The use of fixed- and random-effects models for classifying hospitals as mortality outliers: A Monte Carlo assessment

被引:70
作者
Austin, PC
Alter, DA
Tu, JV
机构
[1] Inst Clin Evaluat Sci, Toronto, ON M4N 3M5, Canada
[2] Univ Toronto, Dept Publ Hlth Sci, Toronto, ON, Canada
[3] Sunnybrook & Womens Coll, Hlth Sci Ctr, Div Gen Internal Med, Toronto, ON, Canada
[4] Sunnybrook & Womens Coll, Hlth Sci Ctr, Div Cardiol, Schulich Heart Ctr, Toronto, ON, Canada
[5] Univ Toronto, Toronto, ON, Canada
关键词
hospital report cards; provider profiling; hospital mortality; outcome assessment; risk adjustment; Monte Carlo methods; random-effects models;
D O I
10.1177/0272989X03258443
中图分类号
R19 [保健组织与事业(卫生事业管理)];
学科分类号
摘要
Background. There is an increasing movement towards the release of hospital "report-cards." However, there is a paucity of research into the abilities of the different methods to correctly classify hospitals as performance outliers. Objective. To examine the ability of risk-adjusted mortality rates computed using conventional logistic regression and random-effects logistic regression models to correctly identify hospitals that have higher than acceptable mortality. Research Design. Monte Carlo simulations, Measures. Sensitivity, specificity, and positive predictive value of a classification as a high-outlier for identifying hospitals with higher than acceptable mortality rates. Results. When the distribution of hospital specific log-odds of death was normal, random-effects models had greater specificity and positive predictive value than fixed-effects models. However, fixed-effects models had greater sensitivity than random-effects models. Conclusions. Researchers and policy makers need to carefully consider the balance between false positives and false negatives when choosing statistical models for determining which hospitals have higher than acceptable mortality in performance profiling.
引用
收藏
页码:526 / 539
页数:14
相关论文
共 43 条
[1]   Factoring socioeconomic status into cardiac performance profiling for hospitals - Does it matter? [J].
Alter, DA ;
Austin, PC ;
Naylor, CD ;
Tu, JV .
MEDICAL CARE, 2002, 40 (01) :60-67
[2]   Long term MI outcomes at hospitals with or without on-site revascularization [J].
Alter, DA ;
Naylor, CD ;
Austin, PC ;
Tu, JV .
JAMA-JOURNAL OF THE AMERICAN MEDICAL ASSOCIATION, 2001, 285 (16) :2101-2108
[3]   A comparison of Bayesian methods for profiling hospital performance [J].
Austin, PC .
MEDICAL DECISION MAKING, 2002, 22 (02) :163-172
[4]   A comparison of a Bayesian vs. a frequentist method for profiling hospital performance [J].
Austin, PC ;
Naylor, CD ;
Tu, JV .
JOURNAL OF EVALUATION IN CLINICAL PRACTICE, 2001, 7 (01) :35-45
[5]   An introduction to multilevel regression models [J].
Austin, PC ;
Goel, V ;
van Walraven, C .
CANADIAN JOURNAL OF PUBLIC HEALTH-REVUE CANADIENNE DE SANTE PUBLIQUE, 2001, 92 (02) :150-154
[6]   APPROXIMATE INFERENCE IN GENERALIZED LINEAR MIXED MODELS [J].
BRESLOW, NE ;
CLAYTON, DG .
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, 1993, 88 (421) :9-25
[7]   Medical profiling: improving standards and risk adjustments using hierarchical models [J].
Burgess, JF ;
Christiansen, CL ;
Michalak, SE ;
Morris, CN .
JOURNAL OF HEALTH ECONOMICS, 2000, 19 (03) :291-309
[8]   Benefits and hazards of reporting medical outcomes publicly [J].
Chassin, MR ;
Hannan, EL ;
DeBuono, BA .
NEW ENGLAND JOURNAL OF MEDICINE, 1996, 334 (06) :394-398
[9]  
DeLong ER, 1997, STAT MED, V16, P2645, DOI 10.1002/(SICI)1097-0258(19971215)16:23<2645::AID-SIM696>3.0.CO
[10]  
2-D