To assess the meaning of hospital-associated death rates, we studied whether mortality within 30 days of hospital admission (30-day mortality) is more informative than inpatient mortality and whether detailed assessment of additional discharge diagnoses helps in understanding death rates. We examined hospitalizations for elderly Medicare patients with principal diagnoses of stroke, bacterial pneumonia, myocardial infarction, and congestive heart failure; these conditions account for 30.8% of Medicare 30-day mortality. Average hospital stays for these conditions were 99.0% longer, and inpatient mortality was 25.0% higher in New York than in California, but 30-day mortality was 1.6% higher in California. We conclude that inpatient death rates depend on length-of-stay patterns and give a biased picture of mortality. Additional diagnoses such as shock and pneumonia were strongly associated with increased mortality, but Medicare data do not reveal which patients had these conditions at the time of admission. Recorded diagnoses of chronic diseases such as hypertension, diabetes mellitus, obesity, benign prostatic hypertrophy, and osteoarthritis were commonly associated with reduced risk of death; such reduced risk is not clinically plausible. Several lines of evidence suggest that chronic disorders are underreported for patients with life-threatening disorders. We recommend great caution in using discharge diagnoses of comorbid conditions to adjust hospital death rates for clinical differences in the patient populations.