.

4.11 Threats to Validity

4.11.1 Internal Validity The response items on the scales used for exploring perceptions may not have been understood in the same way by the respondents. Scale reliability was assessed using Cronbach's alpha18. In general, the scales used for assessing decision-maker’ s perceptions could be considered acceptable given the exploratory nature o f the research and had values for Cronbach's a ranging from .92 to .94 for the 18 items used. The 5 item scale used to explore jo b satisfaction had a borderline reliability with Cronbach’ s a o f only .5253. The same team administered the questionnaires and conducted the interviews in ail the districts, which should have reduced measurement error due to observer variation. To try to avoid this, it was emphasized to the research assistants during training that the questions should be asked the way they were written so that everybody hears the same questions. In addition the researcher and the assistants had had lots o f prior experience in interviewing people at district level and generally understood the organizational culture. The research assistants were given detailed handouts which

18

Cronbach's alpha assesses the reliability o f a summative rating scale such as a Likcrt scale, in measuring an unobserved factor. Alpha is the square o f the correlation between the measured scale and the underlying factor. By convention, a good set o f scale items should have a Cronbach’s alpha o f .80 or higher for about 10 items (Stata Corporation. 1997; Bernard. 1995). 49


spelt out the study objectives and what we were trying to achieve, as well as the different steps of the study. One problem that was encountered in the field was that although it was planned that each questionnaire would be directly administered, for 5 respondents, it was impossible to do this and the respondents filled the questionnaires by themselves. However, the potential bias from this was minimized by allowing this to be done only by the highest officials whom it would have been impossible to find time to interview otherwise. Where such cases arose (especially for DDHSs, CFOs and CAOs), the potentially confusing parts o f the questionnaires were explained verbally before leaving them. In addition, when we collected the questionnaires, we checked to be sure the parts were answered appropriately and where necessary clarifications were sought and corrections were made on the spot. Another potential weakness is that only 5 o f the 10 study districts were re-visited. However, with the exception o f one, the ones not re-visited were generally those from which more complete information was obtained and in which the research team generally experienced a greater sense o f collaboration from district health managers. It is also possible that such cross-sectional information that was obtained may not be entirely representative o f what usually happens. An attempt was made to collect financial information for each o f the 3 past financial years namely 1996/97. 1997/98 and 1998/99. However, only 2 districts had fairly complete information for each o f these years. The strength o f the study is that information was obtained from different sources and some o f the impressions formed during the visits to the districts were cross checked in with national level M O H and M O LG staff who either confirmed them or gave possible explanations. In addition, in each district, people interviewed freely gave their opinions about their collaborators e.g.. D H M T members spoke o f their working relations with district authorities while some o f the administrators interviewed also gave their opinion as to whether they thought the health people were doing their

50


jo b properly or not, and what they thought should be improved.

4.11.2 External Validity To cater for variations due to regional and decentralization timing differences, the districts were stratified accordingly and a representative district from each region or decentralization category was randomly selected. Unfortunately due to insecurity in parts o f the Northern Region and Western Region, some districts were excluded from the sampling frame for those areas. However, the general picture that emerges can reasonably be expected to represent decision-making for public health resource allocation in Uganda. Although the study is specific to Uganda, some o f the problems and issues identified are those typical o f a poor sub-Saharan African country. However, further research in other sub-Saharan African countries similar to Uganda would be needed before extrapolating findings to other countries.

4.12 Protection of Human Subjects and Ethical Considerations The study proposal was submitted through the relevant ethics committees at the Johns Hopkins School o f Hygiene and Public Health as well as the National Council for Science and Technology Ethics Committee in Uganda through Makerere University Medical School. The relevant approvals were obtained before data collection began. Individuals and groups to be interviewed were given full information regarding the objectives o f the study and their consent obtained prior to interviewing them using consent forms approved by the relevant authorities mentioned above. They had the option o f stopping the interview if they felt uncomfortable with it. A ll the prospective interviewees were assured o f confidentiality. To ensure confidentiality no names were written on the questionnaires. Only the researcher had access to the code book for districts and individuals

51


throughout the study. The identity o f districts and individuals were protected throughout. In general, there was a very good spirit o f co-operation from the respondents. They found the subject interesting and discussions were quite lively. Throughout the study only one respondent stopped the interview after in itia lly granting consent because he had to attend to a jo b commitment. As much as possible we tried to make appointments for the interviews in advance although it was hard to find officials who were not busy. Decentralization has produced such a flurry o f activities that the majority o f officials at district level are quite often busy. Ugandans are generally very sensitive about signing their names to any document because o f the history o f political turmoil that the country has gone through. People are. however, generally w illing to be interviewed and to provide information especially about health services when asked in private and assured o f confidentiality. Consent was obtained verbally but a written record signed by the researcher and witnessed by the research assistant was kept for each respondent. For the FGDs, consent was obtained both for their participation in the discussions and for having their voices tape-recorded. Written consent as is practiced in the United States and other countries is hard to obtain in Uganda for the above mentioned reasons. So verbal consent and completion o f the questionnaire were considered adequate.

4.13 Utilization and Dissemination of Results Target audiences for the results include the following: - district health managers, administrators and politicians - these are likely to be interested in changing their spending allocations in order to obtain the most favorable reactions from the public. Eventually these persons are likely to want to allocate funds to those health activities that give them the most benefit e.g., meeting needs o f the people, and improvement o f health status. The study has attempted to identify information requirements for more

52


effective planning as well as bottlenecks to current delivery o f health services. - M inistry o f Health - are interested to see how districts are adhering to national health policies since decentralization has conferred on the M O H a monitoring and policy formulation role. They are also in the process o f developing standards and guidelines for different health activities. A preliminary discussion about the study observations was done by the researcher with the Director General o f Health Services o f Uganda10 - M inistry o f Local Government - the M OLG is currently working on developing a model to be used for the central government to provide equalization grants to districts20. - M inistry o f Finance and Economic Planning - they have to decide on how much money districts really need in order to run the services. They are also likely to be interested in seeing whether decisions are being made in the interests o f the majority o f intended beneficiaries o f services, the communities in rural areas. There have been concerns that districts should adopt transparent systems for resource allocation. - United Nations agencies (UNICEF. WHO. UNDP. UNFPA). - NGOs working in the health sector in Uganda, especially those in the study districts. - Academic and professional, institutions and organizations The main means o f disseminating the findings locally in Uganda w ill be through briefing sessions to discuss the findings with officials o f the participating districts as well as copies o f papers to be produced out o f this work. Matters o f relevance to national policies w ill be discussed with officials o f the M O H during the Annual General Meeting which groups together all stake-holders

19

The D G H S is the highest technical official in the M O H responsible for overall direction o f health services in the country. :o

Equalization grants arc provided for in the national constitution o f 1995 which ushered in decentralization, but a law to regulate it is still under consideration. The M O L G Finance Commission has made some technical proposals to guide the process (M O L G , 1999). 53


in the health sector in Uganda. International dissemination shall be done by means o f 3 papers to be submitted for publication in peer reviewed journals. The 3 papers shall be drawn from each o f the 3 thematic papers which are respectively on; - how decisions are currently being made in allocating resources at district level. - information available and how the decision-making process could be improved. - what capacity changes/inputs are needed at district level in Uganda in order to improve resource allocation decision-making.

54


PART II: TH EM E PAPERS

CHAPTER 5 :PROCESS A ND R EASO NS FOR C UR R EN T D EC ISIO N M AKING PRACTICES (PAPER 1)

C HAPTER 6 :D E V E L O P IN G

IN F O R M A T IO N

FO R R E S O U R C E

ALLO CATIO N DECISIO N-M AK ING (PAPER 2)

CHAPTER 7:TO W A R D S

B U IL D IN G

C A P A C IT Y

FOR EFFEC TIV E

DECISIO N-M AK ING IN DISTRICT LEVEL RESOURCE A LLOCATIO N (PAPER 3)


C H A PTER 5: PROCESS AND REASO NS FOR C U R R E N T DECISIO N-M AK ING PRACTICES (Paper 1)

5.1 Introduction This paper w ill present the findings o f a case study into how decision-making for allocation o f health resources is currently taking place in Ugandan districts in the context o f decentralization. In answer to the question "how are decisions being made and why?” it w ill show that at the moment resource allocation decisions in Ugandan districts are mainly driven by historical categories, availability o f funds, presence o f guidelines from central authorities and donors, extent to which problem is thought to affect a large number o f people or cause death, demand by influential persons, and expected personal benefits. It lays the framework for the next two papers which are aimed at providing an insight into the second big question: How can the decision-making process be improved? The word “ improvement" throughout this work is used to refer to an allocation practice that places resources for health programs and interventions in such a way as to bring about, or are most likely to bring about better health outcomes for the people o f that district. A better health outcome should be objectively verifiable improvements such as better immunization coverage rates, coverage o f pregnancy by Ante-Natal Care, as well as impact measures such as lower infant, childhood and maternal mortality rates, and a higher life-expectancy as defined by the UNDP Human Development Report. In principle, a measure that combines both morbidity and mortality into one composite indicator such as the D A L Y or H E A LY could be used to provide a common indicator. This would give us a means o f comparing the degree o f need in different areas o f the country as well as which interventions are most efficacious. Whereas the priority health problems which cause most mortality 56


and disability and for which cost-effective interventions exist have been identified and expressed in form o f the Uganda National M inim um Health Care Package, how to translate this into effective public health programmes at district level remains a major challenge. In real terms, the interventions considered in the calculations may not achieve the levels o f effectiveness that are assumed. It was therefore the intention o f this study to identify the main factors influencing the decision-making process in regard to allocation o f resources available to the health sector at the main implementation level, the district. It is widely believed that the district level (or its equivalent) is the ideal level at which national policy and community needs meet and health services are planned for and managed (WHO. 1988).

5.2 Background and Objectives Following decentralization o f all social service provision to the district level in 1993. districts were faced with new challenges o f planning, and overall management o f health services. Functions that used to be done by the central M O H were now devolved in a phased manner to district level over the next 3 years. Health managers at both district and national level expressed the need for information as to what was working and what needed to be improved. Management o f resources is one o f the main challenges districts found themselves facing. This study was conceived in order to make a contribution to promote better management o f the scarce health resources. Other background information and literature review is covered in Chapter 1. The study aimed to achieve the following purposes by the end o f December 1998:

5.2.1 Overall Goal of the Study The overall aim o f the study was to review the process o f decision-making for allocation o f public health resources in decentralized districts in Uganda in order to provide a basis for

57


development o f allocation procedures for district health systems that address health needs o f the people and promotes rational use o f scarce resources.

5.2.2 Objectives 1.


2.

Identify the formal guidelines/criteria that are currently existing in the districts with regard to allocation o f public funds to health activities, i f any. and the informal influences that play a role.

3.


4.

Assess the extent to which decisions being made in allocation o f public health funds relate to the health problems perceived as priorities (e.g., as defined by the decision-makers' perceptions and based on national guidelines such as the national basic health services package) as well as reasons for any observed deviations.

5.


This Paper 1 presents findings and issues related to Objectives 1, 2, 3 and 5. Paper 2 w ill present findings relevant to Objectives 3 .4 and 5. Findings and issues related to Objective 5 as well as broader contextual issues in Ugandan districts w ill be addressed in Paper 3.

58


5.3 Subjects and Methods The following is a brief overview o f the methods used in the study. A detailed description is provided in Chapter 4. Qualitative methods were used to obtain information on decision-making practices from each o f the 10 districts studied. The respondents included ordinary citizens who were interviewed as key informants in the district capital. They ranged from retired and active c iv il servants to business people and casual labourers. A semi-structured guide (Appendix L) was used to explore their perceptions and experiences o f the health system. Focus Group Discussions were conducted with District Health Committee members to explore their perceptions and experiences o f the decision-making process using a guide to ensure key issues were addressed (Appendix M). A detailed in-depth interview was conducted with members o f the District Health Management Teams and district administrative and political leaders. Preferences and perceptions o f priorities were obtained using scaled responses with the aid o f questionnaires (Appendices N, O, P. Q. R. S). Where available, copies o f reports, health plans and budgets were obtained and analyzed.

5.4 Results 5.4.1 Respondents There were respondents for both structured as well as open ended questions. There were a total o f 68 persons interviewed using structured instruments (Appendices N, O, P. Q, R. S). The respondents ranged from members o f District Health Management Teams (D H M T), District Local Government Administration Officials e.g., Chief Administrative Officers and/or their deputies, as well as political leaders, namely Secretaries for health and social services, and finance department officers and accounting clerks in the D istrict Health O ffice as well as the D istrict Finance

59


Department. The respondents were drawn from the 10 study districts. Although some districts yielded more and some less the median number o f such interviewees in a district was 7 with a range from 1 to 14. Key Informant interviews were done with 24 district residents found in the district capital who were selected from among people that were deemed likely to know about the state o f health services. For practical purposes, these were chosen from among people found in the main town o f the district. Five out o f the 10 Chairmen21 o f Local District Councils and 5 Chief Administrative Officers (CAO) o f the study districts were interviewed as key informants. There were a total o f 20 participants in the 3 FGDs that were conducted with District Health Committee (DHC) members in the 3 districts in which it was possible to do them22.

5.4.2 District Health Planning and Budgeting The central concern o f this study was how decisions to allocate resources within the public health sector itself are made. In particular the study examined the technical planning as well as the political process involved, as well as the reasons and constraints affecting decisions. The official process o f planning and budgeting for health are described here as put together from various interviews and review o f documents. The main local governance structures are spelt out in the Ugandan Constitution (Uganda, 1995), and the Local Government Act (Uganda, 1997). A politically elected Local District Council (LDC) headed by a Chairperson is responsible for approving all plans and budgets that are prepared by the technical personnel. This district council

21

They arc the political leaders o f the district and are usually elected by popular vote. T>

A t the time o f the field work many D HCs were non-functional since many o f their members had resigned from the District Councils in order to retain their civil service jobs in compliance w ith a new law. In some other districts although a good number o f members remained on the committee, it was not possible to convene them for logistical reasons.

60


(Local Council V, or LC V ) is directly elected by the people and is made up o f 2 councillors for each sub-county. The Chairperson o f the District Council is directly elected by eligible voters o f the district. The Council then elects from among it's members a Vice-Chairperson. General Secretary, and secretaries for Finance. Information and Mass Mobilization, Women Affairs, Youth. Production. Social Services (Health and Education), and Defense. This basic structure o f governance is replicated at the sub-county level or LC III (the county level, LC IV has been phased out), parish level (LC II) and village level (LC I). The main difference is that the village council is made up o f all residents above the age o f 18 (eligible to vote) who choose an executive committee composed as above. It is these village executive committee members, however, who form the electoral college (LC II) that selects the parish executive. Sim ilarly these parish executive committee members form the sub-county council (LC III) that elects the sub-county executive committee (LC III) (see Appendix A). The Local District Council (LDC) has statutory committees through which it functions. Health matters are handled by the Social Services Committee which is made up o f sub-committees for health and education. Other committees include District Finance Committee (DFC), Works and Production. The political planning/budgeting relations are illustrated in Appendix B. The procedure for planning/budgeting laid down under decentralization is supposed to function as follows: Matters are discussed and a district health plan/budget is made by technical staff (see Appendix C), who then send it to M O H for “ comments” . They then present it to the District Social Services Committee for consideration. The committee may recommend changes or approve it in which case it moves to the Local District Council (LDC). In the LDC, it is the Chairperson o f the respective committee who presents plans and budgets for approval by the fu ll council. Once approved, the technical people and administrators can then move ahead with implementation. A ll the members o f the D H M T were reported to participate in making the Annual District

61


Health Plans and budgets in all districts. The supervisor o f the clerical staff in the District Health O ffice who is called the Executive O fficer Accounts (or Higher Executive O fficer in the past) often participates as well especially in giving advice on financial matters. The main budget items on which decisions are being made for resource allocation were obtained from review o f district health plans and budgets, as well as reports, and interviews with officials. The full list o f items, which is generally representative o f ail the items covered in all the 10 study districts is attached together with a key to the reading o f the variables (Appendix G). From the interviews and observations, it was evident that the process o f getting funds released is a very complex one. There is an inter-play o f bureaucratic and political procedures within the district involving various offices and structures and between the district and central government ministries. These were summarized in a series o f 5 flow-charts (available on request). They are a reflection o f the importance attached to following bureaucratic practices and maintaining political balance between the stake-holders involving health technocrats, NGOs, district councillors, and central government. The sitting o f committees o f the LDC are subject to the availability o f funds to pay their sitting allowances. Unlike civil servants, the district councillors have to be paid perdiem for their upkeep since some o f them come from outside the town. They usually have to be paid the same day that they come, which means meetings can only take place when the district has enough money available! In the absence o f relevant technical information to guide decision-making, the concern o f the council is more often to maintain social harmony and order and an appearance o f fairness to all areas regardless o f actual need.

62


5.4.3 General Decision Rules From the interviews and observations, the main influences and determinants o f allocations can be summarized in the decision rules presented in this section. It should be noted that these rules do not operate in isolation. Together they influence the course o f events in allocation decisions. Respondents in all districts mentioned that the allocations are based on national priorities, the previous year’ s allocations, or on guidelines issued by the respective vertical program or donor/NGO. The districts visited operated between 6 to 8 separate bank accounts for health funds each. The commonest accounts mentioned were for funds from donor projects (e.g.. D A N ID A . WES, etc). MOH programmes e.g., STIP, DHSP, and central government conditional grants e.g., that for PHC and for special projects implemented by NGOs e.g.. Onchocerciasis and Guinea worm programs. Some mentioned the use o f criteria such as number o f people affected and the perceived lethality o f the condition.

5.4.3.1 Decision Rule 1: Historical Categories There is a tendency to allocate money because an item was in last year's budget. The MOH issued broad guidelines on what the health plans and budgets should contain. They are meant as a guide as to what should be covered. A ll district budgets had some items in which exactly the same amount o f money was allocated for more than one financial year. Salaries and other items like lunch allowance are generally fixed but one would expect differences from year to year in amounts al located for various items. A number o f respondents mentioned that the amounts to budget are based on the previous approved budget because they had better chances o f being approved. They have to offer explanations for items that do not seem to conform directly to the guidelines. This means that they feel constrained to allocate to activities that do not seem to fall within the approved categories. However, a review o f the guidelines shows that they are broad and there is plenty o f room for

63


fle x ib ility provided there are good reasons. The workplan was often mentioned as the basis for their allocations. On probing how they decide what goes into the workplan, a frequent answer was that they put in activities that they are already doing from last year's workplan. Other answers were; “ central government dictates", “ priority is given i f it is for PHC” , “ it is based on the previous year's allocation” .

S.4.3.2 Decision Rule 2: Availability of Money Most funding for health still comes from the central government and donors. Most o f this money is sent to districts under specific programmes. Thus certain problems such as mental disorders are perceived as priorities but no money is allocated to them since there are no ear-marked funds for it. Although malaria is the major cause o f morbidity and mortality, amounts allocated for malaria are minute in comparison toothers like MCH/FP, TB/Leprosy, and HIV/reproductive health. Estimates o f allocations for various programmes per capita are presented under Chapter 6 (Paper 2). The agenda for these programmes is often set by the central government under influence o f donors who provide a significant input o f funds for health. The most striking example o f this is that in most districts amounts allocated for the National Immunization Days (NIDS) which are meant fora polio eradication campaign were more than funds allocated for routine immunization even though NIDS campaigns are only about twice a year. A ll the funds for NIDS are provided already ear-marked. A ctivities sometimes get introduced into the workplan because a new NGO or donor project has come into the district and has funds to address a specific problem. The recent introduction o f malaria control when funds became available through a vertical programme is a case in point. Some districts had started receiving funds for malaria control from the national programme and this was the only amount allocated for malaria even though as pointed out before the need is great. Another striking example is the controversial issue o f the money that was in itia lly provided

64


to districts earmarked for lunch allowance for hospital staff excluding other health workers. This generated a lot o f discontent among staff working in PHC till the policy was extended to include them. A ll respondents who expressed an opinion on this thought it was wrong to leave out staff in PHC units. The constraints to Rules I and 2 include - previous experience o f actual releases from centre to the district, quality o f team work and communication between the administrators and financial officers (CAO and CFO) and the health managers, and between the DDHS and other members o f the DHMT. It was pointed out by some key informants that sometimes funds arrive but the DHM T members are not aware o f them or what they are meant for. O f those who responded to the relevant questions, out o f 53 respondents, only 15 (28.3%) could mention figures for amounts they believed to have been transferred to the district from central government for health in the previous financial year, while 20 (37.7%) knew the district local contribution for health (most o f which was zero for non-salary items!). Fewer still knew o f the amounts that were actually released to the health departments from the central and district funds (9 out o f 44 or 20.5%. and 13 out o f 45 or 28.9% respectively). On whether they knew the percentage of the district budget allocated for health, 24 out o f 49 (49%) could mention a figure. Those who mentioned a figure for amounts spent on PHC from central funds were 22 out o f 55 (40%) while only 15 o f 53 (28.3%) could mention a figure for district funds spent on PHC. and only 12 out o f 53 (22.6%) mentioned a figure o f the amount spent on drugs and medical supplies. Most (48 out o f 51 or 94.1 %) o f those interviewed about the presence o f different accounts on which health funds are kept knew o f their existence and could mention examples o f such accounts.

5.4.3.3 Decision Rule 3: Availability of Guidelines I f guidelines exist about allocating to a budget item or activity and

65


- D H M T decision-makers are aware o f them ( if they are unaware, it is the same as if the guidelines were not there); - guidelines are understandable; - guidelines are perceived to be realistic; - there are censure measures in case guidelines are not followed; - decision-makers are w illin g to follow them; then they are most likely to allocate to that item as stipulated in the guidelines. An example o f this is provided by one o f the central MOH programmes, the District Health Services Programme (DHSP) funded activities. There are regular visits from program officials to review use o f funds and check the accounts, reports have to be submitted quarterly both financial and narrative. I f there are any doubts on the guidelines, the health managers can seek clarification from the visiting programme officers. I f the guidelines are not easily understandable and; - decision-makers are w illin g to follow them if they could; - they can get clarification easily e.g. on trip to MOH; - they can get clarification from visiting central officials; then they are most likely to allocate to the item (e.g., STI and reproductive health programme activities). Most respondents knew that there are guidelines for use o f funds from different programs implemented at district level. Forty seven out o f 51 answered “ yes" for DHSP (92.2%) as did 49 of the 51 for STIP (96.1%) programs as well as 48 out o f 52 for PHC conditional grant (92.3%). These three are the main sources o f funds for the health budget in all districts. Most o f them (DHSP, 46/50 or 92.0%; STIP, 48/52 or 92.3%; PHC, 45/51 or 88.2%) thought the guidelines are usually followed. A few, however, mentioned that sometimes the guidelines are not followed. Some offered reasons

66


as to why sometimes guidelines are not followed. These were mainly that funds may get diverted to other non-specified routine or emergency activities, usually by the district authorities. Others said some guidelines are unrealistic or are not clear. In regard to the block grant, some D H M T members thought that they would be better o ff if the guidelines clearly spelt out what % should go to health so that they would have all funds earmarked like it used to be in the past! They felt that leaving it open, the district authorities who control the disbursements do not give health it's due. There is the feeling that district authorities see that there are many donors supporting health but they argue that donor support is usually specific which leaves some aspects o f health not catered for. On the other hand, all the 5 Chief Administrative Officers spoken to pointed out that the money sent to districts under block grant is hardly enough to meet the existing mandatory commitments like paying staff salaries. They just do not have any money left over for service delivery unless they delay payment o f salaries. In addition many districts inherited salary arrears at the time o f decentralization which they have been attempting to clear. The same feeling was echoed during the MOH Annual General Meeting (AG M ) at the end o f 1998. In that forum the Permanent Secretary o f the MOH reported that proposals had been tabled to the central government to re-centralize salary payments as it is with primary education teachers operating at the same level as district health personnel. This step seems to go against the spirit o f decentralization which was to give responsibilities and authority to districts. However, it is being deemed necessary in order to guarantee continuity o f services and maintain a reasonable balance. Many districts are finding it a problem to meet their salary commitments and at the same time finance service delivery. The alternative is for the central government to send more funds to the district but this does not seem to be considered seriously as an alternative. The main problem seems to be the fact that capacity to handle personnel in a decentralized system is not yet functioning well in most districts. The impression one gets is that the complaint from health managers that district

67


authorities are not w illin g to fund health services is probably arising out o f inadequate communication or lack o f information flow.

5.4.3.4 Decision Rule 4: Number of People Affected This was perhaps one o f the most cited criterion for allocations in all districts. However, this rule operates in association with the others as pointed out before. For instance malaria is a common problem but fails on the other rules o f availability o f ear-marked funds. In all districts respondents revealed a concern for conditions that affect women and children and often cited the their being the majority, and their needing the services more because o f their vulnerability as the reasons.

5.4.3.5 Decision Rule 5: Problem Causes Death Problems affecting children are considered to be important because they may easily die from them. The same goes for maternal/obstetric problems. Problems were likely to be mentioned as priorities if the problem; - affects a large number o f people; - causes death quickly; - causes slow but sure death; - affects children; - affects expectant women. For example 53 out o f 58 (94.8%) would give immediate priority to cholera. 54 out o f 58 (93.1%) to TB. 47 out o f 58 (81.0%) to epidemic meningitis, 56 o f 57 (98.2%) to prevention o f maternal deaths/obstetric complications, 48 o f 58 (82.8%) to H IV /AID S , 57 o f 58 (98.3%) to prevent deaths o f children below 1 year and 52 o f 59 (88.1%) to diarrheal diseases. Respondents were given a mixed list o f diseases, and phrases descriptive o f common health

68


problems. They were asked to state whether in their view the particular problem was o f first or second priority as an item on which district health resources should be spent. A first order priority item was described to them as one which was sufficiently important and urgent that currently available district health resources should be spent in addressing it. A second order priority ranking was defined as those items which, although they may be a problem affecting people in the district, were not considered by the respondent to be serious or important enough to warrant spending currently available resources now, but could be considered in the future should additional resources become available. The interviewer ensured that what was being asked was clear before reading out the list o f conditions and then checking in the adjacent box corresponding to first or second ranking. The results supporting Decision Rules 4 and 5 are summarized in Table 5.4.1.

Table 5.4.1: Perceived Disease Problems and Priority Categories D IS E A S E O R H E A L T H S IT U A T IO N

F IR S T P R IO R IT Y

S E C O N D P R IO R IT Y

N

Acute respiratory infections (A R I)

46

12

58

Sexually transmitted diseases (STDs)

44

14

58

H IV /A ID S

48

10

58

Birth trauma and neonatal conditions

39

19

58

Low birth weight (less than 2.5 kg)

36

22

58

Death o f children below 1 year

57

1

58

Death o f children below S years

51

7

58

Intestinal worms/parasites

25

33

58

M alaria

57

3

60

Measles and other EPI immunizablc

54

5

59

Malnutrition

36

23

59

Low birth weight

28

28

56

Goitre (swelling in the neck)

10

48

58

Diseases

69


D IS E A S E O R H E A L T H S IT U A T IO N

F IR S T P R IO R IT Y

S E C O N D P R IO R IT Y

N

Heart Diseases

13

44

57

Maternal deaths and obstetric

56

1

57

Diarrheal diseases

52

7

59

Skin infections and conditions

10

48

58

Schistosomiasis (bilharziasis)

7

46

53

Trypanosomiasis (sleeping sickness)

10

39

49

Guinea worm

14

37

51

Onchocerciasis (river blindness)

22

33

55

S

49

57

Tuberculosis (T B )

54

4

58

Epidemic of Meningitis

47

11

58

Epidemic o f Cholera

55

3

58

Eye infections

24

34

58

9

3

12

complications

Filanasis (elephantiasis)

Other

Table 5.4.1: Continued from previous page. These conditions could be categorized as (a) prevalent and lethal, (b) rare but lethal, (c) prevalent but less lethal. Their reasons for assigning problems to first priority were as shown in Table 5.4.2.

70


Table 5.4.2: Reasons Why Respondents Assigned a Problem to First Priority Category (n=60) Frequency Mentioned

(%)

High Number Affected

55

91.7

High Demand by Public

21

35

Causes Quick Death

52

86.7

Affects Children

37

61.7

Affects Women

32

53.3

Affects Men

7

11.7

Effective Treatment or

30

50

Money is Available

10

16.7

Money is Earmarked

14

23.3

It is National Policy

24

40

Criteria

Control Measure Available

Other reasons stated for placing an item in first priority ranking included; “ it depends on the problems identified during supervision", “ high morbidity and m ortality” , “ affects production", “ undermines development” , “ easily spreads", “ it is infectious” , and “ for political reasons". The latter meaning that it is favoured by district politicians, probably as a way o f showing that they are doing something for the people they represent. These reasons reflected a concern for 3 things; (a) those problems affecting vulnerable groups (women and children), for which treatment or remedial action exists and which are in line with national policy, (b) the second was a concern for technical and financial feasibility, (c) a "political pressure” influences. In order to check for internal consistency, they were also asked why they placed an item in second priority category. Reasons included the fact that those conditions affected few people, were

71


not a serious problem affecting vulnerable groups, had no money earmarked for them, or were not in line with national policy. Responses confirmed that they had understood the questions and were generally consistent with themselves in their answers. Other reasons offered for placing an item in second priority ranking were; “ chronic condition” , “ low morbidity and mortality” , “ there are currently enough efforts to contain the problem so it is not a threat” , “ feasibility o f control is poor” , "it is a localized problem” , "does not affect a productive group” , “ it is due to poor sanitation". (This latter respondent meant to say that if the sanitation problem is solved the problem w ill disappear, since there is a nation-wide sanitation campaign, they felt it is being addressed). Others said, “ it is preventable through health education” (same rationale as for sanitation-related problems).

5.4.3.6 Decision Rule 6: Demand by Influential Persons/Groups If an item fu lfills the follow ing conditions, it is most likely to get funds allocated if it; - w ill be seen to answer a popular demand (e.g.. reduce crowding at the hospital as in District 23); - is highly visible as use o f public funds (e.g., physical structure extension or rehabilitation as in District 41); - is politically popular, high authorities emphasize it (e.g.. routine immunization in the past, now NIDS and sanitation in all districts, demand to build a government facility in an area without one even though it is well served by existing NGO facility as reported by key informants in districts 22 and 24); - w ill increase chances o f politician’ s re-election (e.g., had made a promise during

72


campaigns to do it as in District 23, or it is an election y e a r1. Most district politics revolves around issues o f health, education, and roads). - no expected adverse censure effect foreseeable from the central government or donors14.

In all districts large sums o f money are allocated for “ rehabilitation" o f health units. A new hospital ward is being built in one district even if it is not technically necessary and in direct contravention o f M O H guidelines. In all districts allocations for NIDS/Polio campaign generally exceed amounts budgeted for entire routine immunization program (Appendix I). Next to the number o f people affected, the second main factor is popular demand. This could be labeled the “ local political pressure” factor. Other comments from respondents were; “ allocations are based on favouritism", “ it depends on the other urgent district priorities e.g., roads, schools, security, paying allowances o f councillors, etc” (presumably this diverts resources from being allocated to health programmes).

5.4.3.7 Decision Rule 7: Expected Personal Gain Decision-makers are more likely to allocate funds for those activities in which they expect direct personal benefit. The benefit may be monetary, political, or exchange o f favours. There are many opportunities for conflict o f interest between professional and o fficial duties and the expectation o f personal gain in the face o f rising costs o f living. Such benefit is often in form of allowances for field activities or perdiem during trips out o f the district or sitting allowances in meetings or perdiem for training courses. This is especially likely for those activities or funding

2.1

This tendency was reported by key informants in all study districts for 1998 which was an election year for local councils in Uganda. 24

This situation often arises due to lack o f clear legal stipulation o f censure measures for aberrant districts in the relevant legal documents governing decentralization.

73


sources for which verification mechanisms are weak or there are no censure measures for doing it. Activities for which accountability in form o f health outputs or outcome results are not demanded but which pay allowances w ill get preference. A ctivities in which a transaction might result in a “ hand-shake"25 are also more likely to get funds. Examples o f this rule are; - most districts had planned between 20-70 training courses or workshops/seminars; - large sums far in excess o f M O H estimates26 allocated for “ hospital cleaning" (these are generally locally contracted out, sometimes to groups operated by district officials or politicians as reported by key informants in districts 11, 22, 23, 24, and 42); - high allocations for food/wood purchases (one hospital had allocated about 30 m illion shillings for this item alone, about 20% o f it's budget!); - a number o f hospitals had large amounts allocated for paying staff incentives (up to 100 m illion shillings in District 23. about 50% o f it ’ s official salary budget, and about 300 m illion in districts 31 and 32, also about 50% o f their official salary budgets. These amounts are mostly generated from user-fees. Guidelines on use o f user-fees generally recommend using 50% o f collections for staff top-up and the rest for improving quality o f service delivery such as to provide additional drugs and supplies. Since complaints o f drug shortage are still wide-spread, we may conclude that most user-fee money is going to the staff themselves. In the latter 2 districts, there is a major donor funded project going on. and they were reported to pay incentives to top-up o fficial salaries).

25

Terminology used in Uganda to refer to kickback payments e.g.. on supply o f office materials or drugs to a department, the officers who approved the transaction might expect a reward payment from the person who got the contract. This is done so that they can get more contracts. For the officer, it is in order to supplement their income. There is an official tendering process but key informants generally thought that unofficial payments are exchanged. 26

The M O H suggests not to allocate more than about 5 m illion shillings (M O H , 1998) but allocations o f up to 15 m illion shillings or more were observed in some budgets. It is possible that some o f these were masking minor renovations. 74


- political pressure in districts to allow local business owners to supply drugs to districts and not have to get them from the National Medical Stores; - high allocations for ‘‘rehabilitation” o f units which get contracted out (often at the expense o f service delivery); - district councillors being suppliers to the district departments or getting contracts for services, e.g., in one hospital in District 22, the Chairman o f the Hospital Management Committee was also the one supplying stationery! An indication that people are getting sensitive to these things is that once this became known publicly, the contract was terminated (key informant). Health managers may be influenced in their allocation decisions by their perception o f the concern about services among the people they are supposed to serve. They may perceive this as o f potential benefit or threat to their own career advancement depending on whether they are seen to be doing a good job or not.

S.4.3.8 Other Influences on Allocations This section reports on the results o f questions exploring other influences on resource allocation. In order to get the decision-makers’ perceptions o f how concerned they thought the public are about these services, they were asked to rate how likely it is that members o f the public might judge their (decision-makers’ ) performance in relation to each o f these services. They were asked to think in terms o f their popularity, public standing or prestige, effect on career, etc in a general sense. They then came up with a score on a scale o f 1 (very unlikely) to 5 (very likely). A rating of 4 to 5 was considered to be a perception o f high concern. This was done separately for their perceptions in regard to rural and urban people. The results are summarized in Table 5.4.3.

75


Table 5.4.3: Health Managers Perceptions of Public Concern on Selected Items PERCEIVED AS HIGH CONCERN (RURAL)

PERCEIVED AS HIGH CONCERN (URBAN)

Frequency (7c)

Frequency (7c)

Mother and Child Health (MCH) services

40(71.4)

41 (72.0)

Immunization (EPI) services

47 (82.5)

44 (75.8)

Health Education services

29 (50.9)

33 (57.9)

Family Planning services

17(30.4)

33 (57.9)

Treatment for malaria

48 (84.2)

46 (79.4)

Treatment for ARI

35(61.4)

37 (63.8)

Treatment for injuries/wounds

32 (56.1)

35 (60.4)

Treatment for pregnant women

40 (70.1)

44 (75.8)

Transport to hospital when needed

41 (72.0)

39 (67.2)

Drugs available at health centres

47 (82.5)

43 (74.2)

Affordable treatment at health centres

43 (75.4)

43 (74.1)

Drugs available at the hospital

44 (77.2)

46 (79.3)

Affordable treatment at the hospital

38 (66.7)

39 (67.2)

Treatment for intestinal worms/parasites

24 (42.1)

24 (41.4)

Kind and attentive staff at health centres

44 (78.5)

44 (77.2)

Kind and attentive staff at hospital

41 (71.9)

46 (79.3)

Reasonable waiting time at health centre

43 (75.5)

47 (81.1)

Reasonable waiting time at hospital

37 (67.3)

46 (79.3)

SERVICE/CHARACTERISTIC

Items representing indicators o f quality o f access, interpersonal relations with staff, and availability o f curative services seem to be the ones perceived by the managers to be o f most concern to members o f the public. It was generally felt that preventive services are o f little concern to the public in rural areas. For urban people, respondents felt that while they consider most services to be important

76


(unlike rural residents) in urban areas they are much more concerned about hospital based services and availability o f drugs. They were finally asked to state their own valuation o f the importance o f their job to them in terms o f jo b security, prestige, salary/allowances, ability to help relatives (e.g., to get them a job) and the ability to increase their private practice or business. (The degree o f importance was estimated by assigning a weight on a scale where 1 is very low importance 5 is very high importance). The responses are shown in Table 5.4.4 (M ultiple responses).

Table 5.4.4: Respondents Perception of Important Causes of their Job Satisfaction (n=66) ITEM

FREQUENCY

%

Job security

56

84.8

Job prestige

40

60.6

Salary and allowances money

49

74.3

Ability to help relatives

11

26.7

Ability to help personal business

16

24.2

It is possible that allocations to different items are influenced by the perception o f importance o f the item in running health services. To explore this idea, respondents were asked to categorize the importance o f budgeting for each item as follows:

V. "V ita l” - item must always be budgeted otherwise service w ill be considered to have stopped, a serious breakdown in the departmental function; E. "Essential” - item is usually necessary for routine function, should be budgeted for normal

77


operation o f good (acceptable) quality services; N. “ Nice” - “ it would be nice” , item should be budgeted i f first class provision o f services (above average quality) is expected). Across ail districts, items considered vital for the functioning o f the Health Department as defined above included epidemic preparedness, promotion o f safer sex, staff recruitment, maintenance and repair o f vehicles, STD treatment, school immunizations, monitoring and co ordination meetings, staff salaries, TB/leprosy control, procurement o f drugs, family planning and supervision o f peripheral health units. In the context o f decentralization districts are supposed to be responsible for delivery o f health services. In order to explore whether they really feel in control, they were asked about each item in the budget how much they felt in control. Control was defined as follows: F - Yes, managers have full control (can decide to budget or not, including amount); L - Limited, managers have limited control (e.g., item is decided by the centre but local managers decide on amounts disbursed); N - No. managers have no control whatsoever, item is provided for by the centre, including amounts). They were generally unanimous in feeling they have no control over lunch allowance, salaries, rehabilitation o f health facilities, and training. Other items over which most felt they had no local control over budgeting was for supervision o f health units, malaria control, fam ily planning, M CH services, routine immunization. National Immunization Days (NIDS), STD treatment, TB/leprosy control, and epidemic preparedness. O f this list it is only National Immunization Days/polio eradication, training and epidemic preparedness which are still o fficia lly national functions for which the M O H has primary responsibility along with setting national policy and monitoring compliance with standards.

78


Limited control was thought to be exercised jo in tly with central authorities on money for community outreach activities, travel allowances, maintenance o f buildings, vehicle maintenance, office equipment and supplies, and staff uniforms. As to whether actual payment out o f district block funds or local revenue is effected or not depends on the discretion o f the Chief Administrative O fficer (CAO ) and the C hief Financial O fficer (CFO). The time is often shorter for activities whose funds are placed on separate accounts, hence the observation that most districts operated on average 6 bank accounts each for Health Department funds! Donors in particular prefer that District Health Offices operate separate accounts for their funds. The new proposal o f pooling funds together under the Sector Wide Approach (SWAP) is aimed at streamlining this. Items for which it was judged that money was likely to be released i f the item appeared in the budget included promotion o f safer sex, promotion o f water source protection, fam ily planning. M C H services, training, monitoring and evaluation meetings. NIDS/polio eradication campaign, hygiene education, health education and community empowerment programmes, inspection o f drugs shops, community care o f AIDS patients, STD treatment, and epidemic preparedness and control. Items for which it was felt unlikely that funds would be released even if they were planned included staff incentives and uniforms, staff recruitment, community based rehabilitation, developing units in under-served areas, and control o f locally endemic problems that are not covered by national programmes such as trypanosomiasis, schistosomiasis, eye and skin diseases. Finally, funds may eventually be released but not according to the planned schedule. To test this idea and identify items that are likely to be released in time and those likely to delay, they were asked to indicate their subjective opinion based on their experiences o f how likely it was to obtain funds for planned activities in time for scheduled implementation. They used the follow ing rating: 1. very unlikely - funds almost never get released in time, only very rarely are they paid in

79

Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.

time; 2. unlikely - in many cases funds don't get released in time but sometimes they are provided in time; 3. "half/half” 27 - sometimes they get released in time and equally likely, sometimes they don't get released in time; 4. likely - in many cases funds are released in time but sometimes they are not released in time; 5. very likely - funds are almost always released in time, except in very exceptional circumstances are they not released in time). Items likely to be funded in time were condom procurement and distribution, promotion of safer sex. effective STD treatment, school immunization, community and home care o f AIDS patients. TB/leprosy control, HIV/STI control, and office supplies. Those on which there was general consensus as most likely to delay included staff incentives, staff recruitment, maintenance o f buildings and equipment, promotion o f service provider collaboration, support to user-fees. IM CI. malaria control, lunch allowance, support to those with physical disability, drug procurement, and dissemination/enforcement o f treatment guidelines, and locally endemic special health problems. Assuming that the funds are adequate and all the requirements o f personnel, materials, and time for activity implementation are met. the planned intervention may get carried out. The main constraints w ill then include the right knowledge and skills for carrying out those tasks and ensuring proper financial accountability. It is not uncommon for funding sources, either M O H programmes or donor agencies like UNICEF to delay disbursement o f the next funds if accountability for funds is not satisfactory.

27

A Ugandan expression that means the situation is neither good nor bad.

80


They were asked to rank as first or second priority a selected list o f functions or activities that are supportive to the health system. The results are depicted in Table 5.4.5.

Table 5.4.5: Health Support Services Perceived as Priorities FIRST

SECOND

PRIORITY

PRIORITY

Paying lunch allowances

35

23

58

Paying salaries in time

57

1

58

Paying adequate salaries to health workers

55

5

60

Maintenance of health department vehicles

45

13

58

Provision of adequate drugs and supplies

59

1

60

Support supervision to peripheral health units

46

13

59

Maintenance of medical equipment

46

13

59

Maintenance of health unit buildings

16

42

58

Maintenance o f health unit compounds

17

41

58

Paying allowances for outreach activities

40

18

58

Maintenance of health worker's bicycles

31

27

58

ACTIVITY/FUNCTION

(N)

Other important supportive items volunteered as priority areas by respondents were: "do on-job training for s ta ff', "improve the planning process” , “ provision o f a vehicle", “ supply medical equipment to health units” . Respondents were given a list o f service attributes to rank. They were asked to indicate which o f the services they felt was first priority (worth spending the currently available public resources available from government or NGO now) or second p rio rity (item maybe important but would not spend currently available resources on trying to achieve it now). Their rankings are shown in Table 5.4.6.

81


Table 5.4.6: Respondents Priority Categorization of Selected Health Services Attributes SERVICE/CHARACTERISTIC

FIRST PRIORITY

SECOND PRIORITY

(N)

Mother and Child Health (MCH) services

57

1

58

Immunization (EPI) services

55

3

58

Health Education services

46

13

59

Family Planning services

39

18

57

Treatment for malaria

56

2

58

Treatment for ARI

51

6

57

Treatment for injuries/wounds

30

25

55

Treatment for pregnant women

57

0

57

Transport to hospital when needed

42

12

54

Drugs available at health centres

57

0

57

Affordable treatment at health centres

50

8

58

Drugs available at the hospital

50

7

57

Affordable treatment at the hospital

43

14

57

Treatment for intestinal worms/parasites

27

30

57

Kind and attentive staff at health centres

45

11

56

Kind and attentive staff at hospital

41

16

57

Reasonable waiting time at health centre

45

12

57

Reasonable waiting time at hospital

41

16

57

Equity Concerns Respondents know that rural people are less likely to be covered by services. In all cases they mentioned a lower estimate for rural areas compared to urban. An awareness and concern for equity was revealed during the key informant interviews. Most Key Informants felt that disadvantaged people are not being treated as fairly by the health system as they would like to see. Examples mentioned included the fact that poor rural people do not get any preferential exemption from paying say at the hospital or they have to pay the same as urban people even though they have 82


spent more on transport in getting to the hospital and are generally poorer.

Perceptions o f U tility It may be expected that managers have their own assessment o f how important a service is and this may influence the degree o f importance (and hence merit for more resources) that they attach to it. This may be translated into a favourable or unfavourable allocation. To explore this idea, the respondents were asked, using exactly the same list as before, to indicate their assessment o f the degree o f importance o f each o f these services for rural people, then again for urban dwellers. Respondents seem to place most emphasis on availability o f affordable curative care and M CH services at health centres and transport to the hospital for rural residents. Surprisingly, they did not place as much importance on providing for treatment for pregnant women, providing family planning services or treatment for malaria in the budgets. (When asked why. a common comment was that anti-malarial drugs are readily available and they can find them cheaply so it is not so important to spend district funds for this). Similarly, they also did not place much importance on providing for EPI services for rural residents. In recent years all districts have experienced falling immunization coverage rates. This followed the devolvement o f responsibility for routine EPI funding to the districts while the centre retains responsibility and provides funding for the polio eradication campaign. It is possible that this might have sent a signal to district managers that routine EPI services are not a priority anymore! They also did not place importance on providing for access to hospital services for rural people. They, however, considered all services as being important for urban people, revealing their own urban bias. Their perception o f importance o f services for urban people was unidimensional (only one underling factor) while for rural people there were two factors.

83


Strategies for Improvement Current resource allocation decision making can be improved using the strategies outlined in Table 5.4.7 which is a summary o f the main decisions that are being made at the district level and the ways to improve them.

Table 5.4.7: Main Decisions for Resource Allocation DECISION TO BE M A D E

NATURE OF DECISION

LE V E L FOR M A K IN G DECISION

W AYS TO IMPROVE DECISIONS

♦Staff recruitment

special

district

♦Staff training

routine

im provem ent o f staff working conditions

special

♦Reproductive health services

routine

district/ national National programs district

staffing ratios. compare with national

♦Procurement o f essential drugs/equipment

routine

district

♦Purchase o f supplementary drugs/supplies

special DHMT

♦Decisions about interventions - program alternatives - program expansion (e.g.. EPI, FP coverage, etc) - etc *PHC extension activities -E P I - health education - support supervision - etc ♦Allocation to hospitals vs PHC ♦Which hospital services to fund ♦Facility repair and maintenance ♦Vehicle repair and maintenance ♦Equipment repair and maintenance ♦Provision o f transport assistance from rural areas ♦etc

special routine

routine

special routine special routine routine special

central district district district regional district

Clear guidelines. Integrate Inventory management disseminate drug studies/guidelin es

develop health systems reference frame, need o f data on health status. effectiveness o f interventions, costs, society values and preferences Functional analysis

84


5.5 Discussion According to James March (1985) decision-making can follo w existing institutional or organizational rules or be based on the choices people make. The choices are determined by the organizational culture o f the institution. In Ugandan districts health departments generally follow some decision rules based on their perceived interests ranging from political, personal, to the desire to comply with national policy. Resource allocation decision-making is generally not based on objective criteria. The main problem appeared to be lack o f skills and capacity to manage a decentralized health system resulting in a haphazardly functioning district health system. Lack o f awareness or consideration for equity in the system is shown by the fact that poor rural people do not get any preferential exemption from paying say at the hospital or they have to pay the same as urban people. People from rural areas face many challenges in accessing health services. Examples o f attempts to promote equity and rationalize health care utilization and financing system have been reported from the former Zaire28. In Kasongo and Bwamanda respectively, a lower fee was charged to patients who had been referred than for those who by-pass the First Contact Level units (Health Centres) (Van Lerberghe and Pangu, 1988; Criel, Van Der Stuyft. et al, 1997). A ll respondents were asked to rank the list o f disease and health conditions including health technocrats, administrators, and politicians. It is quite likely that the true implications o f some of these conditions were not fu lly appreciated. An example is a check item “ low birth weight” which was deliberately entered twice in the list and drew separate rankings first 62.1% indicating it as first priority, then 50%. This indicates the fact that respondents perceptions are not fixed and that for some o f these items their true rankings can only really be determined after a series o f repetitions till responses are stable. These items that do not differentiate well need to be removed from the list if a scale is to be constructed using such items. It was not the intention o f this study to fu lly develop

28The country changed it’s name in 1997 to the Democratic Republic o f Congo (D R C )

85


assessment scales and to do so was impractical given the scope o f the study. Some respondents on being probed as to their perceived importance for services for rural people mentioned that it is not so important to have treatment for pregnant women at health units since they are well catered for by TBAs. This is a sign that new policies meant to supplement existing programmes may actually end up being perceived by district managers as alternatives. Since they are likely to seek to minimize their activity loads, this might offer a convenient excuse to do less and not more. Most "rehabilitation" favoured by politicians and project officials is a physical structure rehabilitation emphasizing extensions, painting, fittings, etc. Functional rehabilitation is often not seriously considered or planned for. Sometimes one finds magnificent looking health centres that are under utilized. The attractiveness o f physical works is clear for local politicians seeking reelection or donor funded project officials to show "achievements". Sim ilar observations have been made in other developing country contexts (Lafond, 1995). A clear challenge is how to move beyond the physical structure to analyze the performance o f the health system. It is hoped that this process can begin by elaborating a set o f simple resource allocation performance indicators as attempted in the study. The fact that the M O H is not empowered by the current decentralization law to take any serious censure action against districts that do not comply with government guidelines is a serious threat to rational and efficient use o f the meagre resources available. The potential for resources to be used for self-serving interests o f local district health managers and politicians is high. As one district health official commented “ there is decentralization from the centre to the district but now there is 'centralization’ at district level!” This observation is in line with decentralization experiences elsewhere reported by Collins and Green (1994). The Health Sub-District (HSD) policy is aimed at streamlining this by further decentralizing from the district to an electoral constituency.

86


The policy is envisaged to take effect from July 2000 but funds are already being made available to districts to upgrade health facilities that w ill act as the focus o f the HSD. It is not yet clear how the current D H M T w ill evolve to support the new HSDs. Problems o f transport, staff, logistics, financial management, etc w ill have to be faced and solved. There is an urgent need to prepare the prospective HSD teams through some sort o f emergency training. There are 214 HSD in the country so the requirements for building capacity are enormous. There is a different perspective between the central government ministries for health and that for local government. W hile the M O H feels districts should be closely supervised to ensure equity and quality o f services in order to maintain standards (and thus fu lfil their constitutional mandate), the M O LG feels that would defeat the purpose o f decentralization. The M O LG is said to feel that M O H is being paternalistic in dealing with districts. The M O H may give advice but that districts are perfectly legal if they disregard such advice or seek expert opinion elsewhere. This institutional “ turf fight” between the two “ elephants” w ill only leave the “ grass", the districts, suffering the consequences unless they are addressed at the highest levels. The legal loopholes and ambiguities that exist should be reconciled. Decentralization actually substantially elevated the status o f the M O LG at the expense o f other central government ministries. Cutting the powers o f districts would not only be undermining the cherished political goals o f decentralization, it would mean the MOLG giving up some o f it's newly acquired prominence. Local political pressure in resource allocation is a significant factor. Whereas resource allocation cannot be entirely technical since by it's nature it is partly political, it would be in the districts best interests to reach a balance between technical criteria and political ones to achieve an optimal balance. In order for this process to happen, information should be made available to those who make the decisions. During the years leading up to Ugandan independence, one British colonial governor in Uganda, in arguing in favour o f education for the locals is reputed to have said, “ if they

87


are going to be our partners in negotiation for independence, let us at least have informed opponents!" Similarly, the central government should give priority to building up capacity to make informed decisions at the district level rather than seeking to retain influence by retaining central control. As pointed out by Alexander and M orlock (1994) political criteria in decision-making and conflicts gain prominence in absence o f well established technical criteria. While it may be impossible to achieve a high degree o f rational decision-making, we can increase the chances o f success and tip the balance towards more objective decision-making by having clear objectives, knowing what the choices are, and which alternatives are available in improving health status. We would like to argue here that a better health status is possible within the constraints o f available resources since other countries in the region, with a similar economic environment like Uganda, have better health indicators as shown in the W orld Development Report 1993 (The World Bank, 1993). Similarly, it is observed that districts from the same region, at the same socioeconomic level (as shown by Income Index) are seen to have different life expectancies. There is very little information in Uganda on the field effectiveness o f the many innovative interventions being implemented in different districts. Using some o f the resources being expended on such interventions to evaluate them would seem a reasonable investment in order to guide future choices. The scarcity o f meaningful public debate is tied in with the overall political climate in the pace o f democratization in the country. Issues like equity are o f concern to even ordinary citizens. Mechanisms need to be found to promote the involvement o f the general public in giving a voice to their concerns on what is being done with their tax money. As pointed out by Lafond (1995), the general public are often left as passive spectators in the struggle to apportion the scarce health care resources. In the process, allocation practices tend to be skewed by powerful professional lobby

88


groups such as the medical, dental, nurses/mid-wives, and pharmacists associations. Moreover, such pressures often result in allocations that are neither favourable from the point o f view o f allocative efficiency, nor are they equitable. The society's perspective must be taken into account as the Panel on Cost-effectiveness in Health and Medicine recommends. However, this has implications in that the society whether explicitly or im plicitly, has value judgements on different states o f health, and for which groups (Russel, Gold, Siegel et al, 1996). In Uganda, the debate on such issues is still in it's infancy. In talking with respondents during the study, there did not seem to be any hesitation among the study respondents that ways should be found to cater for the more disadvantaged members o f society. Most o f them indicated clear criteria that could be used for identifying such people. However, when it comes to collective decision-making, those whose interests should be protected in the promotion o f equity are often under-represented. In addition the short political life span o f the local councils (3 years), makes them strive to show their electorate that they have achieved results by showing o ff physical constructions, extensions o f existing facilities, or "rehabilitation” o f infrastructure. This is not a problem unique to Uganda. It has been reported in connection with health planning in general, international aid, and the sustainability problem in other developing countries (Green, 1995; Lafond, 1995; Cassels, 1996). The influence o f donors in developing national health policy and priorities for funding certainly influences the decision-making process. A striking example o f such donor influences has been provided by a review paper based on Uganda as a case study (Okuonzi and Macrae, 1995). The feeling that was expressed by some DHC members that they are merely “ rubber stamps” in approving health plans and budgets is not supported by any o f the documents or guidelines that were reviewed. O fficials hinted rather, that the problem is the lack o f capacity in the councillors themselves to comprehend the complex nature o f the subject o f resource allocation since it is a relatively new experience for them. Ultimately, they can only feel ownership o f the process if they

89


are presented with more facts than those currently available to them. This raises the importance o f information that should be generated and packaged in such a way that the ordinary citizen can start to participate in the dialogue. Information that already exists or could be generated from existing sources as was obtained in this study is not being utilized to support decision-making. It is true that in general terms data is scarce but what is available has the potential to provide useful insights to the decision-maker who knows how to obtain the relevant information. The question o f information for decision making is dealt with in the next paper (Chapter 6).

5.6 Summary and Conclusions It would be fair to say that legally and o fficia lly the centre has devolved the power to allocate resources to the districts albeit within tight bounds. However, the players at the district level mostly are hindered by some or all o f the following factors: 1.

Lack o f confidence; - arising out o f lack o f data management skills - in the reliability o f data that is available.

2.

Lack o f conducive systems for data collection, storage, analysis and retrieval.

3.

Work overload due to; - too many training/seminar/workshop activities leaving little time for other implementation either because staff are conducting a training or are themselves being trained outside the district. - strict bureaucratic requirements by central government and donors. - an overflow o f visitors from the centrally managed vertical programmes or other visiting government officials.

90


In order to improve resource allocation decision-making, the issues o f information management and a well functioning district health system are crucial and w ill be explored in the second and third papers respectively. The process through which resource allocation decision making currently takes place has been described and preliminary investigations done o f techniques and items that could be used for further research. The main areas for improvement are summarized in Table 5.4.8.

5.7 Policy Implications

Tabic 5.4.8: Issues Influencing Allocation Decisions and Possible Decisions/Actions to Correct Them MAIN ISSUES

POSSIBLE DECISIONS

LEVEL FOR ACTION

RESPONSIBLE PERSONS/ ORGANIZATION

♦What criteria are being used

generally guidelines are not well disseminated, not known by DHMT or DHC

Central

MOH. MOFEP. MOLG

♦Political concerns

Need legal provision in Local Government Act of censure measures

Central

-do-

♦Staff recruitment, deployment, development and retention

Need to address staff inequity

District but needs central help

DDHS. CAO. DSC. HSC

♦Personal gain

Improve staff conditions

District but needs central intervention

-do-

♦Prevalent conditions

Need of skills in use of available data, and on acquiring new data on effectiveness of approaches, costing

Central with help of educational institutions/c onsultants

91


CHAPTER 6: DEVELOPING INFORMATION FOR RESOURCE ALLOCATION DECISION-MAKING (Paper 2)

6.1 Introduction This article w ill examine the information already available and the requirements for Ugandan districts if decision-making for resource allocation at district level is to be improved. The paper w ill provide insight into the follow ing questions that guided data collection: How could resource allocation decision-making be improved at district level in Uganda? What information is currently available or obtainable? What is the pattern o f distribution o f resources between PHC services and hospitals as well as between different health interventions? The paper w ill present evidence to support the view that major impediments to improving allocation decisions are lack o f information, lack o f skills, lack o f time and lack o f w ill. Weakness o f management structures is another major impediment but is addressed in the third article (Chapter 7). Information that is available could be used to improve on current practices that have been presented before (Paper 1). The conclusion is that improvement is possible within current constraints i f basic data management skills and systems are improved at the district level. The word improvement here is taken to mean a better health outcome for the district as a whole. The final goal is to achieve a better health status as measured by e.g., a higher life expectancy or a composite indicator that includes both mortality and morbidity such as the D A L Y or HE A LY . This paper explores the available information and identifies the gaps and strategies for acquiring information for use in district level resource allocation decisions. It offers a snapshot o f the issues that would be faced in attempting to use a needs-based approach such as the burden o f disease and cost-effectiveness approach to resource allocation in Ugandan districts. This paper does not review 92


the burden o f disease methodology as such. The reader who would like to get a good description o f the methodology can find that elsewhere (Ghana Health Assessment Project Team. 1981; Morrow, 1984; Murray and Lopez, 1994; Hyder, Rotllant and M orrow, 1998). A theoretical framework for rational decision-making using expectation o f u tility based on the approach suggested by Dennis Lindley (1985) is attached as Appendix E.

6.2 Background and Objectives Following decentralization o f all social service provision to the district level in 1993. districts were faced with new challenges o f planning, and overall management o f health services. Functions that used to be done by the central M O H were now devolved in a phased manner to district level over the next 3 years. However, the management structures at district level remained essentially as they had been before decentralization. Health managers at both district and national level expressed the need for information as to what was working and what needed to be improved. Management o f resources is one o f the main challenges districts found themselves facing. This study was conceived in order to make a contribution to promote better management o f the scarce health resources. A general introduction to the study, including literature review have been presented in the first part o f this manuscript. The study aimed to achieve the following purposes by the end o f December 1998:

6.2.1 Overall Goal of the Study: To review the process o f decision-making for allocation o f public health resources in decentralized districts in Uganda in order to provide a basis for development o f allocation procedures for district health systems that address health needs o f the people and promotes rational use o f scarce resources.

93


6.2.2 Study Objectives: 1.


2.

Identify the formal guidelines/criteria that are currently existing in the districts with regard to allocation o f public funds to health activities, if any, and the informal influences that play a role.

3.


4.

Assess the extent to which decisions being made in allocation o f public health funds relate to the health problems perceived as priorities (e.g.. as defined by the decision-makers' perceptions and based on national guidelines such as the national basic health services package) as well as reasons for any observed deviations.

5.


Paper 1 presented findings and issues related to Objectives 1. 2. 3 and 5. This second paper w ill present findings relevant to Objectives 3,4 and 5. Findings and issues related to Objective 5 as well as broader contextual issues in Ugandan districts w ill be addressed in Paper 3.

6.3 Subjects and Methods The information needs o f districts were explored by means o f a semi-structured in-depth questionnaire interview with members o f the District Health Management Team. They were asked to mention those activities for which they currently use data and i f they would like to obtain more data. They were asked to identify potential sources o f such data. The existing data collection was

94


assessed by observing the state o f records at the district offices as well as analysis o f existing data obtained from the records kept at the District Health Office. A variety o f allocation indicators were constructed for monitoring and assessing district performance (Appendix H). Current budgetary allocation practices were studied by reviewing health plans and budgets covering the two Financial years o f 1997/98 and 1998/99. These two years were chosen because they were the most recent for which records and documents were most likely to be available. In general terms they were representative o f existing practices even though 1997/98 saw the introduction o f additional funds in form o f conditional grants for PHC. Financial indicators were standardized into constant 1998 Uganda shillings20, using the Consumer Price Index figures for the respective years™. Relationships between allocation indicators and possible influences were explored using correlation analysis. The indicators were chosen from variables that were meant to show how oriented to PHC the allocations were. The following were used; (a) proportion o f discretionary total district health budget allocated for PHC. (b) proportion o f discretionary district recurrent health budget allocated for PHC and (c) per capita allocation for non-salary recurrent PHC budget and (d) per capita district recurrent non-salary budget allocated for direct and integrated PHC service delivery11. Money allocated for supplementary drug procurement was also taken as being direct service delivery as were mobilization, EEC and health education, etc. A ll financial information was standardized in terms o f constant 1998 Uganda shillings using the Consumer Price Index and

29

The symbol used Cor the Uganda shilling is /= . I US$ = 1.000/=. 30

C P I used for 1997/98 was 352.2 and that for 1998/99 was 354.9. The data was obtained from the Government Official Statistics handbook published annually by the M inistry o f Finance and Economic Planning (M F E P . 1998).

31

as opposed to vertically implemented programmes and other largely supportive allocations like general administration and logistics.

95


adjusting for different district populations by using per capita figures. The information on district financial allocations was obtained from a detailed and careful analysis o f approved district health work-plans and budgets. It would have been ideal to use actual expenditure data for different budget items but this was found to be impractical since expenditure data was largely missing. There is usually a time lag (could be up to 3 years) before audited, itemized final accounts are produced. Since the object o f the study was decision-making for resource allocations, it was deemed that an analysis o f the intentions as reflected in the plans and budgets would fu lfil the purpose o f gaining insight into the influences on the process. However, information on actual payments to health departments in the districts by broad category o f source, i.e., central, donor, and district contributions are generally accurate since they are based on government audited reports. These were obtained from the M inistry o f Health (M O H ) planning department. Missing values on some indicators for two districts were replaced in all the analyses done by using means computed from the other values. Extreme outliers were omitted. In the case o f one district where budget financial allocation data for 1997/98 was completely not available, its 1998/99 data was used instead. The potential bias from this was minimized by the fact that 1997/98 financial data were all converted into 1998 constant shillings.

96


6.4 Results A complete list o f district PHC and hospital budget variables is in Appendix G. Data that was obtained or calculated for selected district resource allocation indicators is presented under Appendix H. Other district specific Financial indicators are in Appendix I.

6.4.1 Main Findings Key findings included the following: - An average o f 59.1% o f total health budgetary allocations were devoted to public health/PHC (non-hospital) services (Standard Deviation 14.7%). I f personnel remunerations are excluded, the average is 70.9% (Standard Deviation 13.0%). - An average o f 66.5% (Standard Deviation 13.7%) o f district recurrent non-salary budget was allocated to PHC. However, if personnel remunerations are included, the average allocation for recurrent PHC budget as a percentage o f total recurrent budget is 54.0% (standard deviation o f 12.8%). - The average PHC capital allocation as a percentage o f total capital allocation was 80.7% (standard deviation 27.1%). - Political concerns, leading to allocation for popular infra-structure oriented programs at the expense o f service delivery; and concern for curative care vs preventive, predominance o f vertical programmes vs integrated. - Lack o f use o f information on service coverage and health status on which to base decision-making. Even existing information is not utilized. It was found that technical health personnel were often vulnerable to act on an a d hoc basis and under influence o f demand for services and local political pressure. The lack o f objective data aggravates this tendency. - Lack o f skills o f staff to use the information that may be available through routine data

97


collection. Information on disease prevalence, incidence, etc though not very accurate, might be obtained from the data that is routinely collected by the health system. Often, health managers do not have skills to analyze and interpret data. - Information on EPI coverage rates is generally available and validated regularly through independent national and district surveys. - Many studies done within the country especially by university based faculty and students generate a mass o f information every year that is grossly under-utilized. Apart from studies done by students in the MPH programme at the Institute o f Public Health at Makerere University12 there was little evidence o f dissemination or use o f such information in the districts in which the studies are done. The IPH has started to compile reports o f studies done by its students and intends to regularly produce and disseminate monographs from them. An example o f the lack o f use o f existing information generated from studies is provided by the case o f rational drug use which is a problem in many districts. Studies have been done on rational drug use and the problems and issues have been documented but district managers were often heard expressing confusion as to what the problems are and how to resolve them. Many studies have sim ilarly been done on nutrition. STDs, and HIV /AID S . However, dissemination o f these findings outside the circles o f the concerned project or those in academia remains very limited. On a number o f occasions districts plan to repeat similar studies done elsewhere in their own districts without a clear plan o f how differently the results w ill be used.

32

The M P H Programme at Makerere is a two-year field based programme which emphasizes “hands-on" training. Part o f evaluation o f students is on basis o f use or potential for use o f research findings. The programme is part o f the Public Health Schools W ithout Walls (P H S W O W ) initiative which was started with financial support from the Rockefeller Foundation.

98


6.4.2 Use of Data for Allocations and Payments On whether data is used in decision-making regarding health budget allocations and payment o f funds, respondents were asked to state separately whether data is used for allocations and/or payments. On which sources o f information are used for the criteria that were mentioned, the commonest responses were use o f routinely collected data, existing guidelines, and use o f their medical knowledge. However, not much direct evidence was seen o f use o f routine data except in 3 districts which had attempted to make summaries o f key infrastructure and service indicators. In one district, the hospital had made a list o f key hospital performance indicators which was o f an impressive quality. However, in the same district the district health services lacked such summaries. The hospital has a Medical Superintendent with an MPH and a university trained Hospital Administrator. None o f the study districts had a university trained administrator in contrast to the hospitals which all now have a social scientist graduate trained in health services management. This situation arose because the M O H made it a requirement for hospitals and provided resources to recruit university graduates and train them in health services management. Some respondents mentioned that the local administrative leaders and politicians are their sources o f information on how to allocate the funds! The possibility o f differences in use o f data between centrally funded, district funded, and NGO/project funded activities were explored but there was no evidence o f any differences. There was no convincing evidence o f use o f data, neither in budgetary allocations nor in deciding on payment releases. In short, even those who said data is used did not provide convincing evidence to support it nor was such evidence seen in more than 2 o f the 10 study districts. Those who said data is not used were further questioned as to the reasons. Their un-prompted responses were noted using a check-list and included; - lack o f skills, poor data reliability, and lack o f time.

99


6.4.3 Financial Allocations Generally districts have adopted the priority areas advocated by the central government, namely; primary education, primary health care, roads and agricultural modernization. In practical terms there is very little money from district local revenue available for health. Most funding for health is from central government by way o f the block grants to districts, PHC conditional grant, and various special M O H programmes, as well as donors/NGOs. The block grants are meant to be allocated by the district for all services for which the district is prim arily responsible e.g., health, education, roads, etc. However, in order to ensure some priority areas are covered, the central government introduced conditional grants which are only to be spent within the sector (e.g., health or education) for which they are intended. However, detailed allocations o f these additional funds between interventions also remained a responsibility o f the districts.

6.4.3.1 What is Money Allocated for? In Paper 1, the current decision-making processes and influences on resource allocation were presented. In this paper results are presented from the analysis o f available information on allocations that was obtained from the health plans, budgets, and reports, and other existing health records. The analysis o f health plans and budgets covered the 1997/98 and 1998/99 financial years. The objective was to obtain a picture o f allocations that were actually made as a way of understanding the influences and relative importance given to different allocation alternatives. Even though the centre provides broad guidelines on the use o f funds, detailed allocation between line items remains at the discretion o f the district. Expenditure is permitted as long as it is according to their approved work-plan. Analysis o f the district health plans and budgets showed great differences among districts in terms o f what they allocated their money for.

100


General Allocations T able 6.4.1 shows the summary o f allocations for two Financial years (1997/98 and 1998/99) which were the most recent for which documentation was found in most districts. Constant 1998 shillings based on the Consumer Price Index are used (1 USS = 1,000/=). Table 6.4.1 is intended to show district differences on those items over which they have a reasonable degree o f control. It excludes allocations for personnel remunerations (salary/wages, incentive payments and lunch allowance), which are presented in Table 6.4.2. Personnel remunerations are largely outside discretionary budget control o f the district health managers. Similarly. Table 6.4.1 Figures do not include allocations for drug procurement because districts all depend on centrally determined allocation o f essential drug kits as their main source o f supply. Although the kits are supposedly allocated taking into account district populations and levels o f health facilities, there were wide differences in population to drug kits as seen in Appendix H. The allocation o f funds for supplementing drug supplies is controlled by the district but depends on perceived or real adequacy o f the drug kits. Table 6.4.3 provides the most comprehensive picture o f district finances because it includes all allocations. A detailed list o f district indicators including other financial allocations is in Appendix H.

101


Table 6.4.1: Summary Statistics • Per Capita Allocations for Non-salary District PHC Recurrent Budget for 1997/98 and 1998/99 Financial Years in 10 Ugandan Districts by Broad Budget Category* (Based on Allocations for 1997/98 and 1998/99 Financial Years; Constant 1998 Uganda Shillings are used; 1 US$ = 1.000/=)

Mean

General

General PHC

Vertical

Logistics

Training

Totals

Administration

Services

Programs

Shillings

Shillings

Shillings

Shillings (US$) 269 (.27)

Shillings (US$)

Shillings (US$) 1547(1.56)

(US$) 315 (.32)

(US$) 748 (.75)

(US$)

703 (.70)

3711 (3.71)

Median

216 (.22)

1274 (1.27)

552 (.50)

206 (.21)

507 (.51)

3051 (3.05)

224 (.22)

Std.

587 (.59)

1219(1.22)

254 (.25)

756 (.76)

2701 (2.70)

deviation N

16

15

16

16

16

15

Missing

4

5

4

4

4

5

0.833

0.835

0.788

0.806

1.011

0.728

CV

Notes: 1.*This table excludes allocations for personnel remuneration and drug procurement. A comprehensive allocation is presented in Table 6.4.3. Additional financial indicators are in Appendices H and I. 2. CV - Coefficient o f Variation

102


Personnel Remunerations Personnel remuneration is o f two main forms; regular salaries and wages, and lunch allowance which was introduced first for staff working in government hospitals and later extended to PHC staff as well. However, in 1997/98, PHC staff were only paid lunch allowance out o f the balance left over after the available allocation was distributed to hospital staff31. Table 6.4.2 shows the breakdown o f personnel payments for PHC and hospital based staff. A ll Figures are in shillings per capita and are presented in constant 1998 currency. As can be seen from the table, salary was not reflected in the budgets o f 5 out o f the 10 districts studied. These districts did not include personnel remunerations in their comprehensive budgets as they were expected to. When asked why, some stated that it was because they have no control over salary, which is usually fixed and determined centrally. Four o f these districts were in the same region.

Table 6.4.2: Summary Statistics • Per Capita Allocation for Personnel Remuneration in 1997/98 District Health Budgets (Constant 1998 Uganda Shillings are used; 1 US$ = 1,000/=; US$ equivalents are in brackets) CV Std Deviation N Missing Median Mean Category 1997 Population

515472

482375

348673

10

0

611 (.61)

514 ($.51)

5 10

5

0.841

0

0.383

5

5

0.779 0.482

PHC Lunch

47 (.05)

561 (.56) 48 (.05)

PHC Total

664 (.66)

605 (.61)

18 (.02) 517 (.52)

Hospital Salaries/Wages

822 (.82)

681 (.68)

396 (.40)

8

2

PHC Salaries/Wages

0.676

Hospital Lunch

305 (.31)

288 (.29)

178 (.18)

10

0

0.584

Hospital Total

1179(1.18)

1123(1.12)

381 (.38)

8

2

0.323

Total Salaries/Wages

1468(1.47)

1099(1.10)

726 (.73)

5

5

0.495

33

The government has moved to narrow the gap between lunch allowance for hospital staff and that for P H C staff in subsequent financial years. The preferential treatment o f hospitals was largely introduced to avert a spate o f strikes by hospital workers. It generated severe complaints and similar threats from PHC staff.

103


Category Total Lunch

Std Deviation

352 (.35)

Median 344 (.34)

Missing

CV

188 (.19)

N 10

0

0.534

1886 (1.89)

1808(1.81)

695 (.70)

5

5

0.369

PHC Salary as 7c of Total Salary

38.5

27.0

21.1

5

5

0.547

PHC Lunch as % of Total Lunch

15.2

15.3

6.1

10

0

0.398

32.4 25.7 PHC Remuneration as 7c of Total Table 6.4.2 continued from previous page

17.1

5

5

0.527

Total Remuneration

Mean

Notes: 1. PHC - Primary Health Care is used here to refer to all non-hospital allocations 2. CV - Coefficient o f Variation

Table 6.4.3: Summary Statistics - Total Per Capita Allocations for PHC and Hospital Services in 1997/98 District Health Budgets (Constant 1998 Uganda Shillings are used; 1 USS = 1,000/=; USS equivalents are in brackets) Category 1997 Population

Mean

Median

Std Deviation

N

Missing

CV

515472

482375

348673

10

0

0.676

664 (.66)

605 (.61) 3344 (3.34)

517 (.52) 1897(1.90)

5 8

5 2

0.779

PHC Total Recurrent

3719(3.72) 4841 (4.84)

5964 (5.96)

2576 (2.58)

5

PHC Capital

2452 (2.45)

2658 (2.66)

1830(1.83)

8

5 2

0.532 0.746

TOTAL PHC BUDGET

7658 (7.66)

9627 (9.63)

4619(4.62)

5

5

0.603

Hospital Personnel Hospital Other Recurrent

1179(1.18) 1723(1.72)

1123(1.12)

381 (.38)

8

2

0.323

8

1 2

0.690 0.483

Hospital Capital

3046 (3.05) 406 (.41)

1188(1.19) 1470(1.47)

9

Hospital Total Recurrent

1577(1.58) 2857 (2.86) 178 (.18)

575 (.58)

9

1

1.416

TOTAL HOSPITAL

3498 (3.50)

3418 (3.418)

1507(1.51)

8

2

0.431

PHC Personnel PHC Other Recurrent

BUDGET

104


0.510

Mean 8553 (8.55)

Median 8971 (8.97)

Std Deviation

N

Missing

CV

3558 (3.56)

5

5

0.416

TOTAL DISTRICT CAPITAL 3426 (3.43)

3956 (3.96)

2003 (2.00)

5

5

0.585

13974(13.97)

5321 (5.32)

5

5

0.444

54.0

50.7

12.8

5

5

0.237

Recurrent PHC Capital as % of Total

80.7

95.3

27.1

5

5

0.336

Capital Total PHC Budget as 7c of

59.1

61.4

14.7

5

5

0.249

Category TOTAL DISTRICT RECURRENT BUDGET

BUDGET DISTRICT TOTAL HEALTH

11979

BUDGET PHC Recurrent as 7c of Total

(11.98)

Total District Health Budget Table 6.4.3 continued from previous page Note: CV- Coefficient o f Variation

Summary statistics from a detailed analysis o f allocations for selected district PHC budget indicators is presented in Tables 6.4.4 to 6.4.6. District specific details o f these indicators are in Appendix I.

105


Table 6.4.4: Summary Statistics * Per Capita Allocation for Selected Budget Items for 1997/98 and 1998/99 (Constant 1998 Uganda Shillings are used; I USS = 1,000/=; USS equivalents are in brackets) Budget Item

Year

Mean

Median

Std

shillings

shillings

Deviation

(USS)

(USS)

shillings

N Missing

CV

(USS) Supplementary drugs

97/98 98/99

140 (.14) 244 (.24)

102 (.10) 163 (.16)

Supervision to Health Units

97/98 98/99

107 (.11) 99 (.10)

76 (.08) 48 (.05)

128 (.13)

8

2

309 (.31)

10 7

0

108 (.11) 123 (.12)

97/98 98/99

28 (.03) 39 (.04)

39 (.04) 17 (.02)

18 (.02) 66 (.07)

97/98 98/99

1 (00) 3 (.00)

0(.00) 0(.00)

97/98 98/99

11 (.01) 29 (.03)

0(.00) 0(.00)

2 (.00) 6 (.00) 15 (.01) 71 (.07)

Transport and vehicle maintenance

97/98

139 (.14) 144 (.14)

86 (.09) 131 (.13)

IEC/mobilization

97/98 98/99

246 (.25)

148 (.15)

92 (.09) 203 (.20)

310 (.31)

97/98 98/99

280 (.28) 191 (.19)

213 (.21) 221 (.22) 138 (.14)

250 (.25)

MCH/FP

261 (.26) 218 (.22)

HIV/AIDS/safe sex/STDs

97/98 98/99

533 (.53) 298 (.30)

650 (.65) 202 (.20)

Malaria

97/98 98/99

111 (.11) 124 (.12)

Meetings/monitoring

97/98 98/99

55 (.06)

Supervision to community Outreach allowances

9 7 8 7 8 7

0.914 1.266

3 1

1.009

3 2

0.643 1.692

3 2

2.000

1.242

2.000 1.364

9

3 1

7 9

3 1

0.986 0.639

7 8

3 2

0.825 0.806

7 9

3 1

0.932 1.141

404 (.40) 375 (.38)

7 9

3 1

0.758 1.258

93 (.09) 116 (.12)

91 (.09) 120 (.12)

7 8

3 2

50 (.05) 107 (.11)

7

91 (.09)

54 (.05) 39 (.04)

8

3 2

0.820 0.968 0.909

97/98 98/99

14 (.01) 38 (.04)

8 (.00) 35 (.04)

16 (.02) 30 (.03)

7 9

3 1

0.789

Stationery

97/98 98/99

22 (.02)

13 (.01)

7

1.318 1.385

97/98 98/99

14 (.01) 155 (.16)

3 1

HMIS/Filing system

26 (.03) 122 (.12)

29 (.03) 36 (.04)

116 (.12)

76 (.08)

88 (.09) 133 (.13)

8

3 2

0.721 1.147

Training/seminars/workshops

97/98 457 (.46) 98/99 | 952 (.95)

484 (.48) 639 (.64)

290 (.29) 924 (.92)

7 9

3 1

0.635 0.971

Transport and night allowance

98/99

Office operations

137 (.14)

9 7


2.448

1.176 1.143

Table 6.4.5: Summary Statistics for Allocations for Routine Immunizations per Child < 1 year for 1997/98 and 1998/99 Financial Years (Constant 1998 Uganda Shillings are used; 1 USS = 1,000/=; USS equivalents are in brackets) Mean shillings (US$)

Median shillings (USS)

Std Deviation shillings (USS)

N

Missing

CV

1997/98

3156 (3.16)

1904(1.90)

3521 (3.52)

7

3

1.116

1998/99

2152 (2.15)

1142(1.14)

2265 (2.27)

9

1

1.053

Financial Year

Table 6.4.6: Summary Statistics for Allocations for National Immunization Days/Polio Campaign per Child < 5 years for 1997/98 and 1998/99 Financial Years (Constant 1998 Uganda Shillings are used; 1 USS = 1,000/=; USS equivalents are in brackets) Mean

Median

Std Deviation

shillings (USS)

shillings (USS)

shillings (USS)

1997/98

720 (.72)

635 (.64)

1998/99

1888 (1.89)

667 (.67)

592 (.59) 2612 (2.61)

District

N

Missing

CV

8 9

2

0.822

1

1.383

6.4.3.2 Influences on Financial Allocations Given the small sample o f districts, missing data, and the fact that donor/NGO inputs are often unpredictable from district to district, the figures reported here show much variation. However, these limitations notwithstanding, there were some identifiable patterns in regard to some budget items. These relationships were examined using both Pearson’ s and Spearman’ s correlations. The correlations that were significant between the budget allocation variable o f interest and selected district indicators are presented in Tables 6.4.7 to 6.4.12. Variables that were examined but

107


found not to be significant are mentioned below the respective table as well. The p-values in the tables are testing the hypothesis that the correlation coefficients are not significantly different from zero. Since it was not feasible to get detailed expenditure information on financial indicators o f interest, the budgetary allocations were analyzed in detail as a proxy for where districts are (at least in intention) putting their money. The following results show the characteristics o f these allocations using indicators o f PHC funding. Excluding salaries, an average allocation for PHC from the total budget was found to be 70.9% (standard deviation o f 13%). The share for PHC drops to 59.1% (standard deviation 14.7) if personnel remuneration is included, a reflection o f the concentration o f personnel in the hospital.

Table 6.4.7: Proportion of Total Non-salary Health Budget Allocated for PHC and Selected District Indicators 1. Summary Statistics Variable allocation for non-salary PHC budget as

Mean .7093

Std. Deviation .1300

N 10

proportion of total budget allocation for non-salary recurrent PHC budget

.6653

.1369

10

is proportion of total non-salary recurrent budget

2. Correlation Coefficients and Significance Levels Variable

allocation for non-salary recurrent PHC budget

Pearson’s r 2-tailed Spearman’s rho 2-tailed

0.862

p-value 0.001

0.866

p-value 0.001

is proportion of total non-salary recurrent pudget

108


Non-significant correlations were with; income and education, total budget allocations, PHC recurrent budget, hospital capital and recurrent budget, total funds released, PHC facilities. Health Centre and hospital beds, drug kits supplied, PHC conditional grant (both budgeted and released), district and donor contributions, measles coverage, distance and travel time from the national capital, area and population. Having money allocated for PHC p e rs e is not enough since as can be seen from District 41, most o f it could be going to capital projects such as “ rehabilitation” o f PHC units and not service delivery. To examine allocations for possible service delivery activities, a useful indicator that was used is the proportion o f total district recurrent health budget (both PHC and hospital) that was allocated to PHC. Salary was excluded from this analysis since district managers have almost no say about it ’ s allocation. An average o f 66.5% (standard deviation o f 13.7%) was obtained. The details and other statistics are as in Table 6.4.8.

Table 6.4.8: Proportion of Non-salary Health Recurrent Budget Allocated for PHC and Selected District Indicators 1. Summary Statistics Variable

Mean

Std. Deviation

N

illocation for non-salary recurrent PHC budget as

.6653

.1369

10

)roportion of total non-salary recurrent budget tl location for non-salary PHC budget as proportion

.7093

.1300

10

)f total budget )ODulation to 1 health centre bed

7511

2521

10

109



Pearson’s r

2-tailed Spearman’s 2-tailed

illocation for non-salary PHC budget as

0.862

p-value 0.001

rho 0.866

p-value 0.001

iroportion of total budget jopulation to 1 health centre bed

0.539

0.108

0.638

0.047

Non-significant correlations included; income, education, total allocations, per capita amounts for PHC recurrent and capital budgets, hospital recurrent and capital budget, total funds released. PHC facilities, hospital beds, clinical officers and enrolled nurses, drug kits, PHC conditional grant (both budgeted and released), donor and district contributions, infant measles immunization coverage, travel burden (in terms o f time and distance from national capital), area and population. Another indicator used as a measure o f financial allocations to PHC was the per capita discretionary allocation for recurrent PHC budget.

Table 6.4.9: District per Capita Allocations for PHC Non-salary Recurrent Budget and Selected District Indicators 1. Summary Statistics (Constant 1998 Uganda Shillings are used; 1 USS = 1,000/=; USS equivalents are in brackets) Variable

jer capita allocation for PHC non-salary ecurrent budget otal per capita allocation for non-salary budget )er capita allocation for PHC capital budget >er capita funds actually released from all >ources ier capita donor contributions

Mean

Std. Deviation

N

shillings (USS) 3719(3.72)

shillings (USS) 1673(1.67)

10

7785 (7.79) 2452 (2.45) 3255 (3.26)

4284 (4.28) 1614(1.61) 1531 (1.53)

10 10 10

814 (.81)

718 (.72)

10

110



Pearson’s r

2-tailed Spearman's 2-tailed

otal per capita allocation for non-salary budget >er capita allocation for PHC capital budget )er capita funds actually released from all

0.824 0.836 0.635

p-value 0.003 0.003 0.049

rho 0.711 0.805 0.498

p-value 0.021 0.005 0.143

iources )er capita donor contributions

0.825

0.003

0.535

0.111

Non-significant correlations were with; income and education, proportion o f total budget allocated for PHC, hospital capital budget, hospital recurrent budget, PHC facilities. Health Centre and hospital beds, clinical officers and enrolled nurses, drug kits, PHC conditional grant (budgeted and actual), measles coverage, travel burden, proportion o f total recurrent budget allocated for PHC. area and population. Most o f the health funding is going to districts in form o f money for vertical programmes. We were interested to see to what extent districts are allocating non-vertical money to deliver or support an integrated PHC service delivery. Allocations for such activities as purchase o f supplementary drugs, IEC and health education, supervision, etc were combined into one indicator for integrated service delivery.

Ill


Table 6.4.10: Non-salary Recurrent Budget Allocated for Integrated PHC Service Delivery and Selected District Indicators 1. Summary Statistics (Constant 1998 Uganda Shillings are used; 1 USS = 1.000/=; USS equivalents are in brackets) Variable 3er capita allocation for direct integrated PHC Jelivery otal per capita allocation for non-salary budget 3er capita allocation for PHC recurrent budget )er capita allocation for PHC capital budget )er capita funds actually released from all sources jer capita donor contributions

Mean 580 (.58)

Std. Deviation 356 (.36)

N 10

7785 (7.79) 3719(3.72) 2452 (2.45) 3255 (3.26) 814 (.81)

4284 (4.29) 1673 (1.67) 1614(1.61) 1531 (1.53) 718 (.72)

10 10 10 10 10


Pearson’s r

2-tailed Spearman’s rho 2-tailed

otal per capita allocation for non-salary

0.745

p-value 0.013

0.675

p-value 0.032

budget 3er capita allocation for PHC recurrent

0.965

0

0.951

0

judget 3er capita allocation for PHC capital budget )er capita funds actually released from all

0.776 0.648

0.008 0.043

0.707 0.644

0.022 0.044

sources )er capita donor contributions

0.856

0.002

0.559

0.093

Non-significant correlations were with; income and education, proportion o f total budget allocated for PHC, proportion o f total recurrent budget allocated for PHC, PHC facilities. Health Centre beds, clinical officers and enrolled nurses, PHC conditional grant (budgeted and actual), district contributions, measles coverage, travel burden, area and population. Allocations for H IV /A ID S were combined with those for other reproductive health activities since in practice there is a lot o f overlap in funding and during implementation. An average o f shs.536/= per capita is allocated for these activities. 112


Table 6.4.11: - Per Capita Allocation for HIV/Reproductive Health and Selected District Indicators

1. Summary Statistics

ier capita allocation for HIV/AIDS/reproductive*

Mean 536 (.54)

Std. Deviation 305 (.31)

N 10

tealth illocation for non-salary PHC budget as proportion

.7093

.1300

10

>f total budget illocation for recurrent PHC budget as proportion

.6653

.1369

10

i f total recurrent budget jer capita district contribution* 87 (.09) 86 (.09) 10 *(Constant 1998 Uganda Shillings are used; 1 USS = 1.000/=; JSS equivalents are in brae


Pearson’s r

2-tailed

Spearman’s rho

2-tailed

illocation for non-salary PHC budget as

0.845

p-value 0.002

0.808

p-value 0.005

iroportion of total budget illocation for recurrent PHC budget as

0.72

0.019

0.802

0.005

-0.526

0.119

-0.738

0.015

iroportion of total recurrent budget )er capita district contribution

Non-significant correlations were found to be with; income, education, total allocation, allocations for recurrent PHC budget, allocations for capital PHC budget, allocations for hospital capital and recurrent budgets, total funds released, PHC facilities, PHC and hospital beds, clinical officers and enrolled nurses, PHC conditional grant both budgeted and released, donor contributions, measles coverage, travel burden, population and area. Total funds released to districts for health from all known sources for the 1997/98 Financial

113


Year excluding salary payments were analyzed. Salary payments were excluded from this particular analysis because the main concern o f the study was with those items over which districts have some decision-making control. The average per capita disbursement (excluding salaries) for the 10 study districts was 3,255/= (standard deviation -1,531/=). There were significant Pearson and Spearman's correlations between per capita funds released with some variables as shown in Table 6.4.12.

Table 6.4.12: Per Capita Funds Released in 1997/98 from all Sources and Selected District Indicators 1. Summary Statistics Variable

Mean

N

Std. Deviation

total district population 1997 515472 cer capita funds released from all sources* 3255 (3.26) otal per capita allocation for non-salary budget* 7785 (7.79) cer capita allocation for PHC recurrent budget* 3719(3.72) copulation to 1 PHC facility 12021 copulation to 1 drug kit supplied 1278 *(Constant 1998 Uganda Shillings are used; 1 USS = 1,000/=;

10 348673 10 1531 (1.53) 4284 (4.28) 10 1673 (1.67) 10 3241 10 670 10 USS equivalents are in brackets)


Pearson’s r

total district population 1997 otal per capita allocation for non-salary budget cer capita allocation for PHC recurrent budget copulation to 1 PHC facility copulation to 1 drug kit supplied

-0.639 0.729 0.635 -0.613 -0.657

2-tailed Spearman’s 2-tailed p-value 0.047 0.017 0.049 0.060 0.039

rho -0.661 0.576 0.498 -0.669 -0.661

p-value 0.038 0.082 0.143 0.035 0.038

Correlations found not to be significant were with; life expectancy index, income index, education index, years o f decentralization, distance from national capital, travel time from national capital, total district area, income*education interaction, travel distance*travel time interaction (called travel burden).

114


The idea underlying these tables is to show which other district indicators these variables o f interest are significantly associated with. Some o f these are important influences driving the allocations. In order to get a national picture o f the sharing o f available financial resources between the centre and districts, disbursements for 1997/98 were analyzed by broad budget centre category. The results are shown in Table 6.4.13.

Table 6.4.13: National Disbursements of Public Financial Resources by Broad Budget Centre Categories for 1997/98 Financial Year (In Constant 1998 Uganda Shillings; 1 US $ = 1,000/=) [Data source: Mugarura C; M O H Uganda, 1999) Budget Centre

Constant 1998

% Share of

Ug shs

Total

National/Central MOH (Central HQ, Devpt,

22,047,446,665

35.2

Mulago National Hosp

14.098,685,187

22.5

Butabika National Hosp

1,097.641,149

1.8

Sub-total

37,243,773,001

59.5

Donors District Hosp

10,824,447,985 6,055.904,832

17.3 9.7

Regional Hosp District Contribution

3,681,483,910 1,442.700,745

5.9

Training Schools

1,435,444,789

2.3 2.3

PHC Cond Grant

1,149.233,890

1.8

NGO Hosp (Govt

769,856,891

1.2

25,359,073,042

40.5

62,602,846,042

100.0

Research Inst)

District

Contribution) Sub-total Total


6.4.4 Health Management Information System The current HMIS system that has recently been started in most districts o f Uganda has been cited by many as a big improvement over the old system (Postma, 1998). However, in practice health workers were said to be unsure o f how to actually use it. Our own observations and interviews in the districts visited over the study period did not show convincing evidence o f HM IS having brought about a change in resource allocations or management o f the district health system as yet. Available information suggested that often the new HMIS booklets were accessible to a few, usually limited to those in-charge o f the unit, who often kept it locked in a cupboard. HM IS summary records observed at the district were generally incomplete for most o f the indicators apart from a few entries on catchment population, immunization data, and perhaps the amount o f money collected from userfees. Even these were often not complete. Information on service utilization such as OPD and Antenatal attendance, and institutional deliveries, and EPI outputs were generally available and fairly complete. A t district level, only 2 o f the 10 visited had fairly comprehensive or near complete records for the last three years covering OPD attendance. ANC attendance, immunizations, and information on finances, logistics, and personnel readily available. Although 9 o f the 10 had a designated clerk to handle compilation o f HMIS information at the District Health Office, only the 2 mentioned above had well established systems for filin g and keeping HMIS records. In some cases even to find complete HMIS records for 2 years before 1997 took some time as papers would be found in different filin g systems or rooms. Sometimes records for one year would be found in different files. The general observation is that information is generated but is not used, and systems for storage and easy retrieval are very poor indeed.

116


6.4.4.1 Criteria Suggested for Resource Allocations Respondents were asked to mention those criteria that in their view should be used to make the allocations to different health interventions. O f the 45 respondents who answered to this question, only 2 (4.4%) said they did not know o f any criteria that should be used. The unprompted responses are shown in Table 6.4.14 (multiple responses were allowed).

Table 6.4.14: Criteria Respondents Would (Jsc in Allocating Resources to Health Programmes (n=43) Criteria

Frequency Mentioned

(%)

High Number Affected

27

62.8

High Demand by Public

10

23.3

Causes Quick Death

24

55.8

Affects Children

18

41.9

Affects Women

16

37.2

Affects Men

2

4.7

Effective Treatment or Control

22

51.2

Money is Available

13

31.0

Money is Earmarked

13

31.0

It is National Policy

19

45.2

Measure Available

When they were asked which criteria should be used for resource allocation in their opinion, respondents answers showed a concern for addressing problems that affect many people, cause death quickly, and affect children. These responses indicate areas in which they might be more interested in developing and collecting information. Other comments offered included (a) bureaucratic e.g., “ allocate according to workplan” (this was the most frequently mentioned, (b) political, “ allocate to each section according to a percentage” , “ allocate equitably according to activity” , “ give mandate 117


to implementers” , “ depends on the political programme o f the district council” , "depends on if health department is taken as priority over other departments” (this refers to the issue o f the amount o f the health budget available to be partitioned between different health programmes), and (c) technical, “ allocate for PHC activities", "allocate according to nature o f activity or output expected” , “ should be based on a needs assessment", "use reports, findings o f the supervision team". As can be seen most o f these responses suggest a lack o f concern or emphasis on objective information on health status as the basis for resource allocation.

6.4A.2 Requests for Information District health managers were asked if they felt they needed further information to decide on allocations, 33 out o f 50(66.0%) said “ yes", while 10 (20.0%) answered "no" with 7 (14.0%) not knowing what to say. On what kind o f information is needed, the respondents would like more information on the follow ing categories:

Technical; - factors affecting acceptance o f interventions e.g., use o f bednets; - current district population and demographics; - why use o f latrines is falling34; - disease specific morbidity; - infant and maternal mortality; - updated health facility data (e.g.. population served, services delivered);

Even though most districts have had an active campaign to promote water and sanitation for most of the last 10 years. 118


- which data is available from previous studies15; - indicators to use for assessing the impact o f their activities; - personnel requirements to implement activities; - prevalence o f diseases; - baseline data (on their district for those aspects relevant to health); - community willingness to pay for health (mentioned by a financial person); - deaths and births in the communities; - environmental health issues; - cost-effectiveness o f interventions as well as burden o f disease (mentioned in only one district); - factors underlying poor use o f latrines;

Bureaucratic; - government guidelines on special health problems; - government guidelines on resource allocation; - availability o f funds and how they should be allocated; - administrative guidelines e.g.. on lunch allowances (which was a big source o f dissatisfaction among PHC workers at the time o f the visits); - which funds are available. We see that respondents feel the need for more technical information and also bureaucratic guidelines to help them in decision-making for their work. Such information is largely either missing, unclear, or needs specialized skills to produce.

35

Sometimes studies that have been done in similar districts are repeated even though the additional information expected is not likely to be different. 119


6.4.4.3 Estimating Disease Burden

Vital Statistics In all districts, there is no information system that collects data about general demographics like births and deaths and population health status apart from the health system. The national system o f registering births and deaths that used to be based at local administrative units collapsed during the years o f political turmoil and has not yet been revitalised. In the absence o f a vital registration system, a way o f gaining insight into the causes o f death is verbal autopsy (Marsh. Husein. Lobo et ai 1995). The potential for use o f verbal autopsy in determining cause o f death profiles has not been much investigated in Uganda. Population projections based on 1991 census are available in all districts, usually broken down by parish, which is the smallest viable community local government administrative unit. The National Statistics office o f the MFEP has produced district specific reports for all the basic demographic indicators. Another census due in the year 2001. is being planned. UNICEF did a situation analysis for women ar.d children in Uganda and this gives figures for IM R by region (Barton and Wamai, 1994; Uganda, 1996; UNICEF - Uganda, 1998). However, since people often move to areas outside their own administrative area to seek health care, health managers need information on their functional catchment areas, which often overlap. Only 2 o f the study districts were found to have population catchment figures for their health units but these were often based on the administrative areas e.g.. a population o f an entire sub county would be mentioned even though a sub-county may have more than one health facility.

6.4.4.4 Effectiveness of Current Interventions How well do current strategies and interventions work under field conditions? An example

120


o f this dilemma is illustrated by the case o f money allocated for numerous training courses/workshops. Most districts allocated between 7.0% to 23.5% (median 15.3%) on training. D H M T members generally felt these training courses were useful in improving performance. However, information is needed on how much performance has objectively been improved as a result o f such training. There is no well developed system to monitor improvements in performance. Input data such as personnel deployment disaggregated by cadre and location were generally non-existent except in 2 districts. The lack o f information on personnel is further evidenced by the repeated calls for payroll lists by the M OH so that they may be re-centralized to ensure smooth payment o f salaries. This was very evident during the Annual General Meeting o f the M O H at the end o f 1998. Often, staff lists were either unavailable or not up to date. O nly 3 districts had complete written inventory o f transport facilities and their locations. In one district the store keeper told the researchers he had no idea how many cars, motorcycles or bicycles the health department has because he is never informed when new ones arrive! (Yet he is the one supposed to keep track o f logistics!). Basic data on outputs o f the health system like immunizations. OPD attendance. ANC attendance, institutional births, was generally available in most districts even though sometimes placed in files that took time to locate (indicating they are not much used). It was observed that districts, through the District Planning Offices, are now in the process o f compiling general baseline data on a wide range o f issues from health to water, agriculture, etc. However, most o f these indicators, while they offer a general view o f the district, were not enough to give the district level health manager the kind o f detailed information that is needed to make informed resource allocation choices. In all districts, IM R and M M R figures found being cited or mentioned in district plans were the national figures yet it is well known for instance that there are wide differences among districts. IM R rates in the North are generally said to be higher than those

121

Reproduced with permission o f the copyright owner. Further reproduction prohibited without permission.

in the South (Statistics Department [Uganda] and Macro International Inc., 1996). Information on various interventions is either already available from the literature or could be compiled by consulting local and international experts. Efforts can be made to develop a set of key information on things like effectiveness o f immunization, malaria prevention and treatment strategies, TB counseling, H IV screening, etc. (Ghana Health Team Project Team, 1981; Morrow. 1984; Hyder, Rotllant and Morrow, 1998).

6.4.4.S Cost of Interventions

Unavailable/unreliable accounting records It was very tedious to try to obtain a picture o f how much money was allocated between different budget categories. This is because during analysis o f the plans and budgets it was noted that duplication and mis-classification was common. For instance, supervision would appear in one general supervision section, then again under STDs, and again under health education, etc. Some items would be discovered in a section under which they had no relation to other items in the category e.g., procurement o f equipment would appear under an item called Equipping Health Units then again under malaria control activities one would find an item “ procurement o f microscopes” . During analysis in this study such mis-classified items were identified and included under the appropriate categories to avoid double counting. We were able to obtain itemized breakdown o f expenses o f all district public health services o f the last financial year (1997/98) from only 1 district. Itemized expenditure o f hospitals for the last 3 financial years (1996/97,1997/98 and 1998/99) were obtained from only 2 districts. The remaining 8 were unable to provide the information. Our observation is that it is probably the former. In most cases the accounting records are so chaotic that to extract timely and complete breakdown o f actual

122


expenditures was almost impossible. In one district an on-going financial investigation at the time o f our visit had paralyzed district health activities. The confusion in the financial management system was evident during the analysis o f health budgets. Sometimes the spreadsheet entries for items costed in m illions would be entered in the same column with items costed in thousands o f shillings without clearly indicating the correct number o f zeros or decimal places. Such anomalies were often easy to spot and correct during our analysis o f the documents by comparing with known rates and other rates in the budgets. However, it was very time consuming. Such a situation makes the extraction o f information for decision-making difficult. It also brings into question whether the plans that are made are actually followed or not. One common observation in many districts is that they planned to carry out “ research" to collect baseline data in their own district for say malaria KAP. FP, nutrition, etc. even though a lot o f studies done at national level or in their region would offer fairly representative information. A summary o f the main disease and health problems with the current level o f concern and corrective efforts is presented in Table 6.4.15. The level o f concern and current effort to address the problems have been scored based on the impressions formed during the course o f the research on a 3 point scale. The last column o f the table summarizes the action or information needed to make an improvement.

123



Table 6.4.15: Current Problems, Degree of Concern, Efforts Being Made and Action and Information Needed DISEASE/HEALTH PROBLEMS

LEVEL OF CONCERN

EXISTING PROGRAM TO ADDRESS IT AS DETERMINED IN THE STUDY

- Sexually Transmitted Diseases and HIV/AIDS

T- 3 P- 3

T -3 P -3

- Reproductive health related conditions

T- 3 P 3

T -3 P -2

- acute infectious non-immunizable diseases (e.g., ARI, diarrheal diseases, malaria, skin and eye infections, etc)

T-3 P 2

T-2 T- 1

ACTIONS NEEDED TO IMPROVE

Disease problems seen as priorities requiring health programs

124

Further integration needed

Information needed on effectiveness of interventions and costs


DISEASE/HEALTH PROBLEMS

LEVEL OF CONCERN


- chronic infectious diseases (e.g., worms and parasites, trypanosomiasis, onchocerciasis, etc)

T -3 P 1

T-2 P- 1

Information needed on burden caused by these

-infectious immunizable diseases

T -3 T -3

T-2 P- 2

Information needed on programme implementation effectiveness, alternatives and cost implications

- polio eradication

T-2 P- 3

T -3 P- 3

Information needed on programme implementation effectiveness, alternatives and cost implications

- malnutrition and nutritional deficiencies

T -3 P 2

T-2 P- 1

Information needed on district specific burden and intervention alternatives

- heart diseases and cardiovascular problems

T -3 P- 1

T- 1 P- 1

National information on burden and associated factors is needed

125



DISEASE/HEALTH PROBLEMS

LEVEL OF CONCERN


- meningitis and cholera epidemics

T -3 P- 3

T-2 P- 2

Information on locally feasible control alternatives. effectiveness and cost implications

- home injuries

T- 1 T- 1

T- 1 P- 1

Most serious problem is bums, need information on burden, domestic prevention measures

- vehicle injuries

T -3 P- 1

T- 1 P- 1

Information on burden and costs

(T - technical; P - political; 3 - high; 2 - moderate; 1 - low)

126


6.5 Discussion It was not the intention to develop instruments for assessing information needs during this study. Rather, an attempt was made to look at what is currently available in light o f a theoretical framework for better decision-making. We attempted to use the data that was available in the routine setting to see how it could be used by district level managers to improve their decision-making rather than make extraordinary efforts to collect refined data which usually takes long to plan, collect and analyze, not to mention the expense o f surveys. Little evidence o f actual use o f data was found. Attempts at data summarization that were seen in a few places showed often incomplete records which were generally not up to date.

6.5.1 Allocations for PHC The East African Burden o f Disease Study recommended shifting public resources from conditions whose treatment benefits individuals and whose spillover benefits to society are limited, in favour o f community and preventive interventions (The W orld Bank, 1995). As we would expect, higher allocation for PHC, as reflected in the proportion o f the budget for PHC. shows a higher commitment and support for PHC and is likely to result in better organized and supervised services. It was a pleasant surprise to find high proportions o f district recurrent non-salary budgets allocated to PHC contrary to what was expected. This is a positive reflection on the commitment o f the government and donors to re-orient services to PHC. However, it should be noted that most o f the health budget is consumed by personnel remunerations as presented in this paper. Since most personnel are based in hospitals which are located in urban areas, clearly the concentration o f health resources is still largely in favour o f curative and hospital interventions. The potential for freeing up government and donor resources to further support PHC w ill only be realized i f there are

127


significant re-distributions o f personnel. It w ill be d ifficu lt to remove personnel from hospitals since in many cases they are fu lfillin g a crucial role and many hospitals are themselves still under-staffed. A strategy that could be tried is the preferential acceleration o f recruitment for PHC. In many districts staff working in PHC units are predominantly nursing aides who are usually trained on the job without having had any formal health training at all. The government ban on recruitment o f new s ta ff6 was applied indiscriminately to all government departments. The PHC sector suffered disproportionately worse consequences. It has recently been lifted in order to enable recruitment for the new Health Sub-District (HSD) policy. Districts which have a higher proportion o f total budget allocated for PHC also have higher proportions o f recurrent budgets allocated for PHC.

6.5.2 Use of Data In the end what emerges is that there are no significant correlations with any o f the variables one would expect if the health managers were using health outcomes, health status or even health service utilization data to decide on allocations. The same goes for the centre to district allocations. They are significantly correlated with the number o f health facilities and district population but nothing else. This confirms the information from the centre, that they use population and existing infrastructure to allocate to districts. However, it casts doubt on the central view that they also use other factors such as how far the district is from Kampala, whether it has donors, etc. none o f which shows up in the correlations. The practice o f using infrastructure simply perpetuates the inequity between those districts which already have more and those that do not. We would like to argue that the policy makers should seriously consider giving additional central support to districts (on top of

36

This was imposed as part of measures to curb government expenditure under lending conditionalities by the World Bank/IMF financial institutions. 128


the one based on population and facilities) taking into account current health status. Some o f the indicators reported in this study could be used in assessing the extent o f additional support. It should be noted that examining for linear correlation does not rule out other non-linear relationships. However, there was no evidence that there was any such significant association. Examining o f residuals and log transformations o f the data did not show any significant difference from the un-transformed data. It is possible that bootstrapping or a larger sample size might have discovered other relationships not found in this study but this is unlikely. A t the moment, basic data is so disorganized that it is d ifficu lt to make meaningful managerial decision-making in a timely fashion. One major priority is to establish a culture o f basic data and information use by building on or improving the new HMIS system that was established two years ago. The question is should managers have a very comprehensive HMIS with an exhaustive list o f indicators which hardly anyone uses in practice or should they aim at something that actually works for an average district? I f a small set o f key indicators are agreed on and districts get to actually reliably collect information on those and use them for decision-making, there might be a better basis for building future systems. We have attempted to highlight what could be some o f these indicators for which data are already being collected or can be found.

6.5.3 Allocations for Service Delivery The more districts have, the more they are likely to allocate to service. Districts with higher PHC conditional grant budgeted and released amounts also had higher per capita allocations for PHC recurrent budget. Higher PHC recurrent budgets were associated with higher allocations for direct service delivery. Although dropped from the final model due to multi-collinearity, it is interesting to note that correlation between proportion o f recurrent budget allocated to PHC with number o f PHC facilities

129


in general (which includes different sized units) was not significant but that with Health Centre beds was significant. A possible explanation lies in the fact that most unit based PHC activities usually start at the bigger units (Health Centre IV ) which are also the ones that have beds. Smaller facilities tend to get involved later or not at all. A district with more big health centres therefore is likely to allocate more for recurrent activities. This is an indication o f the likely effect the new policy on Health Sub-Districts (HSD) w ill have. The HSD policy envisages the creation o f 214 health sub systems, with a big health centre as it ’ s pivot. Each HSD w ill be expected to offer a comprehensive range o f basic public health and curative services, including caesarian sections. Not surprisingly, the more money that was allocated for PHC recurrent budget in general, the more that was also allocated for integrated direct PHC service delivery. This indicates that the w ill for allocating to service delivery depends on money that is available after other fixed commitments have been met. This is the district’ s “ disposable income” after central projects, and other commitments have been taken care of. Allocations for HIV/reproductive health are generally higher in districts with higher proportion o f total budget allocated for PHC. Sim ilarly, districts with higher district budgetary contributions (most o f which goes to pay un-skilled staff) are associated with a decreased allocation for HIV/reproductive health. This is not surprising since districts that spend more on salary for health staff are those who have fewer centrally paid staff, and rely most on staff trained on the job. Such staff are not likely to have skills that are usually needed to implement H IV control and reproductive health activities. Districts with higher allocations for H IV /A ID S and reproductive health also tend to have higher proportions o f total budgets allocated to PHC. This is largely a reflection o f the significant amount o f resources going for HTV/AIDS and reproductive health activities. These have the effect o f boosting the PHC component o f the budget but it is probably detrimental to other PHC activities such as immunization. In fact, infant immunization coverage or

130


lack-there-of did not appear to be significantly correlated with any o f the variables checked. If districts were sensitive to preventive programmes, or there was a link between their allocations and delivery o f immunization activities, one would have expected some association. There is an opportunity cost o f having staff involved more in one vertical activity or the other as something else is bound to suffer.

6.5.4 Infrastructure Effect The total funds released to districts from all sources are most reliably predicted by the PHC facilities and hospital beds. Although there were also positive correlations with population, this was over-shadowed by the physical facilities. It points to the fact that most centre-to-district allocations are still very much dependent on the number o f existing facilities. The problem with this approach is that districts with more facilities get more central funds. This has the effect o f central allocations actually perpetrating inequity instead o f contributing to narrowing the gap between the better-off and the more disadvantaged districts. This is exactly the opposite o f what was intended in providing for equalization grants in the national constitution o f 1995. It is therefore imperative that a formula for central disbursements to districts take into account real health needs, e.g., based on life expectancy and other health status measures than continue basing on existing infrastructure. The shift to evidence based allocation could happen gradually starting with population to facility and need for coverage with services indicators. This could then move to use o f health status (mortality and morbidity) information. Districts that spend most o f their PHC budget on doing capital projects such as rehabilitating physical infrastructure o f health units are likely to have less available to support PHC service delivery. Whereas most hospital capital projects tend to be specially funded from project or central funds, capital projects at PHC units are often done at the expense o f funds for supporting PHC

131


service delivery. This problem is most vividly illustrated by district 41 which had much less allocated per capita for service delivery activities compared to other districts. However, to its credit, this district also had completely integrated it's service delivery and had zero allocation for vertically implemented activities. It is also important to make a distinction between district hospitals (first referral level) and regional hospitals (specialist referral) since their resource consumption patterns are different. There is a prestige factor in having one’ s district hospital being upgraded to a higher (e.g., regional) status, with specialists, and receiving referrals from neighboring districts. However, the implications o f the investment need to be taken into account in a planned manner to avoid a runaway recurrent cost problem (Mahapatra and Berman, 1995).

6.5.5 Equity There is wide variation in district allocations for similar items as can be seen from the coefficient o f variation figures for each budget item shown in the respective tables. O f course, the wide variation is probably due to the small sample size. However, it is also likely due to the fact that different districts have different projects and sources o f funding that include international bilateral funding agreements, and various NGOs. There are also wide variations in per capita allocations for staff remuneration. The differences are due to the different numbers o f staff and health facilities and not preferential treatment since all staff o f a particular cadre and experience get the same salary and lunch allowance regardless o f location. One might want to argue for an additional allowance for staff serving in remote and disadvantaged areas in order to promote equity. The wide differences in lunch allowance between PHC and hospital staff is an issue o f concern. However, the government is moving to address and in a short while the differences w ill no longer exist. It should be noted that PHC staff have opportunities to participate in various programme activities and generally benefit

132


more than hospital staff from other duty and training allowances. Since funding is overall still inadequate to meet the needs, and there is no nationally agreed information as to what services should reasonably cost, there is no convergence o f allocations to some stable figure. One problem that comes out is the inequity in coverage by hospital beds that exists among districts. It would be desirable to clearly state the national targets in regard to infrastructure in order to promote equitable health infrastructure development. Since hospitals are a big drain on the tax payer, the temptation by different districts to expand hospitals or upgrade their hospital should be resisted. Genuine need could be assessed using established criteria and a certificate o f need granted. Such is only possible if there is a clear national policy that is backed by a strong political w ill. A clear distinction should be made between planned expansion to meet growing population needs or fill an existing gap from politically influenced expansion. In the case o f the latter, there could be a firm national policy not to use central government or centrally channeled donor funds.

133


6.6 Summary and Conclusions During this work we have analysed information that was collected from districts and provided estimates, o f some o f the potentially useful basic district health management indicators appropriate for Uganda. These could be used in monitoring and assessing performance in relation to allocative efficiency and equity.

It is evident that currently, little use is made o f information

that is already there or could be obtained from use o f available sources. Completeness o f information is a major issue in most o f the districts studied. We have tried to document some o f what is available and where the gaps are in order to prepare the ground for a needs-based approach to resource allocation decision-making. However, information flow and decision-making in organizations takes place in a wider organizational context (Simon, 1967). In Paper 3, we shall examine some o f the contextual and organizational factors that were observed in Ugandan districts which have a bearing on the subject o f improving resource allocation decision-making. It w ill be argued that decision-makers at both district and national levels should keep the wider picture in mind as they introduce new policies otherwise they w ill not have a good chance o f achieving the intended results.

134


CHAPTER 7: TOWARDS BUILDING CAPACITY FOR EFFECTIVE DECISION-MAKING IN DISTRICT LEVEL RESOURCE ALLOCATION (Paper 3)

7.1 Introduction This paper is concerned with addressing the question o f building up capacity at district level for effective decision-making in resource allocation. One o f the main challenges o f running a decentralized health system in developing countries is how to make informed choices on alternative uses o f the scarce health care resources. This problem cropped up in Uganda soon after decentralization started in 1993. Since then, many changes have taken place, with districts having primary responsibility for delivery o f health services while the central M O H retains responsibility for policy formulation, setting and monitoring standards, training, and emergencies. This meant that district level managers now have to decide themselves how to allocate the health share o f the available resources in between competing health problems. In Paper 1, we examined the current process o f decision-making in regard to resource allocation in publicly Financed health services in the context o f decentralization in Ugandan districts. We pointed out that decision-making is being influenced significantly by past practices, central and donor allocations, real or perceived national guidelines and donor influence, and local district factors such as political expediency and expected personal gain by health technocrats. There is little use o f a needs-based, evidence driven approach in allocating health resources. In Paper 2. started to look at how the situation could be improved by exploring the available information and identified information gaps that could be filled in order to guide resource allocation decision-making. We saw that even though information may be scarce, what is often available is not exploited largely because the district staff lack confidence and skills or are overwhelmed by the 135


many things they have to do to cope with decentralization. This paper presents the results o f an assessment that was done o f the functional capacity in the 10 study districts to make allocation decisions based on use o f evidence and information. It identifies key aspects o f capacity and health system functional status that include issues o f use o f service utilization and health status data, personnel constraints, district management structures, logistics, and the problems o f bureaucracy. It identifies existing inequities between districts and argues for a stronger central support to weaker districts through provision o f additional grants based on level o f need. A range o f district performance and resource allocation indicators is offered from which managers at both district and national levels could choose for monitoring progress. However, the ideal is to have direct measures o f health status so that resource allocation can be according to real need. Such measures that have been used elsewhere include the D A L Y and H EALY. This paper is attempting to provide information on the gap existing between potential use o f these measures at district level and the current management constraints. We shall highlight those areas considered crucial in order to increase the chances to make better resource allocation decisions. It is assumed that we are able to use a theoretical framework to predict what should be useful towards achieving the goal o f better health outcomes and eventually, improved health status. It is assumed that decision-making as it is now can be improved within the means currently available. We shall examine some o f the contextual and organizational factors that were observed in Ugandan districts which have a bearing on the subject o f improving resource allocation decision making. It w ill be argued that decision-makers at both district and national levels should keep the wider picture in mind as they introduce new policies otherwise they w ill not have a good chance o f achieving the intended results. What institutional, policy, structural and personnel capacities should be strengthened or put in place for that to happen is the focus o f this paper.

136


7.2 Background Following decentralization o f all social service provision to the district level in 1993. districts were faced with new challenges o f planning, and overall management o f health services. Functions that used to be done by the central M OH were now devolved in a phased manner to district level over the next 3 years. Health managers at both district and national level expressed the need for information as to what was working and what needed to be improved. Management o f resources is one o f the main challenges districts found themselves facing. This study was conceived in order to make a contribution to promote better management o f the scarce health resources. The study aimed to achieve the following purposes by the end o f December 1998:

7.2.1 Overall Goal of Study: To review the process o f decision-making for allocation o f public health resources in decentralized districts in Uganda in order to provide a basis for development o f allocation procedures for district health systems that address health needs o f the people and promotes rational use o f scarce resources.

7.2.2 Study Objectives: 1.


2.

Identify the formal guidelines/criteria that are currently existing in the districts with regard to allocation o f public funds to health activities, if any, and the informal influences that play a role.

3.


137


4.

Assess the extent to which decisions being made in allocation o f public health funds relate to the health problems perceived as priorities (e.g., as defined by the decision-makers’ perceptions and based on national guidelines such as the national basic health services package) as well as reasons for any observed deviations.

5.

Based on the findings o f the study, outline a resource allocation decision-making model of the process that is observed in Ugandan districts, and suggest ways to improve it.

Paper 1 presented findings and issues related to Objectives 1. 2. 3 and 5. The second paper presented findings relevant to Objectives 3, 4 and 5. Findings and issues related to Objective 5 as well as broader contextual issues in Ugandan districts are addressed in this third and last article.

7.3 Subjects and Methods An assessment o f the functioning o f the district health system was conducted by building on the approach suggested by Kielman et al (1991). The main tools for information gathering were a structured questionnaire administered to members o f the D H M T by an interviewer as well as secondary data collection and observations. A total o f 44 D H M T members were interviewed. Frequency o f meetings, communication, knowledge o f what funds were available, etc were used as signs o f presence or absence o f good teamwork. The presence or absence o f information displayed on boards for the staff to see was noted. Key in-put and output information such as staffing, finances, and service indicators were obtained from records and reports at district health offices, as well as from the MOH. Key indicators were calculated with the aid o f a spreadsheet (QuattroPro Version 8 ).

Selected district indicators were used to explore expected relationships among variables

138


using correlation and multiple linear regression analysis. Information on life expectancy, educational status and income37 were obtained from the UNDP Human Development Report for Uganda 1998 (UNDP, 1998). Response variables for the multiple linear regression were Life Expectancy Index as a measure o f health status. Out-patient Department (OPD) attendance in new cases per person in the district and Ante-natal Clinic (ANC) attendances in new cases per expected pregnant woman in the district38, as measures o f health department service outputs. The indicator chosen for coverage with preventive measures was infant measles immunization coverage. Under 5 mortality rate figures specific to each district would have been preferable as a measure o f health status but were not available. Health services utilization data is regularly collected from all units. It was assumed that the reliability and/or completeness o f the information could generally be assumed to be uniform across the country. Where data for a particular district was not obtained the mean o f the data from the other districts was used during the analysis. It was considered necessary to do these adjustments rather than exclude these districts altogether since the sample size o f 10 districts is small and resources were not available to do a more extensive study e.g.. by replacing districts or increasing the number. Since this was a case study, the main interest was to obtain new information and document existing practices. Possible explanatory variables and interaction terms were selectively used in the pair-wise correlations using both Pearson’ s and Spearman’ s correlations. Variables with significant correlation coefficients on either Pearson or Spearman’ s methods were later used in building the multiple linear regression models. A probability for inclusion was set at p - .10. Analysis was done using SPSS for Windows Version 8.0 Professional. In all cases the final models were examined for the potential problem o f multicollinearity by looking at the Tolerance19and Variance Inflation Factors (V IF )40 which were largely found to be favourable. To correct for multi-collinearity in building the models, a variable that was no longer a significant cause in change o f variance at the set entry probability upon introduction o f a new variable was eliminated from the model. To reduce the bias in the data that comes from extreme outliers, a better picture is obtained by examining the case-wise diagnostics data on the 10 districts presented in Appendix K. The diagnostics presents the actual and predicted values, as well as standardized residuals.

39

Tolerance values range from 0 to 1. A value closer to 1 means the variable has less o f it's variability explained by the other independent variables (Norussis. 1994). 40

V IF is a measure o f the increase in variance due to the multiple correlation among independent variables. It is desirable to keep it as close to 1 as possible though values less than 10 arc not a problem.

140


7.4 Results

7.4.1 District Profiles A collection o f up to 50 indicators o f district performance and resource allocation was obtained using data obtained in the study from existing sources such as health plans, budgets, and reports (attached as Appendix H). The main influences on district decision-making can be summarized as internal district organizational factors41 as well as influences from the central government and role o f donor/NGO projects. Regional and geographical factors were important in defining the financial and health situation o f districts. There were 3 characteristics under which districts could be described (a) “ high funding-high verticalization" (b) “ better-developed-health districts” or “ health-development” group and (c) “ poorly-developed-district-health" or “ health-backwardness” group. The first group o f districts is characterized by high donor contributions, high allocations for logistics, high recurrent PHC budget, high total per capita health budget, etc as well as a number o f vertically funded programmes, hence it ’ s label as a “ high funding-high verticalization" group. A typical district in this category is Masaka District. The second group o f districts is characterized by more favourable population to staff ratios and drug supplies, as well as high allocations for PHC. Examples o f such districts with a high “ health-development" factor are districts 23 and 32 which are outliers and generally stand out quite favourably on health status from the others in their region. The third characteristic o f “ health-backwardness” was seen in districts that are typically low income, low education status, long distance from the national capital, long travel time, have high allocations for general administration (to cover the long distances, and large area, etc). They have favourable population to health facility and bed ratios but their populations are more dispersed than those in

4lThese included local political considerations, and perceived interests of decision-makers. 141


districts with better health status. Apparently the available health facilities had little or no association with measures used for health status. Examples are Arua and Moroto districts.

7.4.2 Life Expectancy Is there any connection between the allocations being made to districts for health with their current levels o f health services coverage, utilization and health status? This is important because it affects the district's discretionary budget for allocation to different health programmes. Since there are fixed commitments like paying salaries and keeping infrastructure, the amounts available for district managers to allocate towards improving health are determined by how much they receive in the first place. The reasoning was. if the centre is sensitive to this, then payments to districts should be able to reflect this, with more going to those districts with lower health status. The mean life expectancy index for the study districts based on UNDP published data was .43 284: (standard deviation o f .07463). A multiple linear regression analysis produced a model that explains about 42% o f the observed variance (R: = .415) and would explain about 34% o f variance in life expectancy of districts drawn from a similar population o f districts e.g., other districts in Uganda (Adjusted R: = .342). There was a negative correlation between life expectancy index and hospital recurrent budgetary allocations on bivariate analysis but it was not significant in multi-variate analysis. Details o f the results are in Table 7.4.1 and case-wise diagnostics are in Appendix K. The p-values are testing the hypothesis that the correlation coefficients are not significantly different from 0.

42

This is the same as 43 .3% and is equivalent to about 35.7 years based on full life expectancy at birth o f 82.5 years from the Coale and Demeany Model Life Table West. Level 26 for females! (Coale and Guo. 1989).

142


Table 7.4.1: Life Expectancy and Selected District Indicators 1. Summary Statistics

life expectancy index education index per capita allocation for hospital

Std. Deviation 7.463E-02 .1429 1120

Mean .4328 .5026 1723

N 10 10 10

lon-salarv recurrent budget^ ♦Constant 1998 Uganda Shillings are used; 1 US$ = 1,000 shillings


education index per capita allocation for hospital

Pearson’s r

2-tailed

Spearman’s

2-tailed

0.644 -0.637

p-value 0.044 0.047

rho 0.479 -0.515

p-value 0.162 0.128

icn-salarv recurrent budget ♦Dependent variable: Life Expectancy

3. A N O V A Sum of Squares df 2.082E-02 1 Regression Residual 2.931E-02 8 9 5.013E-02 Total a Predictors: (Constant), Education Index Model 1

Mean Square 2.082E-02 3.663E-03

F 5.684

Sig. .044

b Dependent Variable: Life Expectancy Index

In a multivariate analysis only education was found to remain significantly associated with life expectancy. The relationship is summarized in this equation;

p re d ic te d lein dex = .264 + .bAAeducitidex + e ^ ; .........................................................Equation 7.4.1.

143


where, p re d ic te d lein dex is predicted life expectancy index; and educindex is education index and e u . is an error term specific to these measures.

Other correlations examined and found not to be significant are; income index, total budgetary allocations, total funds released, PHC recurrent and capital budgets, hospital capital budget, proportion o f total health budget allocated for PHC, proportion o f recurrent health budget allocated for PHC. PHC conditional grant (both budgeted and released), donor and district contributions, and the income*education interaction.

7.4.3 Health Services Utilization Is there a connection between budget allocations and health services utilization? Possible relationships were explored. The indicators used to explore possible links between resource allocation and utilization were OPD attendance and Ante-natal C linic attendance as well as institutional un-complicated deliveries (normal births). These could be taken as proxy indicators for morbidity in the population as well as quality and acceptability o f services. However, these may not be appropriate if population access to the reporting facilities is a serious problem. I f the allocations are data driven, we might expect to see some relationship. Average attendances were .4544 new OPD cases per person. .6118 new ANC cases per expected pregnant woman, and .1630 institutional normal delivery per expected birth.

7.4.3.1 OPD Attendance Average OPD new case attendance was .4544 cases per person (standard deviation o f .08552). The influences o f district factors on OPD attendance were investigated using multiple linear regression. The regression model that was obtained explains 54% o f the observed variance (R2 = .541)

144


and would explain about 48% o f the variance in OPD new cases in districts drawn from a similar population o f districts (Adjusted R: = .484). The only significant associations were with income and education. Income alone had a significant correlation in bivariate analysis but got displaced in m ulti variate analysis by the interaction between income and education. Further details are shown in Table 7.4.2 and case-wise diagnostics are in Appendix K.

Table 7.4.2: Out-patient Department New Cases and Selected District Indicators 1. Summary Statistics

OPD attendance in new cases jer person income index income*education

Mean .4544

Std. Deviation 8.552E-02

N 10

.1744 9.272E-02

6.275E-02 4.562E-02

10 10

2. Correlation Coefficients and Significance Levels Pearson’s r

Variable

2-tailed Spearman’s rho 2-tailed

p-value p-value -0.712 -0.744 0.014 0.021 income index -0.756 0.011 income*education -0.736 0.015 *Dependent variable: Out-patient Department New Cases per Person 3. A N O V A Model

Sum of

df

Mean

Squares Square 1 3.561E-02 Regression 3.561E-02 8 3.776E-03 Residual 3.021E-02 6.582E-02 9 Total a Predictors: (Constant), Income ndex*Education Index 1

F

Sig.

9.432

.015

b Dependent Variable: OPD Attendance in New Cases per Person

145


Other correlations examined but found non-significant were with; total budgetary allocations, PHC recurrent and capital budgets, hospital recurrent and capital budget, proportion o f total budget allocated for PHC, proportion o f total recurrent budget allocated for PHC, total funds released, enrolled nurses and clinical officers in PHC, drug kits, PHC conditional grant (both budgeted and released), donor and district funds, number o f PHC and hospital facilities.

A general model to predict OPD new cases is:

p re d ic te d opdnew = .583 - .736in co m e*ed u cation + eopdncw; ......................................Equation 7.4.2.

where;

opdnew is p re d ic te d new O P D cases p e r h ead o f population', incom e ^education is the interaction

between income and education and e0[xlncw is an error term.

7.4.3.2 Ante-Natal Clinic Attendance New Ante-Natal Clinic (ANC) attendance on average was 61.2% o f pregnant women (standard deviation o f 16.9%). A M ultiple Linear Regression model was obtained that explains 67% o f the observed variance (R2 = .666) and would explain 63% o f variance in a district randomly chosen from those similar to the study districts (Adjusted R2 = .625). It shows that the main predictor o f ANC utilization is the total amount o f funds actually disbursed to districts for health. Table 7.4.3 shows the details o f the analysis. Case-wise diagnostics are in Appendix K.

146


Table 7.4.3: Ante-natal New Cases and Selected District Indicators 1. Summary Statistics Mean .6118 .1744 7785

Std. Deviation .1686 6.275E-02 4284

N 10 10 10

budget* per capita funds released from all sources* 3255 1531 1278 670 population to 1 drug kit supplied in a year 718 814 per capita donor contributions* ♦Constant 1998 Uganda Shillings are used; 1 US$ = 1,000 shillings

10 10 10

ANC new cases per expected pregnancy income index total per capita allocation for non-salary

2. Correlation Coefficients and Significance Levels Pearson’s r

2-tailed

Spearman’ s rho

2-tailed

income index total per capita allocation

-0.515 0.669

p-value 0.128 0.034

-0.657 0.688

p-value 0.039 0.028

for non-salary budget per capita funds released

0.816

0.004

0.819

0.004

i'rom all sources population to 1 drug kit

-0.776

0.008

-0.757

0.011

supplied in a year per capita donor

0.796

0.006

0.788

0.007

Variable

:ontributions ♦Dependent Variable: ANC Attendance in New Cases per Expected Pregnant Woman

3. A N O V A Model

Sum of

df

Mean Square

F

Squares .171 15.977 1 Regression .171 8 1.067E-02 8.538E-02 Residual .256 9 Total a Predictors: (Constant), Per Capita Funds Released from A ll Sources 1

Sig.

.004

b Dependent Variable: ANC Attendance in New Cases per Expected Pregnant Woman

147


Other correlations examined and found non-significant were w ith; education, PHC recurrent and capital budgets, hospital capital and recurrent budgets, proportion o f total health budget allocated to PHC, proportion o f total recurrent budget allocated to PHC, enrolled nurses and clinical officers in PHC units, PHC conditional grant (both budgeted and actual release), district contributions for health, PHC and hospital facilities. ANC attendance could be predicted using the Model expressed as; p re d ic te d a n c n e w = .319 + .816fn n dsall + 6 ^ ,,^

......................................................... Equation 7.4.3.

where; an cnew is predicted new ANC attendances per expected pregnant woman .Ju n d sall is total per capita

funds known to have been actually paid to district health departments, and e.uic„cw is an error term.

7.4.3.3 Infant Measles Immunization Coverage Are decision-makers concerned about coverage rates? If so. is there any sign that their allocations are being driven by such concerns like infant immunization coverage rates? Average measles immunization coverage for infants was 64%. It was not possible to construct a model to predict the proportion o f infants immunized against measles since there was no significant correlation between measles coverage and any o f the key variables. Variables tested for correlation with measles coverage but found non-significant included; income and education, area, total budgetary allocations, PHC recurrent and capital budgets, hospital recurrent and capital budgets, proportion o f total budget allocated for PHC, proportion o f total recurrent budget allocated for PHC, total funds released, enrolled nurses and clinical officers deployed in PHC, drug kits, PHC conditional grant (both budgeted and released), donor and district contributions, PHC and hospital facilities.

148


The initial and final models are summarized in Table 7.4.4.

Table 7.4.4: Summary of Models Response Variable

Initial Model

Final Model

Life Expectancy Index (leindex)

Education Index (educinde) Per Capita Allocation for Hospital Non-Salary Recurrent Budget ( rechosp)

Education Index (educinde)

OPD Attendance in New Cases per Person ( O P D N E W )

Income Index (incindex) Income Index-Educalion Index Interaction (incindex *educinde)

Income Index-Education Index Interaction

ANC Attendance in New Cases Per Expected Pregnant Woman (.ANCNEW )

Income Index (incindex ) Non-Salary Per Capita Budgetary Allocation (totpcalf) Per Capita Total Funds Released from all Known Sources in 1997/98 (fundsall) Population to 1 Essential Drug Kit (poptokit) Per Capita Donor Contributions

7r Infant Measles Immunization Coverage (measlcov)

None significant on bivariate analysis

1incindex*educinde)

Per Capita Total Funds Released from all Known Sources in 1997/98 (fundsall)

149


7.4.4 Ownership and Teamwork A common observation in many districts was that although on personal contact many o f the D H M T members seemed to know what to do or were often quite informed about the problems and how they could be resolved, there seemed to be a perception o f powerlessness in addressing the problems. As seen before in Paper 2, the managers perceive the need for more technical information in managing health services. In addition to a lack o f skills, organizational and motivational factors further hinder development o f the health system. Available data is often not analyzed to provide even basic information that could improve managerial decision-making. The following sections present some o f those factors considered o f urgent importance in order to improve the capacity for better decision-making.

7.4.5 Capacity for Financial Management Problems with accounting was the single most cited reason for delays in payments o f funds for all types o f activities. Other reasons were late submission o f work-plans and budgets. Each district operated on average 6 different accounts to hold funds for vertical programmes, or donor and other NGO supported activities. This often placed an impossible burden on the accounting staff (usually one clerk at the District Health Office). Often the vote books and ledgers were not kept up to date and clerks had to spend some considerable time compiling expenditure data. In many districts, they were not in a position to do this during the time o f the visit. This problem is referred to in a MOH report (MOH, 1997) which found similar problems. The fact that most District Health Offices do not have trained Administrative staff could be a significant factor. The technical people are too busy trying to run the numerous programs and deal with visitors to pay any meaningful attention to how things are organized. This phenomenon has been cited as a common managerial problem in organizations (March, 1994).

150


In the interviews with D HM T and financial staff, the phrases “ poor accountability” and "excessive bureaucracy” were the most commonly mentioned in relation to why funds delay. Other reasons cited were late release by the centre and lack o f funds due to low revenue base o f the districts41. As mentioned previously, availability o f district officials to sign checks when needed was an issue in most districts. In all districts, it was felt that once a request was initiated by the line officer for an activity which appeared in the approved work-plan, funds would usually be released provided the right steps were followed and funds were available. In a number o f districts it was often found that an item would be planned and budgeted for but never actually implemented in that financial year, necessitating it ’ s reappearing in the plan. In some cases, such items were seen to have been rolled over for about 2-3 financial years! Examples o f such include purchase o f vehicles, renovation o f health facilities, training. District health plans and budgets usually start with a preamble in which an account is given o f the previous year's financial and implementation performance. These documents often pointed out that for a number o f budget items on which funds had not been spent it was because funds were not received in time, or were not received at all and so implementation could not take place. The "Background to the Budget, 1998/99” , a government document produced as a technical companion to the annual government budget, mentions the tendency by many districts to make unrealistic forecasts o f funds expected from revenue, central government and donors (MFEP, 1998). A recent news report quoted one official saying Uganda lacks qualified accountants. He cited a figure o f only 200 professional accountants whereas Kenya has 2,000 and Tanzania has 1,000 (East African Business News, 1999).

43

Districts sometimes rely on central grants to meet their own obligations e.g., paying allowances for district local council members to meet, or settling salary arrears o f locally recruited staff.

151


7.4.6 Bureaucracy Although innovative approaches are being tried in some districts e.g., prepaying for fuel, etc, most districts still use the traditional steps in getting money to fund activities. From an analysis o f the planning and budgeting steps, as previously noted in Paper 1, there were up to 26 key steps identified from budgeting to actual release o f funds. In the process there are many delays and opportunities for something to go wrong in between allocation and having money in the hand for implementation o f health activities. These steps could be changed but would imply the agreement o f the key ministries involved, namely MOLG, M O H , and MFEP. One problem that came out is the use o f outdated financial guidelines and regulations, as well as keeping many bank accounts and manually maintained accounting systems. There was talk o f introducing computerized accounting systems which would greatly improve the situation. A ll districts visited have got computers and most have a system o f power back-up to safe-guard against the numerous power black-outs. A computerized accounting system could be introduced without much problems. There is enough local capacity in the country to do the necessary training and offer maintenance support. The complex steps in getting funds does not apparently prevent diversion o f funds or prevent a feeling o f suspicion. A t the time o f the study team's visit to one district, the District Local Council had ordered the suspension o f financial releases to the health department pending investigations into alleged financial impropriety.

7.4.7 Office Management D ifficu lty to trace records, incompleteness o f records, disorganized state o f records were encountered in all but 2 o f the 10 study districts. The personnel supervising general administrative and office activities are generally people who entered the service as ju n io r clerks and worked their way up into supervisory positions. Information available from observations and Key Informants

152


suggests that such people often rose due to their “ connections" to powerful people in or out o f the district rather than as a result o f their managerial competence. In addition, traditional civil service provisions make it almost impossible to ever get rid o f a non-performing administrator. At best, such a person is usually simply transferred to another department and another equally incompetent person from another department is brought in! The same problem used to be experienced in the running of hospitals. The situation in hospitals was only saved by the intervention o f the M O H which made it a pre-condition that in order for one to be appointed as a Hospital Administrator one had to have a minimum o f a university bachelor's degree and to undertake a course in health services management. There is a huge potential pool o f recruits into administrative positions because there are many university graduates who are jobless. Some districts have taken the lead by recruiting graduate level administrators to manage day to day non-professional activities o f the health department. This frees the DDHS to concentrate on the professional aspects o f public health practice. When one talks to a DDHS in Uganda the story is often told o f how they spend between 40%-50% o f their time doing non-professional routine things like attending meetings, handling disputes, reviewing vouchers and payrolls, etc. They then spend another 30%-40% o f their time in training in or outside the district. Either way, between 60% to 70% o f their time they are not in office. One would wish that they are spending the time out in the field supervising health services but observations suggest otherwise!

7.4.8 Information Management Displays o f HMIS summaries and/or wall charts showing basic output information e.g., immunization data (commonly) were found in only 3 districts. One respondent summarized the problem this way, "we do not have a culture o f using data” . How true indeed! Systems for storage, indexing and easy retrieval o f records had obviously been present some decades ago as evidenced by old filin g cabinets and labels in most districts. However, the

153


systems went into a state o f disorganization during many years o f neglect that arose out o f the political problems o f the country. There was an obvious need o f file cabinets in most district health offices visited. Files were often found piled high on top o f desks or shelves. Secure, lock-able cabinets are generally a luxury and where they exist, they are in the offices o f either the DDHS or some vertical programme office or project co-ordinator. Under such conditions records quickly deteriorate or get lost altogether. Not surprisingly, it was hard to find records more than 3 years old in 7 o f the 10 districts.

7.4.9 Policy Interpretation It was common to find health staff thinking that their jo b was to implement what had already been decided by the centre (M OH or M O LG ) or donors and NGOs. A close examination o f existing guidelines and documents from the centre did not, however, support this view. Clearly there is a big gap between what the documents actually say and what the people at the district think the guidelines are.

7.4.10 Planning and Budgeting A ll the members o f the D H M T were reported to participate in making the Annual District Health Plans and budgets in all districts. The supervisor o f the clerical staff in the District Health O ffice who is called the Executive O fficer Accounts (or Higher Executive O fficer in the past) generally participates as well. Plans and budgets where items were sometimes listed with other unrelated items were found in all districts although varying in degree. In all districts there was no attempt to separate costs between capital and recurrent. The only separation was between hospital and PHC services. The

154


researcher spent many days looking at each item in all the plans and budgets to come up with dis aggregated allocations as presented in Paper 2.

7.4.11 Corruption in Health Units Respondents were asked whether in their view, informal payments are being made in the health units in their districts. O f the 44 who responded to the question, 31 (70.5%) thought that informal payments are made, while 8 (18.2%) thought there are no informal payments in their units and 5(11.4% ) did not know. They were asked to state their subjective assessment o f the likelihood that a patient attending the hospital or health centre w ill be asked to pay a bribe. In most cases respondents felt people are more likely to be asked to make informal payments (bribes) in hospitals than in PHC units. This might be a reflection o f the bias inherent in the fact that respondents were all generally urban dwellers and more likely to hear o f such illegal payments at the hospitals than in PHC units.

7.4.12 Management Structures How functional are the management structures that were described in Paper 1 (also see Appendices B, C, D and E)? This question was important because those groups are recognized as the legitimate bodies charged with management o f health services. It is in them that most o f the official decision-making takes place. There are 3 main health management groups in all the districts. These are the core District Health Team (D H T) made up o f the section heads o f the technical departments o f the District Health Office. Then there is the so-called Extended D HT which actually is also referred to as the District Health Management Team (D H M T). This has the core DHT as well as the Hospital Superintendents and Co-ordinators o f NGOs running health programs. These two bodies are the main technical groups which plan and budget for health services. Whereas the DHT

155


is supposed to do the day-to-day running o f the health system, composed o f 7-8 people (maximum about 10 in some districts) and therefore meet more often (as often as weekly in some districts), the D H M T is usually larger (20-30 people in some districts!) and ideally meets every quarter. Then there is the District Health Committee (DHC) which is composed o f about 7 elected members from the District Local Council (DLC), with the DHT participating as ex officio with no voting powers. The DHC is a sub-committee o f the District Social Services Committee, an organ o f the District Local Council that oversees health, education and social welfare services. The DHC is charged with the power to review and approve plans and budgets for health activities, and present them (through the Secretary for Social Services) to the full DLC for final approval. Respondents were asked how often the meetings are usually held for DHT and D HM T (regardless o f whether they themselves attended or not). Generally core DHT meetings were reported to be regularly held ranging from once a week to quarterly, whereas D H M T meetings were said to be planned for quarterly but in practice it generally happened bi-annually. Some respondents offered these additional responses for the DHT; “ we meet as needed", “ we meet for consultations anytime so there is no need for a meeting” , “ we meet when there is an urgent need". The additional responses for D H M T were; “ we meet as needed", “ we meet twice a year” , “ the D H M T is not functional", “ the D H M T meetings are irregular". Physical attendance in meetings was rated as good or excellent by 41 out o f 43 (95.4%) for D HT and 32 out o f 41 (78.0%) for D H M T respectively. The extent o f participation in the discussions was rated as good to excellent by 38 out o f 44 (86.4%) for D HT and 35 out o f 42 (83.3%) for D H M T respectively. Since the respondents were D H M T members and usually sit as ex o fficio members on the DHC, they were asked about regularity o f DHC meetings, attendance o f members and the extent to which members participated in discussions. The regularity o f meetings in the last financial year was

156


rated as good to excellent by 24 out o f 42 (57.1%), while 6 people (14.3%) rated them as bad to very bad, 4 (9.5%) said it was neither good nor bad. and 8 (19%) did not know about DHC meetings. On physical attendance to the DHC meetings, 26 out o f 42 (61.9%) thought it was good to excellent. 4 (9.5%) thought it was bad to very bad, 3 (7.1%) thought it was neither good nor bad, and 9 (21.4%) did not know. Asked how they would rate the extent o f participation in the discussions by DHC members, 26 out o f 43 (60.5%) thought it was good to excellent, 2 (4.7%) thought it was bad, 5 (11.6%) said it was neither good nor bad, while 10 (23.3%) did not know. Those who said meetings are irregular were asked to mention the reasons for having meetings irregularly. The answers were given in an open format. The commonest mentioned reasons were....“ DDHS is very busy", "members o f the DHT are busy” .

7.4.13 Staff Welfare How do staff feel about their official remuneration benefits? This has a bearing on their levels o f commitment to their jobs. Salary was used as the basic indicator. Respondents were asked to state their perceptions regarding the adequacy o f salary and wage payments in terms o f meeting their food and other basic needs in one month. The following scale was used:

1. very inadequate, not enough to survive on, can't buy food for a month; 2. inadequate, barely provides food for one month, cannot provide for other needs; 3. "half/half"*4, somehow adequate for feeding and solving a few fam ily problems; 4. adequate, generally enough to feed the fam ily and solve some problems; 5. very adequate, definitely enough to feed the family well, pay school fees, and solve most

44

A Ugandan expression meaning that the situation is neither good nor bad.

157


fam ily problems. They were asked to score separately for hospital. District Health O ffice and PHC units. Their responses are summarized in Table 7.4.5.

Table 7.4.5: Health Managers’ Perception of Adequacy of Salaries and Wages for Staff at Three Levels of the District Health System SALARY ADEQUACY

DISTRICT HEALTH OFFICE

PHC UNITS

HOSPITAL

Frequency (%)

Frequency (%)

Frequency (%)

5: very adequate

2(4.7)

1 (2.3)

1 (2.3)

4: adequate

3 (7.0)

1 (2.3)

5(11.4)

3: "half/half’

6(14.0)

4 (9.3)

11 (25.0)

2: inadequate

26 (60.5)

22 (51.2)

22 (50.0)

1: very inadequate

6(14.0)

15(34.9)

4(9.1)

9: do not know

0 (0.0)

0 (0.0)

1 (2.3)

43(100.0)

43 (100.0)

44(100.0)

Total

They were again sim ilarly asked to rate the regularity o f staff payments at the different levels o f the health system. The follow ing scale was used:

1. very irregular, almost never get paid on time; 2. irregular, generally not paid on time except only occasionally; 3. "h a lf/ha lf', sometimes on time, sometimes not; 4. regular, generally paid on time, though occasionally late; 5. very regular, almost always paid on time. Again they were asked to score separately for hospital. District Health Office, and PHC units.

158


The results are summarized in Table 7.4.6.

Table 7.4.6: Health Managers’ Perception of Regularity of Salaries and Wages for Staff at Three Levels of the District Health System

DISTRICT HEALTH

PHC UNITS

HOSPITAL

OFFICE

Frequency (%)

Frequency (%)

3 (7.0)

2(4.7)

4(9.1)

4: regular

29 (67.4)

6(14.0)

28 (63.6)

3: “ half/half’

5(11.6)

7(16.3)

5 (11.4)

2: irregular

5(11.6)

20 (46.5)

5(11.4)

1: very irregular

1 (2.3)

8(18.6)

1 (2.3)

9: do not know

0 (0.0)

0 (0.0)

1 (2.3)

43 (100.0)

43 (100.0)

44(100.0)

REGULARITY OF SALARY

Frequency (%) 5: very regular

Total

When asked who usually decides on staff deployments, most thought decisions on staff deployments are usually made by the DDHS (27 out o f 42 or 64.3%) or the Chief Administrative O fficer (6 out o f 8), while 14 out o f 42 (33.3%) mentioned the M O H /M O LG as the ones usually deciding. (There is a lot o f confusion regarding personnel issues since decentralization started in general, with district managers often not knowing what powers they have over hiring and/or firing o f staff or not knowing how to exercise them).

159


7.4.14 Logistics Most o f the work o f the district health office is aimed at providing support to communities to improve their health. Transport therefore becomes crucial for members o f the health team to reach the people. They were asked if transport is usually available to them when needed. It was evenly split with 22 o f 44 (50.0%) saying “ yes" and the other 22 (50%) saying “ no” . It was generally reported that facilities are available for doing routine maintenance for vehicles in 8 out o f the 10 districts. In the other 2, vehicles have to be taken to Kampala, the national capital city for servicing. Servicing for motorcycles and bicycles was reported to be available in all districts. Asked to rate the functional status o f medical equipment at the district hospital and PHC units from 1 (very bad) to 5 (very good to excellent). 25 o f 44 (56.8%) respondents said it was good to excellent at the hospital. 14 (31.8%) said it was “ h a lf-h a lf’. 3 (6.8%) said it was bad, while 2 (4.5%) did not know. For the PHC units, 20 out o f 43 (46.5%) said it was good to excellent, 16 (37.2%) said it was equally good and bad (half-half), and 7 (16.3%) said it was bad to very bad. Most respondents said there were no facilities in the districts visited to do routine maintenance o f medical equipment. Surprisingly, even some respondents in districts which have the regional medical equipment workshops thought there were no facilities. In some cases they were aware o f the existence o f the facilities but mentioned that the team was over-stretched as it has to cover a number o f districts. A vailability o f drugs and other medical supplies was rated as being good to excellent by 32 out o f 44 (72.7%) respondents, as bad by only 2 (4.5%) while 9 (20.5%) said it was “ sometimes good, sometimes bad” and 1 person (2.3%) did not know. As for PHC units, the rating was 24/43 (55.8%) saying it was good, 3/43 (7.0%) saying it was bad, while 16/43 (37.2%) said it was in between (sometimes good, sometimes bad). Since the main source o f drugs in all units is the Essential Drugs Programme drug kits, the

160


respondents were asked to assess to what extent the kits alone met the drug requirements o f the hospital and PHC units. For the hospital, 21/43 (48.8%) rated adequacy o f kits in meeting requirements from good to excellent (generally better for the small district hospitals), while 5/43 (11.6%) thought it was bad and 17/43 (39.5%) said it is in between. As for PHC units, 17/43 (40.5%) thought kits are good for meeting requirements, 21/42 (50.0%) thought they are in between, and only 4 (9.5%) thought they were inadequate in meeting drug requirements. On average there were found to be a population to I drug kit supplied o f 1,278 persons (standard deviation o f 670).

7.4.15 Supervision of PHC units Supportive supervision o f staff in rural health centres is crucial for improving quality since they are often less skilled, and are the first point o f contact between the communities and the health system. On regularity o f supervision to PHC units, 28/41 (68.3%) reported that they supervise units monthly, 10/41 (24.4%) reported going quarterly. The others either said supervision is irregular or did not know. A ll the members o f the D H M T were reportedly involved in doing supervision to PHC units. O f these the DDHS, Assistant DDHS, DHV, DHI, DHE, TBLS. D A ID , DCCA. MOH officials, DNO were said to regularly visit the hospital for supervision as well. Asked to mention constraints to doing regular and proper supervision o f PHC units and hospitals, the unprompted answers that were given are summarized in Table 7.4.7 (multiple responses were allowed).

161


Table 7.4.7: Reasons Mentioned as Constraints to Doing Supervision to PHC Units and Hospitals

Frequency o f Mention by: PHC units

Hospital

- mentioned

16

22

- not mentioned

28

21

- mentioned

8

8

- not mentioned

36

35

- mentioned

19

11

- not mentioned

25

32

- mentioned

32

16

- not mentioned

12

27

8

6

36

36

Mentioned as Constraint to Supervision Lack of time

Lack of team spirit

Lack of money

Transport problems

Lack of supervisory skills • mentioned - not mentioned

Other constraints mentioned in regard to supervision o f PHC units were; “ health workers don't change so one gets frustrated” , “ insecurity” , “ bad roads", “ lack o f motivation", “ lack o f personnel", “ weather” , “ work overload". For supervision o f the hospitals, additional constraints or reasons mentioned were; “ DHT are not serious” , “ the hospital is an NGO and they resist to be supervised by D H T” , “ consultants don’ t accept to be supervised” , "lack o f promotion o f s ta ff’, “ no need since they know what they are

162


doing", “ the hospital is not yet decentralized” , “ the hospital people resist to be supervised by the DHT” . It should be noted that although some o f the districts visited have regional hospitals, the running o f regional hospitals are supposedly delegated to the districts in which they are located. Planning and budgeting for them, as weli as overseeing their operations therefore o fficia lly are delegated functions and come under the day-to-day supervision o f the Hospital Management Team led by the Medical Superintendent but under the overall authority o f the DDHS and the DHM T).

163


7.5 Discussion

7.5.1 Life Expectancy An attempt was made to examine possible links between a number o f health services and general development indicators with UNDP estimates for life expectancy. One would expect that districts with lower health status as indicated by life expectancy at birth should receive extra attention from the central government and donor agencies. The evidence available suggests that worse o ff districts are in that situation for reasons other than their own resource allocation practices. It should be pointed out that although good service delivery may contribute to good outcomes it is not easy to attribute this backwards (Wouters. 1995). This study attempted to situate resource allocation decision-making in the context in which it currently takes place. The review by Wouters (1995) o f quality o f services in Niger reports low levels o f compliance with diagnostic and treatment standards. In Uganda too. although many interventions have been planned and implemented since the 1978 Alma Ata declaration on PHC. observations are that health status is not in keeping with what one would expect for the country's current economic status (The World Bank. 1993). This might, to a large extent, be due to inefficiencies in the allocation and use o f available resources (The World Bank, 1994). In this study the observed lack o f evidence for a data driven or evidence based allocation seems to lend support to the view that the system is functioning in an inefficient way. The available resources could thus be used to improve both allocative and technical efficiency as a step towards getting more societal value for money. This study makes the argument that although in principle decentralization is an opportunity to bring services closer to the people, improvements in health are not likely to be realized i f the current allocation practices continue. In particular, attempts need to be made to bring more objective criteria into the allocation decision-making process at both district and national level. Service

164


outputs and standards need to be stated more explicitly. The relation between inputs and service outputs and outcomes should be developed as a starting point for assessment o f performance o f district health systems. I f we know the deficiencies in the current resource allocation processes we can begin to identify whether the expected levels o f cost-effectiveness can be achieved using the interventions currently in place or if resources are being wasted with little prospect o f improvement. As Wouters (1995) points out. quality improvements in a developing country may imply significant investments in both fixed and variable costs. Before national and district decision-makers can facilitate the improvement in health status, there should be objective basis for making decisions, albeit not entirely technical, such as is provided by use o f composite indicators such as DALYs and HEALYs. These should be based on objective and reliable

information. Furthermore,

implementation should be in a context o f a district health system framework that is, at least in principle, organized along the lines o f the PHC philosophy, embodying quality, cost-effectiveness and equity. In general it is observed that districts that have lower life expectancy also tend to be further away from the national capital, and have lower income index. However, it was only education index and per capita allocation for recurrent hospital budgets that were significantly correlated with life expectancy. Rather surprisingly, income and income-education inter-action did not have significant correlation with life expectancy in a multi-variate analysis. One would expect that increased wealth should translate into better health care, better living conditions, better nutrition, and ultimately a better health status. It is possible that a relationship exists and would have been detected with a larger sample size. The amount allocated for hospital recurrent budget is a reflection o f the number o f hospitals. Some districts have more than one hospital, the larger ones studied had between 3-5 hospitals. Districts farther from the national capital also generally tended to allocate more per capita for

165


general administration, a reflection o f the fact that their overhead costs are higher e.g., transporting personnel, drugs and supplies, as well as linking with the national capital. In general it is worth considering that the national authorities and donor agencies should provide additional support on basis o f such non-health services factors such as education, distance and geographical location. These districts need additional resources to compensate for their extra problems o f relating with the national capital from a longer distance. It is only the central government which can do this. The frame-work is already provided for in the constitution through equalization grants. An analysis o f key indicators (Appendix I) shows differences between districts e.g., Hoima District and Bushenyi share cultural, language, education and income similarities, yet Bushenyi has a worse life-expectancy than Hoima. In-fact Bushenyi has a higher income than Hoima. Factors that seem to be important for Hoima District which is clearly an outlier, include a good coverage with health facilities, a long presence o f district based technical experts from NGOs and bilateral agencies, as well as regular contact with academic institutions. Kapchorwa District enjoys better health mainly because o f geographical location which makes it less susceptible to malaria which is the major cause o f mortality in Uganda. The relatively strong donor Financial support found there is a recent phenomenon just like in Moroto which also has recently received strong donor funding but still has poor health. A nomadic life-style with frequent insecurity is also partly responsible for the poor heath status o f Moroto.

7.5.2 OPD Attendance We could expect that a district that has an allocation practice that promotes better quality health services through availability o f drugs, motivated staff, proper supervision o f services, etc should have should also have better utilization o f it’ s services. To test this idea an indicator of service utilization, new OPD cases per person, was correlated with a number o f financial allocation

166


indicators. None o f them had a significant correlation with OPD attendance. The only significant correlations found were between OPD attendance and income and education. In multi-variate analysis the interaction between income and education was the single most important predictor o f OPD attendance. Districts w ith a higher income*education had less OPD attendance. The likely explanation is that the districts with better income and education also have more private clinics, people can afford these private facilities, and people tend to use these private facilities rather than the government funded ones from which most o f the utilization data was obtained.

7.5.3 ANC Attendance In contrast to OPD attendance, we see that Ante-Natal C linic (A N C ) utilization was not significantly correlated with income or education. In general, most private mid-wives in Uganda offer mainly delivery services. Expectant mothers have more confidence in government units for ANC services since they are sure o f being attended to by trained staff and having some laboratory investigations done. In addition, they know that when delivery gets complicated they w ill usually be referred to government units by these private mid-wives. ANC attendance was significantly correlated with total health budget released to districts, drug kits, and donor contributions. In m ulti variate analysis, the only significant factor that remained in the model is total funding released. This is probably a reflection o f that the general facilitation o f service delivery that comes with better funding (through better mobilization, staff allowances, availability o f supplies, etc) works as a synergistic package. Districts with more funds released are associated with higher A N C new case attendance. The fact that it is the total funds released and not any other variable possibly points to the importance o f the overall functioning o f the district health system. It is the total package o f services and activities that bring about better utilization o f ANC services and probably not single vertical interventions. The

167


positive correlation between AN C attendance and donor funding is not surprising given that most donor funding tends to go for MCH/FP and related services. It is a sign that the investment is at least producing service outputs. It is to be hoped that over time, this should be able to translate into better health status. However, the link between ANC attendance and lowering maternal mortality has been questioned (MacDonagh, 1996).

7.5.4 Infant Immunization Coverage Infant measles coverage is a good indicator o f how well preventive services are reaching and being utilized by the community. It was rather disappointing not to find any significant linear association between observed coverage and financial allocations. I f districts are sensitive to the quality o f service delivery and coverage, one would wish to see some attempt to monitor coverage and allocate resources accordingly. The sample average infant measles coverage o f 64 7c is disappointing but not surprising since a recent nation-wide study had identified a number o f problems with the national immunization system. The focus on National Immunization Days (NIDS) forpolioeradication is channeling significant centrally allocated resources to the exercise but routine funding has been left to the discretion o f the districts. It is possible that districts do not see supporting routine immunization as a priority for them to fund since the centre traditionally did it. In the attempt to eradicate polio it is possible that a lot o f energy o f the health system is being diverted from routine immunization as pointed out in a review article on the subject Taylor and his colleagues (Taylor, Cutts and Taylor, 1997). This is likely to be a result o f taking personnel time in special mobilization, training, as well as the additional strain on logistics such as vehicles and the cold chain. Most districts allocated more per capita for the 2 days o f NIDS than the routine immunization activities for the entire year. In addition mothers may now think that it is not necessary to go for routine immunization since the “ big day” is coming! In the absence o f good mobilization,

168


and staff time to disseminate correct information, mothers are likely to get a wrong message from the entire exercise. The richer districts in general had the same or worse immunization coverage than the poorer ones. This is probably because o f lower utilization o f public facilities (where immunization services are generally provided free) in richer districts, the result is that fewer mothers or attendants are exposed to health messages for immunization. The link between utilization o f curative services and preventive services such as immunization have been noted elsewhere (Yazbeck and Leighton, 1995; Bhuiya, Bhuiya and Chowdhury, 1995).

7.5.5 Getting Value-for-Money The link between resource allocation and utilization is hypothesized in this paper on the basis that an allocation that is favourable to PHC should somehow get translated into better utilization as a proxy for better quality. It is not possible to make this link with certainty given the design and data scarcity limitations o f this study. In addition interpretation o f utilization data should bear in mind the many influences on it including the well-known distance, economic, cultural, quality, and accessibility barriers (Barlow and Diop, 1995; Knowles. 1995; Criel. Van der Stuyft and Van Lerberghe, 1999). The accounting system needs urgent attention and should be modernized with the introduction o f computerized accounting software. The additional cost o f setting up this w ill likely be well off-set by having more timely financial statements, and saving time o f the personnel involved, and having more accurate accounts too. The hardware is already in place in all districts. What is needed is to acquire the software and do on-job training. The system can be introduced in a phased manner. This step may need active support o f the central government. Some o f the numerous steps can be eliminated while still maintaining a good degree o f transparency and accountability for funds. Experience has shown in any case, that the present elaborate system

169


does not deter those officials who want to divert public funds. As a review o f the experience o f The Gambia shows, although decentralization and better management may bring benefits, the anticipated benefits are not likely to be achieved to a reasonable degree for the resources expended. It is necessary that national policy as articulated by central government and donor agencies takes into account the need for not only devolving duties and resources, but also giving them the skills needed for decentralized decision-making as well (Smithson, 1995; Conn. Jenkins and Touray, 1996). Mechanisms need to be developed for monitoring and assessing performance in relation to objective measures o f progress in improving health status. The increasingly strong presence o f donor inputs should be channeled and re-oriented to building up technical capacity o f the district level public health system. Only in this way w ill there be a reasonable chance for eventual sustainability that goes beyond a financial definition (Hiscock. 1995; Smithson. 1995). The term “ public” here is meant to include both government and NGOs that are now being called Private Not for Profit (PNFP) (Giusti, Criel et al, 1997). The involvement o f the general public and technical personnel in the resource allocation debate needs to be actively promoted. I f either side is left out it w ill be more and more d ifficult to ju stify choices that are inevitable given the resource limitations to an un-informed body o f health practitioners and the public who demand to see that results are being achieved and equity is being promoted (Stronks, Stribis, Wendte et al, 1997).

170


7.6 Summary and Conclusions It is evident from inter-district comparisons that allocations are not data driven at the moment. There are some basic organizational constraints to the functioning o f the district health system as highlighted in this paper. These are issues that affect the entire district health care delivery system. Current strategies hinge on training individual health workers. We would like to argue here, that the context in which these workers operate are o f crucial importance as well. Improving decision-making has to take these general factors into account. There is a need to establish systems that are likely to survive the absence o f an individual worker. Focus should be on building systems rather than the current focus on training individual workers. Training is not bad per se, but it should be more productive to engage some o f the quite remarkable sums o f money in strengthening the institutional framework and team spirit. Decision-makers need to link the allocation o f resources to more objective evidence o f need than rely on the whims o f a few individuals who know how to get what they want. The central government and donors should be cognizant o f the existing differences in health status and take them into consideration in their allocations rather than rely largely on population and existing infrastructure as is currently the case. The proposed policies o f Health Sub-Districts (HSD) and Sector-Wide Approach have a chance to correct some o f the problems observed. However, the human side o f districts should be kept in mind since individuals and organizations influence each other. The organizational interests are sometimes in conflict with wider interests o f the society as a whole. Interests o f the individuals making the decisions might over-shadow the declared goals o f promoting equity, efficient resource use and quality as a means to improved health status unless measures are taken to promote use o f evidence in resource allocation decision-making.

171


7.7 Policy Implications Policies need to be developed that facilitate development o f district capacity to cope with decentralization. Such policies could focus on personnel standards e.g., making it a requirement to have professional administrators seconded by districts to the health department, standards on hiring accounting staff, etc. The wide differences between districts in terms o f health status need to be addressed by putting into operation the equalization grants from central government which are provided for in the constitution. Health specific indicators need to be developed and adopted. Such indicators should focus on the use o f data which can be made available with some improvements in the existing system rather than creating a new system. The policies need to spell out clearly the goals that the health departments should strive to achieve, and state the kind and scope o f assistance districts can expect as well as penalties for not maintaining basic standards o f practice. The capacity o f health personnel at district level to interpret and operationalize national policies should be developed. A review o f training effectiveness o f the numerous short seminars and workshops is urgently needed through which should be evaluated the advantages o f such vis-a-vis more advanced, longer training. Finally a public dialogue involving all stake-holders into the resource allocation problem should be encouraged. Such consultations could focus on clarifying the objectives society is trying to achieve through use o f available resources, different alternatives available, and why choices have to be made. This w ill hopefully pave the way for the stake-holders to use clearer and more objective criteria in deciding the priorities for allocation.

172


PART III: INTEGRATING REMARKS

CHAPTER 8:

GENERAL DISCUSSION AND CONCLUSIONS

173


CHAPTER 8: GENERAL DISCUSSION AND CONCLUSIONS

This last chapter presents an overall discussion o f the subject o f resource allocation decision-making in light o f what was found in the field. B rief comments are made in view o f some o f the available literature that was found. It also points out areas that would need further investigation.

8.1 General Comments In the effort to increase the share o f public resources allocated to health, as is so often heard echoed, a meaningful approach is to rationalize the use o f existing resources within the health sector. Only then, w ill health managers and advocates convince resource allocators that indeed the public is getting it's money's worth from the health sector. Case studies o f other developing countries such as Ghana and Pakistan showed that it is indeed possible, even in a developing country setting to move to use o f a needs-based, evidence driven approach to resource allocation combining the use o f burden o f disease and cost-effectiveness analysis (Hyder, Rotllant and Morrow, 1998). In the previous East African Burden o f Disease study referred to earlier (The W orld Bank. 1994), the recommendations that were made were aimed at the national level and not the district. Even then, they are said to have been largely ignored in coming up with the final list o f basic health services. These are said to have reflected donor and political interests more than what was actually found during the burden o f disease and cost-effectiveness exercise (a key informant. 1999). This study, therefore, explored the potential for introduction and use o f measures o f health outcomes and health status at the district level such as the D A L Y or HE A LY , through improvement o f information and strengthening district capacity to cope with decentralization. We have attempted to provide insight into the main questions guiding the research: How

174


are decisions about priorities for allocating public financial resources being made in decentralized district health systems in Uganda? We identified 6 main decision rules based on availability o f funds, historical categories, presence o f national guidelines, number o f people affected, popular local demand, and decision-makers expected benefits. On the next question o f how the process can be improved as a pre-requisite to promotion of better health, we have described some o f the existing problems and opportunities. These include excessive bureaucracy, persisting vertically directed focus o f implementation, lack o f confidence and skills by decision-makers at district level in use o f information to guide resource allocation, and overall health system factors o f personnel deployment, staff perceptions, and logistics. We have made the argument that improvement as defined in this study, .i.e., moving towards an allocation that is driven by measurable needs is possible. We have shown how some o f the available data and information sources could be used to guide decisions aimed at promoting aspects o f quality and equity. We have outlined a battery o f indicators o f district health profile and resource allocation that use available or obtainable information as a basis for future monitoring and assessment o f performance as well as comparisons across districts. The main resource allocation decisions being made including a comprehensive list o f budget variables for both PHC and hospital services have been documented.

8.1.1 Re-orienting to PHC and Equity Evidence suggests that the central government and donors are moving towards doing more for PHC but serious inequities in health status persist between districts. In keeping with the spirit o f the Uganda Constitution (1995) which argues and provides for the central government to provide equalization grants to areas in special need, it is argued in this study that allocations o f resources to districts by central government and donors should keep in mind this national objective. The ground

175


work has been laid through the elaboration o f techniques and indicators that can be used to analyze district resources and guide central authorities and donors in channeling support to districts. District managers themselves should be able to find the data collection techniques used in this study applicable to improving their own management. Resource allocation is influenced by different

stakeholders e.g., health workers,

administrators, politicians, the public. These stakeholders have their own criteria and interests in influencing allocations. In practice, official criteria may not be followed. Local bargaining plays an important part in determining how resources are allocated. As Gladwin (1989) points out. many carefully planned projects usually fail because the technocrats fail to understand the reasons why the decision-makers who make choices act as they do. Policies may be set and guidelines issued at the national level. This does not mean they w ill get translated into action to achieve the desired results. A recent example o f that is the introduction o f national supervision guidelines. Although there are signs o f starting to try out the guidelines in a few districts, most o f the others had not made a move even though the guidelines were made available and training done (a key informant, 1999). It is expected that under decentralization the policy goals o f quality, equity and efficiency o f services w ill be met. These goals are often in conflict and have to be operationalized in a complex setting o f logistic, bureaucratic and political realities. Uganda has for long stated that the policy is to re-orient services towards PHC but the reality is that most resources are still geared to curative interventions and hospitals45just as they did in the past. The analysis o f budgetary allocations shows that more resources are now being directed to PHC, especially since the introduction o f the Conditional PHC grants. Estimates by M OH o f actual expenditure on PHC, shows only an approximate 10% increase in real terms growing from 25% in

45

Hospitals take a significant part o f the health budget especially in form o f resources spent on personnel. In many cases a significant amount o f resources allocated to P H C often goes to curative services or extending physical structures or providing drugs.

176


1992/93 financial year to 34% in 1997/98. The hospital share o f the total budget declined in real terms over the same period from 70% in 1992/93 to 52% in 1997/98, while expenditure on management (M O H headquarters) increased over the same period from 5% in 1992/93 to 14% in 1997/98 (Mugarura, 1999). It should be noted that analysis o f budgetary allocations indicates that on average, the share o f total district health budget allocated to PHC in 1997/98 was 59%. The disparity might be partly explained by the fact that the budgetary allocations figures might include sources not reflected in M O H expenditure estimates but these are likely to be small amounts since the major sources o f funding to the district are central programmes, and bilateral and multi-lateral agencies. There are district based donors who might incur expenses not reflected in the MOH figures e.g., the value o f salary for expatriate personnel working in districts and goods obtained from abroad as donations may be hard to estimate. It should be noted that allocations do not mean all that money w ill be released. However, the MOH estimates suggest that over the last 3 financial years there has been an improvement with disbursements for central funds averaging between 80% to 90% o f amounts budgeted. Another interesting observation in the M O H data quoted above is an estimated 63% o f total household expenditure on health being met by private sources with 19.8% and 17.5% from donors and central government respectively in 1997/98 (Mugarura, 1999). So, the share for PHC may have increased in relative terms but the greater part o f health care costs at the household level is largely being met by individuals.

8.1.2 Public/Private Mix In recent years, there is an increasing focus on the role o f public and private sectors in relation to quality and efficiency o f health services (Bitran, 1995; Giusti. Criel et al, 1997). Information on costs and quality, as well as efficiency in both public and private facilities needs to

177


be augmented. Key informant reports indicate that in Uganda some private practice is subsidized by the informal use o f publicly funded facilities, mostly hospital diagnostic and operative facilities. The extent o f this is not well documented although it is widely known that most people in private practice are in fact also government employees. It is thought that people perceive quality in private facilities to be better mainly because o f staff and drug availability, and shorter waiting time. However, people in Uganda generally recognize that they are more assured o f the technical competence o f health workers in government facilities (Okello, Konde-Lule et al, 1997). The absence or lack o f enforcement o f regulation o f private practice, generally has the effect o f subsidizing private practice from public funds! An example o f this happens when a health worker employed in a public facility uses it for the treatment o f patients (e.g., surgery) who were first seen in the private clinic. Such patients w ill normally pay directly to the health worker and the public facility actually loses. This complaint came out from most key informants in all districts. There are also often institutional tensions between publicly funded and private facilities, profit and non-profit. One suggestion is to further distinguish between those private facilities which are geared primarily towards profit for the owners (e.g., private clinics) and those which have a prim arily charitable role, such as those founded by missionaries, and other NGO charitable NGO facilities. The former could be called Private-for-Profit (PFP) and the latter Private-not-for-Profit (PNFP) (Giusti, Criel et al, 1997).

8.1.3 Compliance with Standards Although there is a growing presence and importance o f the private sector in health care provision, the government remains the main provider o f curative care in many rural areas as well as practically the only provider o f preventive services. In the interests o f maintaining some minimum set o f national standards and safeguard the interests o f minority groups, the central government

178


through it ’ s role under decentralization o f setting policies and monitoring compliance with standards consider it a priority to build up capacity at district and lower levels. It should then have the possibility to intervene e.g., by sending additional resources i f such goals are threatened in any district.

8.1.4 Use of Available Resources In one district, the comprehensiveness o f the records kept and indicators calculated at the hospitals was often in sharp contrast to what was seen at the district level. This is also a reflection o f the dichotomized nature o f the health system in Uganda where hospitals function largely independently o f the rest o f the district health system. The M O H in the past focused most o f it's attention and resources on hospitals with the excuse that PHC units were outside their control. The issue o f the introduction o f lunch allowance mentioned before further points to the bias o f the MOH towards hospitals. Expertise could be tapped from the hospital or NGO to improve district level systems if bureaucratic stereotypes and administrative tu rf rivalries can be overcome or reduced. These are mostly between the traditional District Health Team (D H T) and hospital management. This is only possible if the hospitals are integrated into the functioning o f the district health system as the Hoima District (Uganda) experience suggests (Criel and Pariyo. 1997). The new HSD policy is geared towards using available facilities. The HSD entity w ill be an electoral area or constituency46. The biggest health facility in that constituency such as a Health Centre type IV or hospital w ill be staffed and equipped to act as the focal point for the HSD. They w ill use existing facilities whether these are traditionally government facilities or NGO (M OH, 1998; Kiyonga, 1999).

46

This is a reflection o f population concentrations.

179


8.1.5 Ownership and Control There is generally a perception on the part o f district health managers o f not being allowed to make their own allocations or payment decisions at the district level47. However, a review o f the various circulars and guidelines that were sent to the districts, as well as key-informant interviews with officials at different levels showed a rather different picture. It is true that guidelines were sent and in some cases boundaries are specified on limits to spend on different activities e.g., it is stated that not more than 30% o f the PHC conditional funds should be spent on malaria. However the details o f which intervention to use e.g., promotion o f bednets, community-based distribution of chloroquine, etc and the amounts to spend are generally not spelt out. This is definitely quite different from the days before decentralization when the centre used to state how much to pay say for supervision or lunch allowance, how much for fuel, etc.) A ll the interviews conducted with officials at the M O H and M O LG concurred with the observation that emerges from the study that indeed districts have real power. This is further evidenced by the incident mentioned before where one district had seriously strayed from the guidelines in putting up a new hospital ward and even ignored directives from the MOH to stop. The M O H has a constitutionally stated mandate o f assuring provision o f quality services to the citizens and the clear role o f setting standards and monitoring compliance w ith those standards (Uganda. 1995: Uganda. 1997). The M O H, however, is said to lack adequate and clear legal authority to enforce those standards (A key informant48, 1999). The Local Government Act (1997) which is the main legal document referred to in relation to district operations under decentralization clearly gives power to districts and is vague on the role o f central ministries. It was reported that the MOH

47

This came out in all the 3 FGDs. It was also mentioned by many o f the respondents during the structured interviews. 48

Identity o f key informants who made specific statements is being withheld in honoring the commitment for confidentiality. 180


attempted to take up the issue with the M O LG which drafts the laws but they were told that MOH was being paternalistic and districts have the right to refuse M O H directives (A key informant, 1999). Another official suggested that the M O H may not have been very convincing as to why they should retain some power o f censure over districts in the use o f health resources (a key informant, 1999). The apparent ambiguities in the law are presumably being addressed through an on-going review o f the Local Government Act. One central official, however, felt that although M OH seems to have a point, they did not articulate their interests convincingly enough. The situation is bound to get more complicated in the short term as districts move to implement the sector-wide approach (SWAP) under which all government and donor resources w ill be made available to districts in one pool. Unless the potential legal loopholes are addressed, districts w ill practically have more resources with a free reign on how to spend them. The potential for abuse o f these funds by districts which has been noted repeatedly in a number o f districts since decentralization may actually increase as has been known to happen in other countries (Collins and Green, 1994). That districts have real power was confirmed by all the Chief Administrative Officers interviewed. They, however, pointed out that the legal power they received under decentralization is not adequately matched with the resources to undertake the new responsibilities under decentralization. On this aspect the views o f the CAOs differed from the councillors o f the Health Committee and the health technocrats, most o f whom did not feel really in-charge. The new Health Sub-District (HSD) policy is aimed at addressing some o f the problems of resource management that are currently being faced at district level. However, the resources w ill be stretched even more, since the policy envisages a doctor at every Health Centre IV. In effect this w ill create 214 new HSD teams, each serving an average population o f about 20,000. The new HSD is meant to address the problems o f over-centralization at the district level that followed centre to district decentralization. It is expected to bring services nearer to the people, and free the District

181


Director o f Health Services to focus on planning and co-ordination o f HSD operations (M O H . 1998; Kiyonga, 1999). The current legal “ laissez-faire" state o f affairs and duplicity o f vertical programmes that occupy the attention o f M O H officials makes it d iffic u lt for the M OH to monitor let alone enforce basic standards. It is proposed to put central resources in a central “ basket" for districts to use in an integrated manner. This is the focus o f the new Sector Wide Approach (SWAP) that is expected to be implemented soon.

8.1.6 Information Costs Information gathering is expensive and unless it is going to substantially change the decision, managers should attempt to use information that is already available from other sources. An analysis o f the cost o f expected information could be done. This would examine the decisions to be made and the information requirements. By looking at whether the new information to be generated would make a difference in the quality o f the decision made or not. as well as comparing this with potential cost savings, managers would then be able to decide to pay for more information or use their best guess!

8.1.7 Improving Allocation Decision-Making In order to promote a more evidence based resource allocation, a "per performance" or "reward and censure” approach could also be tried, e.g., making the release o f funds to districts contingent on not only having provided a good financial accounting (which is tedious but rather easy to do if systems are in place), but also for districts to account for the actual use o f those funds, e.g. what is the outcome, any improving coverage, etc. This brings in the question o f allocative efficiency. Which services should public funds be spent for? Which ones should be left to private

182


funding? The East African Burden o f Disease Study recommends use o f 3 classifications o f interventions, community, preventive and curative. An argument is presented to leave certain aspects o f health care to be funded by private individuals, especially for those conditions which are both rival and exclusive49 (The World Bank, 1993; The World Bank, 1995). Information on the effectiveness o f the many different public health interventions being employed is urgently needed. Some o f these interventions are probably quite effective and lessons could be learnt from their applications, and adopting them on a wider scale might lead to cost savings. Others on the other hand, may be o f doubtful benefit, in which case their continued use constitutes a clear waste o f scarce resources. A t the moment, implementation o f different initiatives seems to depend on the particular preference o f the actors or their financiers but information on the objective basis o f such interventions is sorely missing. There are generally good political and pragmatic reasons why things are the way they are. What we are arguing here is that the process can be improved within the means currently available, even though technical and political considerations must co-exist. The process is right now skewed towards political considerations since objective technical information is generally un-available. This makes the decision-makers more liable to being influenced on political considerations, or to weaken their case in arguing for one intervention alternative over another (Bolman and Deal, 1991; Alexander and Morlock, 1994). The M O H for instance had good technical reasons to oppose the construction o f a new hospital ward in one study district but apparently did not articulate it in a way that was convincing to the district. One might think that i f there was practical information on the alternatives available, and what the likely consequences o f the choice that the district leadership was making on political grounds would be, they might have been prepared

49

Rival here means use by one person prevents use by another, and exclusive refers to the fact that benefits are limited to the individual getting the service, e.g.. malaria treatment as opposed to a health education session. 183


to listen and a misallocation o f scarce resources might have been averted. I f such information had been available, the political leadership might have been better prepared to resist pressure from the electorate for a new ward as the solution to their problems o f over-crowing at the hospital. Such alternatives could have included re- organization o f the hospital services, a better management of the referral links between the levels o f care and better capacity to manage problems at the health centre, nearer to where the people actually live. Eventually the solution chosen by the district leadership works to the detriment o f overall health o f the district by aggravating the recurrent cost problem. In fact on a follow up visit to this particular district, they had started negotiating with the M OH for equipment and additional staff for their new ward! Such recurrent cost problems are increasingly a problem in many developing countries, in a situation where funding for health is not expected to increase in real terms and makes the problem o f sustainability o f health services even more acute (Lafond, 1995). Since much o f health funding is being provided from outside the district (M OH, projects or NGOs) it should not be so d iffic u lt to implement a system o f performance accounting. Even though it might appear to create problems for the implementers in the short run, it would also give those who are doing a good jo b the means to advocate for more resources. A related benefit would be to have the effect o f making people question their local and national leadership and demand for better accountability in the use o f resources. One M O H vertical programme has started to enter service contracts with some NGO health providers on an experimental basis. Expected service outputs are being stipulated in return for financial inputs from government sources. However, it w ill be necessary to carefully evaluate the inputs and expected outputs. It is quite possible to have a situation where such a service contract could place a greater burden on the contractee than the means provided by the contractor (government). One example that illustrates this problem is the case o f one NGO institution that did a calculation o f the cost implications. It turned out that in order to implement the

184


new set o f activities being proposed by the central M O H programme, it was necessary to deploy a doctor full time. The effect o f this would be to take the doctor away from existing clinical duties o f the NGO institution. Patients, usually the more disadvantaged, would suffer more by doing this unless a new doctor was hired. This would have had the effect o f deviating the institution from it's declared mission o f caring for disadvantaged communities (personal communication, Giusti, 1999). It would be imperative to consider not only direct service delivery costs but indirect and opportunity costs as well. However, the idea o f service contracts and performance assessment in relation to publicly funded services is an attractive one. If the experiment with NGO units works well, it could conceivably get extended to health units operated by the local government authority as well.

8.1.8 Rational Allocation Framework A scientific management view-point such as Bureaucratic Theory with its impersonal roles o f "offices", clear, written rules and guidelines, dealing with "cases" rather than persons would seem to be an appropriate answer to the problem o f nepotism and recruitments based on “ connections". However, it is impossible to formulate rules to cover every aspect o f the functioning o f a district health system. As the concept o f bounded rationality (March, 1994) would tell us. districts are limited in the amount and quality o f information they can get or afford. It is d iffic u lt to find the kind o f data on morbidity and mortality on the basis o f which burden o f disease type o f modeling or other quantitative techniques such as expectation o f u tility could be used. However, the use o f burden o f disease approaches have recently been demonstrated using data from Ghana and Pakistan, two countries whose data constraints are not much different from Uganda. There is potential for further exploration in Uganda, using district level data. W hile the East A frica Burden o f Disease study provided information on the Uganda M inim um Health Care Package at the level o f national policy, the investigators also point out the need for data and techniques that can be more readily used by

185


district level health managers. Having one person, the DDHS, as the highest authority on health in the district has the advantage, as Weber points out, o f allowing the possibility for rapid decisions, free o f the necessity to compromise. This is useful in the running o f a public health authority which often has to deal with complex scientific matters and disease emergencies. According to Weber bureaucratic control implies the exercise o f control on the basis o f knowledge which is both technical and factual based on experience in the organization. However, information acquired in training decays and may not be refreshed fast enough, routinely collected information is often inaccurate or irrelevant. It should be noted that much as a rational decision-making framework is a desirable ideal, districts are staffed by human beings who do not always act rationally even when provided with the right information. One can hope to approximate to this ideal situation by clearly identifying the desired objectives and then trying to act in a way that is generally consistent w ith those objectives over the long run. Short-term irrational behaviour may not be a major problem but when inconsistencies persist over the long run, then the already bad health situation might only get worse. It is important to make efforts to get better value in terms o f health status for the money being spent since resources are unlikely to increase significantly in the short term. What matters most is how people view the reality and work together and achieve society's objectives and not so much what rules exist. The rules are designed as means but often risk to become an end unto themselves (Mintzberg, 1983). On the other hand a certain degree o f order and efficiency is necessary. The suggestion by Kanter (1991) that the reality is more complex than a dualistic view o f the world, seems quite relevant here. Often it is not a question o f a clear cut this or that, right and wrong. The main value o f various theories and approaches is to give us a framework o f studying the question at hand and be the basis for possible explanations that might be obtained from our research. Different theories and approaches are often different faces o f the same reality even though they may

186


sometimes appear to be contradictory. Indeed perhaps contrary to the rational view, there is no "one best way" o f allocating resources. The fact that society makes choices which may change over time makes it unlikely that one best way serves for all time. Rather as Mintzberg (1994) pointed out what is needed is strategic thinking. Districts need to use knowledge o f D H M T and district leaders from all sources, i.e., from their own insights from experience and from the data from research and epidemiological or burden o f disease type o f studies. These should then be synthesized into a vision for improving the health status o f the district population and setting a course o f action for the immediate and long-term future. It may be necessary, where feasible, to increase the levels o f resources to reduce conflict among different interests competing for the scarce health care resources, as advocated by Alexander and M orlock (1994). Given the rather severe lack o f data on morbidity and mortality on the basis o f which quantitative models o f resource allocation like those used in developed countries could be applied, it is important to start the process by documenting what is going on and laying the basis upon which more quantitative information can be collected. An attempt has been made in this work to contribute to the process. Better understanding o f current practice can suggest how it could be improved using established management techniques that have been shown to work well in developing countries (Kielman. Janovsky et al. 1991; Maier, Gorgen. et al, 1994; Smith and Morrow. 1996).

8.1.9 Decentralization and Democracy The review by Kanter (1991) challenges the long held view o f scientific management rationalists and Weber's theory that bureaucracy is the only rational (economically efficient) form o f organization. Society’ s perspective implies the use o f other criteria, not only economic. Decentralization is part o f society’ s attempt to distribute the power to allocate resources. The imperative to maintain social order and appear to be fair to all parts o f the country or district often

187


overrides economic and technical considerations. A political approach offers an alternative perspective. Resource allocation in real life is more often a political activity than a technical one. The centre has played an important role in beginning the process o f promoting equity through the introduction o f PHC conditional grants. Unfortunately, they were largely introduced in response to a perceived lack o f commitment by district authorities to fund PHC. This is ironic given that a major justification for decentralization was to bring control and accountability for delivery of essential social services to the district, nearer to the intended beneficiary, to whom district leaders are now more directly accountable. Political expediency o f pleasing their electorate through high visib ility projects such as extension and renovation o f infrastructure with little attention to functional capacity has been the main driving force behind some o f the main allocations for capital projects by districts. As pointed out by Collins and Green (1994), decentralization sometimes has some negative consequences. It does not always mean that the district administrative authorities and politicians who are nearer to the people have the people’ s best interests at heart. Moreover, they as leaders o f their people, often take it for granted that what they are doing is alright since it is on behalf o f those same people. The situation is made more complicated by the low levels o f general education, widespread poverty, and low political awareness. It is a paradoxical situation because on the one hand, people have to take control o f their own affairs, but on the other, they are not sufficiently empowered with the necessary information to make informed choices. Should decentralization proceed gradually, giving people time to leam along the way? The risk here would be that the process can get derailed by the numerous problems and people would probably give up before results can be seen. In the case o f Uganda, the un-precedented speed with which decentralization was done has resulted in problems that perhaps could have been anticipated and prepared for better. However, most district people would not surrender their new found power back to the centre in-spite o f the problems that exist. There have been many remarkable developments and achievements along side some predictable

188


problems. In the struggle for democratic control over their own affairs people are bound to make mistakes but they w ill hopefully learn along the way. As the population gets more informed about their rights and responsibilities under decentralization, they w ill hopefully demand for greater accountability on the part o f local and national health bureaucrats, administrators and politicians. In order to begin to improve the process o f resource allocation, it is important to first understand how and why things are happening at the moment. Modeling how people make real world decisions is a first step in understanding specific decision-criteria that are used by most individuals and organizations. Equipped with this insight, policy-makers might be able to influence the decision making process by formulating new policies and regulations that are intended to make things better for the targeted group (Gladwin, 1989).

189


8.2 General Conclusions The need for objective and practical information, that is understandable to a decision-maker who has no technical training in health or economics is obvious. It is quite a huge challenge to extend this further to include the tax payer, who ultimately should have the say as to what resources should be allocated for. As one Ugandan official put it, “ people don't go to parliament to discuss DALYs, they discuss the costs in terms o f money, but the benefits measured in D A LY s is hard for them to understand, people see money, they don’ t see D A L Y s" (a key informant). This study attempted to use data and information that is largely available or can be collected to enable district and national decision-makers to improve the way they allocate their health resources. It is argued that the centre should support the districts in building up the required capacity and some suggestions o f what can be done are offered. Finally, a collection o f about 50 district profile and resource allocation indicators was generated from available data. It is hoped that district health managers w ill be able to use some o f the same indicators and techniques and spreadsheets developed in this study to prepare the ground to improve their own analysis and resource allocation decision-making. A main effort in the future should be to build up the kind o f management and information handling skills and logistics that are needed for a decentralized health system in all districts. It is hoped that this research has contributed to that process. Information has been generated to guide development o f policies and further research on resource allocation in the decentralized health system in Uganda. The information is o f potential interest and benefit to other developing countries as well.

190


8.3 Future Research A rational decision-making framework was used to guide data collection and analysis. In this framework, decisions are made to maximize expected u tility o f the subject. A First step has been made with the identification o f the main decision alternatives being made, information availability, as well as the main constraints faced. Decision-maker preferences were explored. This being an exploratory study, it was not possible to make a clear distinction between the decision-maker's personal u tility and his/her perceived societal u tility. Such a distinction could be pursued further in future research. Preliminary scales for assessing decision-maker perceptions and utilities have been outlined but w ill need future research aimed specifically at testing them more rigorously. The unique aspect o f this research was to see how the gap between actual and desired resource allocation decision-making could be narrowed. Preliminary information has been collected on opportunities, operational constraints and possible solutions in the use o f an evidence based approach to resource allocation at district level. Such an evidence based approach is the use o f measures o f burden o f disease e.g., the D A L Y or H E A LY , and cost-effectiveness in resource allocation decision-making. The setting o f this study in a poor Sub-Saharan African country like Uganda has identified issues that might be expected in other countries in the region. The incorporation o f decision-maker perceptions and expectations in trying to explain what is happening was also important and unique. It included personal factors such as prestige, income, job tenure, as well as health outcomes into the decision-maker’ s subjective expected u tility judgement. It has outlined the elements o f decision-making rules being used in practice and their constraints. Future studies o f public health resource allocation in a Sub-Saharan African country might build on this by seeking to strengthen, clarify, or disprove them. National estimates o f some basic resource allocation indicators have been made. Policy makers could build on and use these to monitor progress o f districts in the struggle to promote equity and improve health status. Future research should examine

191


the field effectiveness o f the interventions identified in this study that are being used and come up with more refined cost estimates. V ital statistics are also needed on the district populations. It is important that policy makers think o f developing a system o f collecting population based vital statistics on a routine basis. The up-coming census o f the year 2001 provides an opportunity for up dating the demographic estimates but a routine system should be developed perhaps on an experimental basis in selected areas. Finally, the effect that training courses o f district level and health unit operational level staff have on actual performance should be form ally evaluated. Training is currently occupying a significant role in the use o f personnel time and share o f the health recurrent budget. The currently on-going multi-centre evaluation o f the effectiveness o f IM C I w ill provide information on one o f the main training strategies being implemented in many developing countries. Another on-going initiative in Uganda is the recent introduction o f national guidelines for the supervision o f health services. Different strategies in use o f resources at various levels o f the health system could be evaluated, preferably with the active participation o f the implementors.

192


PART IV : REFERENCES

APPENDICES

BIBLIOGRAPHY

193



APPENDICES Appendix A: Structure of Community Representation and Local Governance in Uganda

Structure o f Local Councils

RDC

D L C (LC V )

I Reports lo ihe President)

l A ccountable lo D istrict Electorate)

IC A O ) To p A d m in is tra to r tall departments report here)

LC II!

LC II!

(S 1C C h ie f)

IS 1C C h ie f )

L C II I Parish C h ie f)

194

|S 1C C h ief)


Appendix B: Hierarchy of Groups Involved in Health Planning and Budgeting

Planning and Budgeting - Link W ith Political Structure District Local Council (Elected Politicians) (Approves Plans/Budgets) 1 Social Services Committee (Health Education. Welfare)

DHM T Health Department (Plans with NGOs)

.

l

...

Finance Committee

1 Works

Production


Appendix C: Basic Structure of District Health Management Teams in Uganda

District Health Management Team DDHS {Overall Technical Head)

Deputy DDHS

Medical Superintendent

Medical Officer

I District Hospital

1T0 Head HSDi

N'GO Co-ordinator

DHV

TBLS

DHE

DHI

D A ID

Executive Officer

(Public Health Nuising

(TB/Leprosy)

{Health Education)

Environmental Health)

iDrus Inspectorate)

i Accounts i lOversees Clerical Staff)

196


Appendix D: Hierarchy of Officials Managing District Health Finances

Financial M a n a g e m e n t CAO (Overall A d m in istr a tiv e H e ad ) D e p u ty C A O

DDHS ( Health D e p a r t m e n t ) Executive Officer (A c c o u n ts )

C h i e f Fina nc ia l Officer (District Treasurer) A c c o u n t a n ts

A c c o u n t s Clerk

197

Internal A u dito r

Tax Officer ( B u d g e t Officer)

Appendix E: Steps in Building a Rational Decision-Making Model

The follow ing are the steps needed to derive a rational decision-making model based on the expectation o f u tility frame-work suggested by Dennis Lindley (1985):

1. Make a list o f the different decisions being made for resource allocation or district health care in general. 2. List all the main events o f relevance that may happen following a particular decision, and the outcomes or consequences o f decisions or public health interventions carried out. 3. Make best estimates o f probabilities o f outcomes given particular decisions. These could be estimated based on information available from studies (e.g., estimates using Delphi technique, etc). Some o f the probabilities would then be calculated in accordance with the laws o f probability and the Theorem o f Extension o f the Conversation. Coherence could be checked using calculations based on Bayes' Theorem. 4. Make best estimates o f utilities attached by decision-makers to different consequences or outcomes o f their decisions or interventions implemented. The outcomes would need to be compared to the most desirable situation (given the country's circumstances) and national standards as appropriate. Where such do not exist, a scale could be developed based on the best practice frontier (BPF) or Data Envelopment Analysis (D E A ) technique. This would consist o f comparing each outcome in a particular district (or region) with the best possible outcome that has been achieved by a district (or region) in a similar situation (even though it may not be identical). This information would then be summarized with the aid o f decision tables and/or decision trees as in the following example (much simplified compared to real life!):

198


Uncertain Events

1

2

3

Decision 1 Decision 2 Decision 3

u(On) u (0 21) u (0 ?1)

u(O i:) u (0 22) u (0 22)

u(On ) u (0 23) u (0 „)

Probabilities

(h

o2