Hospital death rates vary markedly, even for the same disease. The authors studied a representative sample of 1,126 congestive heart failure patients and 1,150 acute myocardial infarction patients in hospitals with unexpectedly high disease-specific death rates (targeted hospitals) vs. all other (untargeted) hospitals in California, Illinois, Minnesota, and New York, using both inpatient deaths and deaths within 30 days of admission. Death rates in targeted hospitals were 5.0 to 10.9 higher per 100 admissions than in untargeted hospitals. However, 56 to 82 percent of the excess could result from random binomial variation, even if all hospitals provided the same quality of care to the same age/sex/race mix of patients. The authors measured severity of illness and quality of care using detailed medical records abstracts; at the individual patient level, higher severity and lower quality were both associated with higher probability of death. However, they found only small and insignificant differences in quality between targeted and untargeted hospitals; even at a 95 percent confidence bound on the estimated difference in quality, quality differences could explain only 0.3 or fewer of the excess deaths per 100 admissions in targeted hospitals. Severity differences were also small for hospitals treating congestive heart failure patients. For myocardial infarction patients, however, severity differences explained up to 2.8 excess deaths per 100 admissions in targeted hospitals. There is some evidence that targeting hospitals with consistently high death rates over periods longer than one year may better identify potential quality problems.