Evaluation of hospitals regarding outcomes and baseline risk -- an application to percutaneous coronary intervention

Monitoring of hospital performance is increasingly used in the process of improving the quality of health care. Various approaches to quality assessment and publication of the results are discussed. Therefore, a method acceptable for clinicians in the hospitals is urgently needed. As the case mix of the patients differs considerably across hospitals, any assessment based on outcome measures has to be adjusted for the patients' risk profiles. We use a method of calculation and presentation that relates the observed complication rate of one centre to the mean rate that would be expected for similar patients in the other centres. We argue that this procedure might especially meet the demands of clinicians in the centres. Data from 44 hospitals in the German Cypher Stent Registry are used to exemplify the proposed procedure.

A large variation in the cumulative incidence of MACE within 6 months (range 0-12.2%) as well as in the prevalence of selected risk factors was observed across hospitals. Using a logistic prediction model, the differences in risk profiles of the patient populations in the centres resulted in a wide range of expected MACE rates (2.55-5.44%). The outcome of seven centres differed significantly from the expectation.

In comparisons of the outcome quality of hospitals, adjustment for the patients' baseline risk by an appropriate method is indispensable.