The Public Health Care Planning Problem: a Case Study using Geographic Information Systems Sophie D. Lapierrey
Justin A. Myrickz November 16, 1998
Gail Russellx
Research partially supported by NSERC Grant OGP0184219. CRT and Dept. of Math. & Industrial Engineering, Ecole Polytechnique, C.P. 6128, succ. Centre-ville, Montreal, QC, Canada H3C 3J7,
[email protected] . z School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA 303320205,
[email protected] . x Gail Russell, Oce of Planning and Evaluation, Fulton County Health Department, Atlanta, GA 30308,
[email protected] . y
1
Abstract We address the problem of designing new networks for the delivery of public health care services in the United States. The paper is based on a case study design conducted with the Fulton County Health Department (Atlanta, GA, U.S.A.). The research contribution this paper makes is twofold. First, it presents the issues of delivering health care services through a mix of xed health centers, satellite facilities, and mobile facilities. Second, it gives insights on how to use geographic information systems to design new health care service networks. Keywords: Public health care planning, satellite facility, mobile facility
2
1 Introduction Since 1948, when most public health centers in the United States were built under the HillBurton Law, nursing practices and population needs have changed. Today, public health departments in the U.S. still continue to care for the poor and protect the population against contagious diseases; but nursing has changed and its practitioners require more specialized training as well as team e ort to improve their knowledge and skills. Larger nursing teams are required to deliver health services than was the case in 1948. Moreover, population needs have also changed. The size of the low-income population has increased and the gap between the rich and the poor has widened. Today's public health departments often have to care for more low-income households spread out over a larger territory. All these changes are forcing public health departments to reconsider the way they provide their services to the population. One major challenge public health departments are facing is to deliver services closer to where their customers live or work, while facilitating the e orts of the larger teams needed. These objectives can be reached by reducing the number of health centers and increasing the use of satellite and mobile facilities. We refer to a satellite as a facility where a service is given by sta that comes from another facility called the satellite's base location. An example of this would be u shots given to the elderly at a nursing home by local public health nurses. A satellite can also be a mobile facility if services are given within the vehicle used for transportation to satellite sites. An example of this would be mammograms being given in a mobile facility located in a parking lot and operated by a local hospital | the base location | where the mammograms are read at the end of each day by radiologists. The concept of satellite facilities is not new but technological advances have made the use of mobile facilities, in particular, more prevalent. Several health care organizations have been innovative in using mobile facilities to deliver primary care (Anderson [1], Hodgson et al. [4]), to o er screening mammograms (Bewerse and Williams [2], Destouet and Evens [3], Rubin et al. [12], Sickles et al. [14]), and to allow hospitals to share expensive imaging equipment (Rajagopalan [10]). What has not been addressed though, is how to integrate all satellite facilities with xed facilities, taking into account how they might alter the entire 1
delivery network. This paper lls this need. To date, public health departments have used a reactive approach | which has been relatively successful | in addressing new needs on a local basis. Since they are responsible for monitoring population needs, public health departments are often the rst agencies to notice any spread of poverty or other social changes. They respond to these needs by o ering new services or by trying new delivery strategies, such as using mobile facilities to supplement existing facilities. But these changes constitute minor modi cations since they do not alter the core method of delivering health services. In order to fully exploit the potential of mobile and other satellite facilities, a system-wide approach is needed to address the major restructuring that the changing health care environment has imposed on today's public health departments. In the case study presented here, we examine major delivery changes made using such a system-wide methodology. Our approach goes beyond the simple assessment of changing population needs and market segmentation, the most common way health care systems are analyzed [11]. We employ a geographic information system (GIS) to fully utilize all elements of the network and ensure that we obtain the best-designed delivery system. To our knowledge, the most comprehensive GIS applications in health care planning are the Toronto [7] and the Melbourne plans [6], both being population need assessment and market segmentation analyses. These GIS applications, however, did not use extensive models like the one we use. This paper is structured as follows. First, we present an overview of the case study, giving the context in which the Fulton County Health Department (FCHD) was attempting to design a new health care delivery network. Second, we present modeling issues related to using xed and satellite facilities, before providing a brief description of the planning methodology used. This is followed by an in-depth presentation of the case study, with an emphasis on the availability of data and the GIS technology used at each stage of the development of the new health care network. Finally, we discuss what has been learned about integrating di erent types of facilities in re-designing Fulton County's health care system and the merits and diculties of using GIS to do so. 2
2 The case study Fulton County is the most populous county in the state of Georgia with 648,491 people [15]. It is 75 miles long, has a total area of 532 square miles, and has 11 municipalities, the largest one being the city of Atlanta. The county has an extensive highway network and a well-developed public transportation system with trains and buses. Fulton County's population is 50% black and 48% white, with a small Hispanic population of 12,500. The greater Atlanta metropolitan area is the fastest growing in the United States, thanks to a diverse economy. Fulton County is characterized by great auence in the north and pockets of poverty in the central and south central areas. Atlanta is the fourth poorest city in the United States among those with at least 250,000 citizens. In 1989, 18% of Fulton County residents were living below federal poverty level standards; children were particularly a ected, with 32% of those under ve years of age living below the poverty level. Figure 1 shows the distribution of residents with income below 185% of the poverty level, by census tract. Note the concentration of poverty in the county's central area.
[Locate gure 1 about here] [Locate gure 2 about here]
Each year, the Physical Health Services division of the FCHD receives 150,000 visits from clients at its 20 health centers (Figure 2), as well as visits to schools and other community facilities. Its annual budget in 1993 was $15 million, most of that coming from Fulton County. The division's main goal is to help families maintain a high level of health and wellbeing through routine health screenings and wellness programs. Public health nurses, along with support sta , provide services in the majority of physical health programs including child health, adolescent health, family planning, communicable diseases, nutrition programs, high risk infant programs, and refugee services. The most common services delivered by the 88 nurses are childhood immunizations, u shots, pre-school physical exams, sex education, gynecological exams, and testing for communicable diseases. Dental hygienists and dentists provide comprehensive dental services to pre-school and school age children. The existing 20 health centers were built under the 1948 Hill-Burton Act, most of them 3
located near elementary schools. Since then, Fulton County's population has expanded and taken on a di erent distribution, clinical practices have changed, and buildings have aged. The entire network of health centers has to be rebuilt or relocated into rental space. The Health Commissionner launched a major re-design project in 1994 and requested the utilization of the latest GIS technology.
3 Methodology The design of any health care network will necessarily be unique and the data available for input will vary from case to case. Therefore, the methodology needed for designing a new delivery network must be exible and interactive. The methodology developed by Lapierre et al. [5] ts these requirements. Several assumptions underlie this methodology and before presenting it, we elaborate on the key assumptions and the issues addressed. For most public health programs, participation in prevention programs and access to medical care is of primary importance; when designing a new health care network, the goal is to keep the same participation rate or to increase it. However, opening or closing facilities can result in a degradation of service for some people and an improvement in service for others. Moreover, changing the type of delivery strategy | a xed or a satellite facility | can alter the number of participants. Therefore, it is essential to have a model for predicting population participation for a given network. Identifying the factors leading to participation in a program | or lack of such participation | requires public health expertise. Work done in this eld typically looks at the socio-economic characteristics of an at-risk population and their impact on participation. Nevertheless, the two most important factors to explain participation are generally recognized as being distance | the more distant a facility, the less likely people are to participate | and size of facility | the larger a facility, the more likely it is that people will seek services. In our case study, this translates into the following assumptions: (1) the number of participants should increase as we add more facilities | the facilities will be closer | and (2) xed health centers should draw more distant customers than satellite facilities since the satellites are smaller. 4
Whether to use a xed vs. a satellite facility must also be decided. The trade-o is that xed facilities can care for a larger number of customers than a satellite facility and so cost less to operate per customer, but they involve a larger initial investment. Rather than focussing on cost, however, health care planners are more accustomed to using the concept of minimum population size to justify the presence of services. The two concepts are inter-related, and the methodology we use takes both into account. We begin with a cost-bene t analysis and then translate it into number of customers per type of facility. The assumptions and concepts noted above led to the development by Lapierre et al. [5] of the methodology presented in Figure 3. This methodology consists of ve steps. Step 1 requires collection of electronic data such as population census data and maps. Step 2 models customer behavior; i.e., given a set of xed and satellite facilities, where customers will go if they participate in prevention programs. Then, the maximum distance a satellite facility or a xed facility can be located from any customer is determined based on the participation rate targeted by the health care agency. Obviously, the customer behavior model strongly depends on the level of detail of the collected data. In step 3, costs are estimated for xed and satellite facilities in order to de ne the minimum number of customers needed to justify the opening of xed facilities. In step 4, this minimum number is taken into account as xed facilities are rst located so as to serve as many customers as possible. Satellite facilities are then added to serve areas not well served by the xed facilities, the criterion here being the maximum distance customers will have to travel. Once these locational decisions have been made, satellites are assigned to a base location and are designated as being either xed or mobile. At this point, the decision is also made as to which mobile satellite locations will be served by the same mobile unit. Finally, in step 5, schedules are drawn up to determine when satellites will be visited by a travelling nursing sta or a mobile unit. This research project was a multidisciplinary e ort involving health care planners, nurses, and operations researchers. An interactive approach was adopted in order to allow each group to share its expertise at every step of the process. The assessment of population needs was made on Atlas GIS at the Oce of Planning and Evaluation of the Fulton County Health Department. Quantitative analysis of the proposed network was performed 5
using the CAPS Logistics Toolkit (an advanced GIS used in transportation analyses) at the Georgia Institute of Technology. A qualitative analysis of the network was performed by administrators and nurses of the FCHD. We elaborate on this work in the next section.
[Locate gure 3 about here]
4 Design of FCHD delivery network The general motivation for the FCHD to re-design its delivery network and to re-think its delivery strategy is to provide better access to higher quality health services at lower cost. There is no precise number of health centers to be located, but 20 centers were viewed as too many because of the diculty of coordinating nursing sta | especially at the smaller centers | and of the high cost of building maintenance. Administrators and some FCHD sta recognized that by adding xed and mobile satellites to the health care delivery network, services to the population would not su er if the number of health centers were reduced and, in some cases, could even be improved. However, concrete proposals were necessary to convince remaining sta and taxpayers of the validity of this view. This project is designed to provide such information.
4.1 Step 1 { Data Collection Three types of information are needed for this project: customer data, details about the sites for existing and proposed facilities, and area maps. Relevant customer information for this project involves the customer's location and socio-economic characteristics. To begin with, such information is gathered from current customers in the following way: addresses are sorted by zipcode area and compared to census data to identify the socio-demographic characteristics of existing customers. To achieve a ner level of detail, customer addresses can be instead matched to a geographic database | to extract the longitudinal and latitudinal coordinates | and then grouped at the census tract or block group level. This latter technique is more complex to perform; its accuracy depends on the quality of street maps, customer databases and address matching software and can be quite variable. For example, in a test of two such software packages 6
(each with di erent database information), accuracy ranged from 60% to 90%. If databases of existing customers are not computerized, questionnaires sent to customers or senior sta can provide the needed information. In the case of Fulton County, although the county has address matching tools, the health centers' customer databases are not computerized. Information on the health centers' current customers was therefore extracted from questionnaires and nursing sta expertise. We thus formed a socio-demographic pro le of the main customers of the Fulton County health centers. They were identi ed as: (1) those with income below 185% of the Federal poverty level; (2) women between 15 and 44 years old (for family planning and the Women, Infants and Children Supplemental Food program (WIC)); (3) children under 5 (for childhood immunization, the Child Health Program (CHC) and dental services); (4) the elderly (for u shots); (5) and young males (for treatment of STDs). Once the socio-economic pro le of current customers has been established, the relevant census data for an entire region is examined to predict potential customers within each socio-demographic category. A desired level of geographic detail is selected according to the precision level needed by the location model: smaller zones provide more precise information on customers but will require more computation. For this case study, census data at the block group level were extracted, giving counts on each of the identi ed socio-economic categories. The census data were not broken down by zipcodes because, in Fulton County, the resulting areas do not provide the desired resolution and often overlap the county's boundaries. For this project, we found the relevant census data free of charge at a local university library, but it is now available free on Internet. A second category of data concerns the sites where health centers, satellite facilities, and mobile facilities can be located. In theory, facilities can be located anywhere, but in practice, health care administrators generally have speci c ideas of where health centers and satellite facilities should be located. For this project, we considered potential sites for the health centers to be the same as the existing health care centers, and the potential sites for the satellites to be the 40 di erent housing projects and the 110 elementary schools located within the county. These sites were geocoded using the address matching capabilities of Atlas GIS and Tiger les (free electronic road maps made available by the U.S. Census 7
Bureau). Electronic maps of transportation networks can be used for at least two other purposes: to better display solutions, and to compute network distances. Displaying the main roads helps planners to associate the sites of health centers and mobile facilities with speci c neighborhoods. For example, Figure 1 would have been much more dicult to read without the display of main highways. Sophisticated geographic information systems can use information on transportation networks to compute distances between customers and facilities, and can even calculate travel times if we have information on each road segment's average speed. For this project, we got detailed road maps from the Fulton County Urban Planning Agencies and purchased the census boundaries from a commercial provider. A ordable collection of data can be a tedious task. The cost and quality of databases can
uctuate greatly, depending on whether we purchase them from the government (usually cheaper) or from a private company (usually more expensive, but more up-to-date and with appropriate formatting.) Data collection took up 80% of the time spent on this projet, making this step a non-negligeable task!
4.2 Step 2 { Customer behavior The second step consists in modeling customers' behavior, i.e., where customers go when they decide to seek health services. Customer behavior models can be very sophisticated, such as the one presented by Parker [9], but in practice, models do not need to be extremely precise. FCHD agreed that the two main factors a ecting choice of a facility were the distance and size of the facility. Distance has an important impact on accessing health care. According to a nursing sta survey, most residents of Fulton County go to the nearest health care center, this assumption was later con rmed by similarities between the boundaries made from allocating census block groups to the nearest health care center and the boundaries drawn by nurses shown on Figure 2. Quantitative analysis performed with CAPS Logistic Toolkit on the current network of health care centers (see Table 1) shows that the population does not have to travel more than 11.27 miles. The center of census block groups and straight line distances were used to compute these distances and such computations take less than two 8
minutes to perform. Therefore, for most services, locating health centers such that the population would not have to travel more than 11.3 miles guarantees the same access as the existing network. However, it is believed that the poor and the elderly cannot travel very long distances, three to ve miles being the maximum distance they can travel. It should also be noted in Table 1 that the poor and the elderly have to travel, on average less than the entire population | 1.44 and 1.79 instead of 2.04 miles. If mobile services can deliver services targeting this less mobile population, than the health centers can still be located within 11.3 miles. The size of a facility is also important, because smaller ones o er a limited number of services. The current network has a few smaller health care centers, but the nursing sta of FCHD strongly supports the design of a network in which every health care center o ers the same spectrum of services. There is a fear at FCHD that less health care centers, even if complemented by mobile facilities, can lead to travel diculties in accessing health care services. For certain services, the target population is mobile and motivated enough to travel long distances. In modeling customer behavior, we need to take into consideration both distance and size. We may assume that residents will go to the nearest health facility but we cannot assume they will do so if it is a smaller satellite facility o ering a limited number of services.
[Locate table 1 about here]
4.3 Step 3 { Costs/Minimum number of customers In step 3, we have to decide what is an appropriate size for the xed facilities. The methodology suggests that we perform an economic analysis based on the cost of serving a given population size with xed facilities versus mobile satellite facilities. For this project, we did not have access to actual dollars values and we therefore had to work with a network proposal from the Health Commissionner in order to de ne the minimum number of customers to justify a xed facility. As previously mentioned, 20 health centers are too many and less health centers combined with mobile facilities will allow FCHD to provide better quality and better access to its services. The Health Commissioner and the Executive Director for Physical Health 9
Centers thought that eight to eleven health centers, combined with three mobile facilities, could be as e ective as the existing 20 health centers. These health centers should draw enough customers to support a large full time multi-disciplinary team but each health center should not be too big to keep the sense of community health care. The Commissioner gave us a proposal of eight health centers as a place to start. A quantitative analysis with CAPS logistics { with the population allocated to the nearest health center { shows the smallest health center would attract 35,000 residents; the largest would attract 150,000 residents, 76,000 of whom are poor. Following, a discussion with the Health Commissionner and the nursing sta , we agreed that under the new system, health care centers would have to serve at least 35,000 but no more than 100,000 residents with no more than 40,000 poor residents.
4.4 Step 4 { Locational decisions In step 4, we locate health centers rst and then we locate the satellite facilities and the base location of the mobile facilities.
4.4.1 Health Centers The methodology was designed under the assumption that locational models were needed to rst generate a basic proposal that health experts could later modify to account for factors not in the model. With FCHD, we found out that the opposite was needed: the basic proposals were generated by experts and the model helped to improve them by using quantitative criteria. We started with two proposals | one from the Health Commissionner and one from the nursing sta | and helped FCHD to design a health care network that met with everyone's agreement, including that of the population. We rst worked on the Commissionner's proposal which consisted of eight health centers. We ran a simple improvement algorithm on the proposal | closing one center at a time and opening one of the 12 health center sites not originally selected | simply aiming at improving access for customers and insuring that none had to travel more than 11.3 miles. The algorithm was able to improve the proposal. Based on this, we look at this proposal 10
plus one health center at the proposed site by the algorithm and the algorithm could not improve this new network of nine health centers. The Health Commissionner aknowledge this new network was better. We then worked with nursing. Their proposal consisted of eleven health centers. It was the result of a \merging" analysis: each new health center consisted of a merger of two or more existing health centers. It was easiest for them to perform their analysis this way, starting with a known clientele and merging small customer bases. We quantitatively analyzed their proposal using the model, then re-ran a new version of our algorithm, integrating the size constraints found in step 3 | between 35,000 and 100,000 residents, with no more than 40,000 poor residents. We met after each step to discuss the latest analysis and models outcome. New health centers were added or withdrawn based on the nurses' knowledge of the area and taking into consideration the Health Commissionner's proposal. When there was uncertainty as to wether a facility should be kept or open, the quantitative analysis was always used to make the nal decision. The nal proposal had ten health centers. Their locations and market share boundaries are shown in Figure 4 and demographic data are shown in Table 2. We can see that the poor population is quite well distributed between the four health centers within the county's central area, an unexpected but welcome result.
[Locate gure 4 about here] [Locate table 2 about here]
This interactive location process lead to the consensus that FCHD could well serve its population with half the number of existing health care facilities, a consensus which was important in order to implement changes. The decision-making process was made easier by the knowledge that any gaps in the health center network could be lled by satellite facilities.
4.4.2 Satellite facilities Fulton County's existing system of health care delivery did not include satellite facilities. In order to help FCHD develop expertise on how to locate and use mobile facilities, we therefore started working on a small problem, the delivery of childhood vaccinations. It 11
was believed that participation in the vaccination program is highly sensitive to travel distance and the existing network of 20 health centers did not provide satisfactory access. With a network of 10 health centers, satellite facilities would be essential. The county's 110 elementary schools and housing projects had been identi ed as potential sites for these satellites, but obviously, a satellite facility could not be located at every one of these sites. We assumed that traveling 11.3 miles would be less of a barrier for the parents of rich children, and therefore concentrated on locating the satellites so that we could improve access for poorer children. Assuming that these would be willing to travel up to 3 miles to seek immunization for their children at a health center, we surmized that they would be willing to travel up to 2 miles for service at a satellite facility. Where the density of poor children is especially high, we can have more satellites to provide better access as long as it does not raise costs. To achieve these objectives, we developed a location model and implemented it on CAPS logistics. We considered all 150 satellites to begin with. We allocated the population of poor children to the nearest health center or satellite and computed the cost to serve these satellites. Although some satellites would be permanent facilities, the ones considered for the immunization program were to be served by mobile units. The total cost to serve the satellites is not based on the number of satellites, but rather on the number of visits that each mobile unit must make. Assuming each satellite has to be visited at least once a month, we estimated that a population of 200 children under 5 needs one half-day visit per month (based on the number of immunizations required throughout childhood and the rate at which nursing sta can dispense vaccinations). The lowest cost network generated by the model is shown in Figure 5 and requires 21 satellites and 32 half-days
[Locate gure 5 about here]
4.4.3 Mobile facilities The last set of locational decisions involves determining the base location of mobile facilities by minimizing travel distance. The Fulton County Health Commissionner had already decided to base the mobile facilities at a secure lot in the central area of the county. It is obvious from Figure 5 that to meet childhood immunization needs, the base should be 12
located in the south-central part of the county. However, because the Health Commissionner's decision was rm, we did no further analysis to pinpoint an exact location.
4.5 Step 5 { Scheduling According to the methodology, the last step requires making schedules for services, sta , and mobile facilities. This implies that all questions related to scheduling are not taken into consideration during the locational decisions. In this case, FCHD simply decided to schedule the mobile units such that each satellite facility would be visited on a periodic basis.
5 Discussion Since this planning exercise which occurred in 1994, where is FCHD several years after? What can we learn from delivering health care services through a mix of xed health centers, satellite facilities, and mobile facilities? What did we learn from using geographic information systems to design new health care service networks?
5.1 Present status of FCHD facilities As a result of merger and consolidation, there are currently 18 health centers in Fulton County, down from the original 20 when the study was done. The plan to reduce the number of facilities to ten has been slightly changed. As of 1998, the proposal called for 11 Regional Health Centers (RHCs) and two smaller Neighborhood Health Centers (NHCs). At the same time, the planning process has been reopened with ve options now being considered: a nine-center plan, a ten-center plan, and three twelve-center plans. Note that all of these options gravitate around the ten centers suggested by our analysis. Deviations from the network proposed by the original study are the result of continuing changes in the population and services of Fulton County. A new Health Commissioner is now in place and these issues are being reexamined in light of community interests. In addition, the political process also plays a role and neighborhood pressure has been a factor 13
in the decision to open or close facilities. In 1996, FCHD started using two mobile clinics | one general medical used by the nursing sta for child health, immunizations, pregnancy tests, etc., and the other used for dental services. It has been problematic for the vehicles to cover the entire county since it is 75 miles long and the vehicles are very large. Progress with the mobile clinics to supplement the xed facilities includes the purchase of three other mobile facilities for a total of ve: two dedicated to dental services, and three for general medical use. These will be scheduled to cover the three areas of the county: the northern, central, and southern region. A wide variety of services are planned for these facilities.
5.2 Value of the planning exercice Utilization of the latest GIS technology was very helpful for designing the new health care delivery network for FCHD. Digitized census data and maps provide helpful feedback in comparing di erent proposals. We found out that basic GIS software, such as Atlas/GIS, are very good for displaying information graphically. However, GIS with advanced tools and programming capabilities, such as CAPS Logistics Toolkit or Arc/Info, were essential for the quantitative analysis of a new network, a task which is dicult to perform on basic GIS software. In addition to sophisticated analysis, CAPS Logistics enabled us to run location models. We discovered that models generating new networks were not very useful to start with as FCHD already had a basic proposal. Models to improve an existing network were more useful. Because it was dicult for the health administrators to formulate their basic proposal based on cost analysis, we used the GIS to extract the needed input parameters for our location models. Our analysis ended up being very useful for FCHD, but required a lot of data and complex software. The data manipulation was, surprisingly, more dicult to handle than expected because of di erent formats which were not always compatible with the GIS software. FCHD was smart enough to buy the same software used by other Atlanta agencies in order to share databases and to team up with researchers from the Georgia Institute of Technology to quickly access more sophisticated GIS software. The technology is there and available but it is the expertise which is dicult to develop. Even if an organization 14
is able to develop the data expertise, in-depth analysis and the use of more advanced GIS software might still be beyond them. Planning a health care system to deliver multiple services to heterogenous population through a mix of xed health centers, satellite facilities, and mobile facilities is not an easy task. As the methodology suggests, in designing the delivery network, it was easier to locate the health centers rst even if administrators had decided that mobile programs would be later added. Knowing that mobile facilities were part of the new delivery network, however, helped administrators to make a decision on the nal proposal and removed some weight from the decisions regarding health centers, since a bad facility decision could be overcome by the use of mobile facilities. Through the planning exercice, FCHD was able to understand that they could adequately but cheaply serve less populated areas with fewer health centers using more satellite facilities frequently visited by a mobile facility. Nevertheless, since implementation of a network takes place over a long period, it was predictable that it would be re-assessed periodically. We were not as resourceful in helping FCHD to plan the use of mobile facilities. When we worked with FCHD, they had no prior experience with satellites and mobile programs. One diculty is that administrators' tendency is generally to think in terms of health centers supplemented by mobile facilities to ll any gaps in the system; only then do they consider where satellites should be located to service the mobile facilities. This intuitive approach involves more trial and error than the models' quantitative results. Perhaps a mistake with our analysis was to present the location of the satellites rst and then the location of the mobile facilities. Using the same models but showing information di erently might make it easier for health decision makers to accept the results of models and GIS analysis. Future research could focus on di erent presentation of outputs.
Conclusion A ve step methodology was used to develop a new health care delivery network for FCHD. The early stages of the analysis were easier to carry out than later ones when the realities of the actual implementation have to be taken into account. The method is data intensive 15
and its quality depends on the sophistication of the GIS software used. However, it provides a solid basis for planning, allows incorporation of expertise from many quarters, and its quantitative results give planners the tool they may need for the justi cation of a network proposal.
References [1] H. J. Anderson (1992). \Health Care for the Homeless: What Role Should Hospitals Play?" Hospitals & Health Networks, July 20, 44{48. [2] B. A. Bewerse and K. J. Williams (1991). \Mobile Mammography Units: An Opportunity for Public Health Sector Capacity Building Through Local Linkages," APHA 119th Annual Meeting { Proceedings, Abstract #4007, 342. [3] J. M. Destouet and R. G. Evens (1992). \Mammography Outreach Program," American Journal of Public Health 82, 302{303. [4] M. J. Hodgson, G. Laporte, and F. Semet (1996). Covering Tour Model for Planning Mobile Health Care Facilities in Suhum District, Ghana," Publication 96-CRT-01, Centre for Research on Transportation, Montreal, Canada. [5] S. D. Lapierre, H. D. Ratli , and D. Goldsman (1998). \Models for the Delivery of Preventive Health Services and Application to Fulton County," Publication #98-64 , Center for Research on Transportation, Montreal, Canada. [6] Metropolitan Hospitals Planning Bord (1995). Taking Melbourne's Health Care Network into the 21st Century: Phase 2 Report, Working Paper, Metropolitan Hospitals Planning Board, Victoria, Australia. [7] Metropolitan Toronto District Health Council (1995). Directions for change: Toward a Coordinated Hospital System for Metro Toronto, Working Paper, Metropolitan Toronto District Health Council Hospital Restructuring Project, Toronto, Canada. [8] B. R. Parker (1990). \In Quest of Useful Health Care Decision Models for Developing Countries," European Journal of Operational Research 49, 279{288. [9] B. R. Parker and V. Srinivasan (1976). \A Consumer Preference Approach to the Planning of Rural Primary Health-Care Facilities," Operations Research 24, 991{1025. [10] S. Rajagopalan (1993). \Allocating and Scheduling Mobile Diagnostic Imaging Equipment Among Hospitals," Production and Operations Management Journal 2, 164{176. [11] P. N. Reeves, D.F. Bergwall, and N.B. Woodside (1984). Introduction to Health Planning, 3rd Edition, Information Resources Press, Arlington, Virginia. [12] E. Rubin, M. S. Frank, R. J. Stanley, W. K. Bernreuter, and S. Y. Han (1990). \PatientInitiated Mobile Mammography: Analysis of the Patients and the Problems," Southern Medical Journal 83, 178{183.
16
[13] G. P. Schultz (1970). \The Logic of Health Care Facility Planning," Socio-Economic Planning Science 4, 383{393. [14] E. A. Sickles, W. N. Weber, H. B. Galvin, S. H. Ominsky, and R. A. Sollitto (1987). \Low-Cost Mammography Screening: Practical Considerations with Emphasis on Mobile Operation," Cancer 60, 1688-1691. [15] U.S. Government (1990) \1990 Census Data", Washington D.C..
17
Fulton County: Demographics of the existing network Max Pop. Poor W15-44 60 M20-24
Health Center
(miles) 7.94 5.48 7.41 4.22 3.90 1.28 2.63 2.05 2.47 2.30 2.18 2.54 2.25 5.83 4.04 1.70 2.70 2.14 5.15 11.27
Alpharetta 54,903 3,889 15,513 4,947 3,097 Roswell 53,560 5,091 14,557 3,492 4,787 Sandy Springs 57,785 5,956 15,292 2,774 7,420 Buckhead 54,705 7,667 13,053 2,392 12,172 Rockdale 36,979 14,974 9,004 2,713 4,601 N'Hood Union 18,386 11,751 4,984 1,702 2,783 Techwood 16,851 7,334 3,976 1,055 1,082 Aldredge 19,130 11,853 4,262 1,572 2,971 Northeast 35,539 9,979 10,025 1,521 4,514 Roy McGee 40,419 19,315 10,694 3,027 6,100 Center Hill 14,542 8,485 3,610 1,547 1,803 Lakewood 22,736 11,750 5,630 2,268 1,796 South Fulton 20,899 14,138 4,991 2,210 2,608 Adamsville 39,383 14,269 10,043 3,269 4,351 Brooks 41,641 15,118 10,791 3,502 5,310 Jere Wells 16,433 7,361 4,324 1,343 2,574 Hapeville 14,819 6,523 3,729 1,546 1,563 College Park 15,453 5,431 4,222 1,221 2,221 Red Oak 52,698 13,828 15,901 4,383 3,473 Fairburn 22,090 4,976 5,530 1,733 2,845 Total 648,951 199,688 170,151 48,217 78,071 Avg distance (miles) 2.04 1.44 2.07 1.98 1.79 Table 1: Demographics of the existing network { from North to South
1,381 1,937 2,191 1,721 2,169 700 2,923 747 1,725 1,593 465 760 853 1,398 1,603 533 457 491 2,149 849 26,645 1.81
Fulton County: Demographics of the proposed network Health Center Max Pop. Poor W15-44 60 M20-24 (miles) Alpharetta 7.9 49,178 3,345 13,279 4,451 3,227 North Fulton 5.7 92,509 8,739 26,675 5,564 7,789 Buckhead 5.8 74,390 9,759 17,235 3,300 15,453 Rockdale/Ctr Hill 4.2 74,381 32,372 18,434 5,905 9,923 Aldredge 3.7 73,603 33,227 18,485 4,695 9,381 N'Hood Union/McGee 3.2 67,913 34,360 18,241 5,560 9,283 Lakewood 3.6 57,523 30,553 14,855 5,989 5,636 Adamsville 5.3 56,932 17,223 14,896 4,341 5,685 College Park 6.7 67,930 23,139 18,889 5,744 7,597 Fairburn 11.3 34,592 6,971 9,162 2,668 4,097 Total 648,951 199,688 170,151 48,217 78,071 Avg distance (miles) 2.47 1.96 2.47 2.42 2.33 Table 2: Demographics of the proposed network { from North to South 18
1,208 3,452 2,436 3,872 4,914 2,855 1,978 2,147 2,537 1,246 26,645 2.27
Figure 1: Poorest census tracts in Fulton County | number of residents below the 185% federal poverty level
19
Figure 2: Location of Fulton County Health Department's 20 health centers
20
Step 1
Collect Data
Step 2
Model Customer Behaviour
Step 3
Determine Costs of Fixed Facilities and Satellite Programs
Step 4 Locate Fixed Facilities Locate Satellites Locate Base of Satellites
Step 5
Schedule Satellite Services and Visits of Mobile Facilities
Figure 3: Methodology for designing a network with xed and satellite facilities
21
Figure 4: The proposed network for the Fulton County Health Department
22
Figure 5: Satellite locations for childhood immunization showing number of half-day visits per month.
23
