Ms. Lmd* Davids ... mm and treatment by radical surgery results in maximal .survival rates ... present, especially if a surgical mortality at the lower end of the.
MEDICAL DECISION MAKING
70 • ESMDM 1990 Abstracts
H-|\M-
hC\]
^70
E P l t J E M I O L O C I C A L A?® R E S O L U C E DATA REQU!II£D FOR A COSTEfFECTIVE^aESS S T L D Y IN G A S F E C I N T E S T I N A L PATIEOTS • K S A I D f are widely used in the treatment of arthritis. Problems exist wjtls «id«-effKts particularly m tM gastro-intestinal tract »g*oci«ted wiih G I erosions or peptic alccif and their complications l 0 | i ^ i and resaoKa data repaired to mdemke this £0st-«ffectiv«sies3 suid^. %5den«ologkal data
was ot>t«ioed from Scottish Morbidity Record 1 (SMRl, cofflpleted foe all disch«i|e», eteaihs aisd tfanifers of nfm-dbuttrk and iKsi-psydHalric pstknts in StaoHish hospitals) ami was saed to provids: 1.
2. 3.
4.
gastric ulcer patients discharged in Scotland. Data on speciHed GI dtagtK>se5 hi Greaier Glasgow Hedih B o » d (pqHiiasion 0.&m). A s a l u t i n g fraiae 10 csamine a r.ai4cn9 s^wpte of T78 of the 364$ G I j»ticms dijcfta^ted frwn O G H B hospitals m mi. Identify patiems %ho imi a co-di^^sis of erthriu. Data On
Csmsm review was wdert^rat 10: Vafidste the. f»ian»y diagsosi* of GI diagnosis. 2. Validate t»B CQHjiagnoJis of Mdirftts. 3. Ascertain liie ms of NSAIDs by the«* patieots (and subsequenay vtiidated by a postal survey of Uiess paiietfts' general fsaciilioners). 4. Establish the re»Mirces used hy these patients (attendance at 0 1 outpatientj, G ! admissions, mtgtif, endoscopies, barioisi ttiealt, bwiwa eraunas. wtd transfusions). A ttmctar^ interview study by telephone of 30 general practiiioftcrs agcertaiijed their management of patiems m NSAIDs with GI ^w^lom*. From this resource aiilisaaon was estunated for ^ese patietHs in comparlswi to tho«« sdmiaed to bospiial. Deiaikd costings of procedures and resoerce «*« were obtained from published data ssd imerviewiof itey perstmnef ia Mtrgery. btood lf«i»fu«ioft artd radjoloty. I > - H»praei S . Kohlt. Dr, KaiB-iara*, PP^CSSOT Michael DrunmKsod. Ms. L m d * Davids 1.
70 • ESMDM 1990 Abstracts EVALUAriON OF A STATISTICAL INFERENCE METHOD EOR MULTI-CRITERIA DECISION MAKING. PR Holtsrave University of Oklahoma Health Sciences Center Decision-mafcers i n the health-care arena often must try to satisfy several c r i t e r i a at once. A popular analytic approach to such problems i a multi-attribute value theory (MAVT) . Simply put, this . approach recommends choosing the available option with the highest sum of importance-weighted value scores across c r i t e r i a ( i . e . , attributes) , Let R be the number of discrete available options, r denote any one of the R options, r ' be an option other than r , C be the number of discrete c r i t e r i a , c denote any one of the c r i t e r i a , iCc) be the decision-maker's importance weight for c r i t e r i o n c, v(r,c) represent the value of option r on criterion c, r be the C-dimensional vector with entries v ( r , c ) , and W(r) be the sum over a l l c of ( i ( c ) ] ( v < r , c ) ] . KAVT recommends selecting the option with the largest W(r). This guidance has been limited, however, because u n t i l now there has been no s t a t i s t i c a l inference method for assessing whether the observed difference between r and r ' ( i n che^C-dimensional space dilated by the set of C importance weights) i s larger than would be expected by chance alone, We have developed an inference procedure which accomplishes this aim. The primary assumptions of the procedure are taken from c l a s s i c a l test theory in psychology and from the randomization test literature in s t a t i s t i c s .
MEDICAL DECISION MAKING
AGREEMENT BETWEEN DOCTOR AND COMPUTER-TAKEN HISTORIES. RP KniH-Jones. GP Crean. F Marlofibanks Dept. of Public Health, University of Glasgow and Southern General Hospital, Glasgow. SccHland. This study compares a questionnaire for collecting clinical history data in patients with dyspepsia to a carefully developed computer-based version (GIADYS) evolved over many years, amed at eliciting the same information. Over a two-year period 130 duplicafe histories were obtained wittiin three weeks of each otfier. The data was compared using Kappa statistics to assess the ctiance-corrected extent of agreement and McNemar's test to investigate wtiether one method or the other tended to obtain more positrve responses, since most indicants were binary. An inWal analysis showed the average kappa for 67 indicants to be 0.38, slightly greater than a value of 0.33 obtained in an pilot study in wtltch a PDP computer had been used. S7 Indicants (85%) showed agreement beteen the two methods of data collection to be significantly greater than chance (P
