Premium
Hierarchical modelling of small area and hospital variation in short‐term prognosis after acute myocardial infarction A longitudinal study of 35‐ to 74‐year‐old men in Denmark between 1978 and 1997
Author(s) -
Rasmussen Søren
Publication year - 2004
Publication title -
statistics in medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.996
H-Index - 183
eISSN - 1097-0258
pISSN - 0277-6715
DOI - 10.1002/sim.1843
Subject(s) - case fatality rate , deviance information criterion , random effects model , statistics , medicine , logistic regression , myocardial infarction , multilevel model , bayesian hierarchical modeling , bayesian probability , demography , emergency medicine , bayesian inference , mathematics , epidemiology , meta analysis , sociology
Models for analysis of trends in hospital and small area variation in case fatality after acute myocardial infarction are presented. The data are from administrative registries in Denmark. Hierarchical modelling in a logistic regression with a Bayesian approach is used. Model selection is undertaken using the deviance and the Bayesian information criteria. There is a modest trend for hospital variation in case‐fatality rates that coincides with the introduction of new treatment strategies. This hospital variation is considerably larger than the variation at the area level. There is no trend for variation of the case‐fatality rates at the area level. Unstructured random effects slightly outperform spatially correlated random effects at the area level. Somewhat high correlations over time within hospitals and within areas were detected for the case‐fatality rates. Heavy‐tailed distributions (T‐distributions) could be an alternative for the random effect distribution in data from administrative registries and compete in the model selection with the normal distribution in this study. Copyright 2004 John Wiley & Sons, Ltd.