A comparison of a Bayesian vs. a frequentist method for profiling hospital performance

Journal of Evaluation in Clinical Practice - Tập 7 Số 1 - Trang 35-45 - 2001
Peter C. Austin1,2, C. David Naylor3,1,4,2, Jack V. Tu3,1,4,2
1Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada
2Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
3Department of Medicine, University of Toronto, Toronto, Ontario, Canada.
4Division of General Internal Medicine, Clinical Epidemiology Unit and Health Care Research Program, Sunnybrook and Women's College Health Sciences Centre, Toronto, Ontario, Canada

Tóm tắt

Abstract

The objective of this study was to compare the classification of hospitals as outcomes outliers using a commonly implemented frequentist statistical approach vs. an implementation of Bayesian hierarchical statistical models, using 30‐day hospital‐level mortality rates for a cohort of acute myocardial infarction patients as a test case. For the frequentist approach, a logistic regression model was constructed to predict mortality. For each hospital, a risk‐adjusted mortality rate was computed. Those hospitals whose 95% confidence interval, around the risk‐adjusted mortality rate, excludes the mean mortality rate were classified as outliers. With the Bayesian hierarchical models, three factors could vary: the profile of the typical patient (low, medium or high risk), the extent to which the mortality rate for the typical patient departed from average, and the probability that the mortality rate was indeed different by the specified amount. The agreement between the two methods was compared for different patient profiles, threshold differences from the average and probabilities. Only marginal agreement was shown between the Bayesian and frequentist approaches. In only five of the 27 comparisons was the kappa statistic at least 0.40. The remaining 22 comparisons demonstrated only marginal agreement between the two methods. Within the Bayesian framework, hospital classification clearly depended on patient profile, threshold and probability of exceeding the threshold. These inconsistencies raise questions about the validity of current methods for classifying hospital performance, and suggest a need for urgent research into which methods are most meaningful to clinicians, managers and the general public.

Từ khóa


Tài liệu tham khảo

Casella G., 1990, Statistical Inference.

10.1056/NEJM199602083340611

10.7326/0003-4819-127-8_Part_2-199710151-00065

10.1002/(sici)1097-0258(19971215)16:23<2645::aid-sim696>3.0.co;2-d

10.1016/0002-9149(88)90918-6

Fleiss J.L., 1981, Statistical Methods for Rates and Proportions

10.1097/00005650-199506000-00005

10.1007/978-1-4899-4485-6_1

10.2307/2348941

10.1148/radiology.143.1.7063747

10.1002/sim.4780141909

Iezzoni L.I., 1994, Risk Adjustment for Measuring Health Care Outcomes.

10.1016/0003-4975(94)91721-3

10.7326/0003-4819-123-10-199511150-00004

10.1097/00005650-199601000-00002

10.1177/0272989X9601600405

10.2307/2529310

Lee P.M., 1997, Bayesian Statistics: an introduction

10.1016/s0140-6736(97)09362-8

Luft H.S., 1993, Annual Report of the California Hospital Outcomes Project.

10.1136/bmj.316.7146.1701

10.1161/01.CIR.85.4.1254

New York State Department of Health, 1992, Coronary Artery Bypass Graft Surgery in New York State 1989–91.

10.1080/01621459.1997.10474036

10.1016/0895-4356(94)00126-b

Pennsylvania Health Care Cost Containment Council, 1996, Focus on Heart Attack in Pennsylvania. Research methods and results.

SAS Institute Inc., 1997, SAS/STAT Software: changes and enhancements through release 6.12.

Scottish Office, 1995, Clinical Outcome Indicators, 1994.

10.1007/978-1-4899-4485-6_2

10.1378/chest.108.1.83

10.1056/NEJM198308113090602

Tu J.V., 1999, Cardiovascular Health Services in Ontario: an ICES atlas, 83

Tu J.V., 1999, Temporal changes in the outcomes of acute myocardial infarction in Ontario, 1992–96., Canadian Medical Association Journal, 161, 1257

10.1161/01.CIR.88.2.416

Wong G., 1985, The hierarchical logistic regression model for multilevel analysis., Journal of the American Statistical Association, 80, 513, 10.1080/01621459.1985.10478148

ZinmanD.(1991)State takes docs' list to heart.NewsdayDecember 18 p.7. New York NY.