Differences in mortality from acute myocardial infarction between coronary care unit and medical ward: treatment or bias?

To analyse the effectiveness of coronary care units in reducing mortality from myocardial infarction 18 hospitals ranging from large urban teaching hospitals to small country hospitals were stratified into four levels of care. Previous analysis had failed to show significant differences in the overall mortality in hospital among levels. There were significant differences in mortality, however, between those patients allocated to be cared for in the coronary care unit and those in the medical wards in the more advanced hospitals. The differences were largest in the hospitals with the most elaborate facilities (level 1) and non-existent in those with the least (level 4). Several analytical approaches to these observed differences indicated that they were: (a) reduced by adjustment for age and severity of infarction; (b) paralleled by differences in coexisting disease recorded on death certificates; (c) no longer significant at level 1 after allowing for differences in coexisting disease; and (d) not significant at any level after exclusion of patients first diagnosed at necropsy. These findings suggest that the observed differences in mortality between coronary care units and medical wards are largely due to bias in selection and diagnosis.