the objectives of this study were to analyze the morbidity and mortality profile in elderly patients hospitalized in two teaching and two non-teaching hospitals in rio de janeiro, brazil (municipal planning area 2.2) in 1999, and to compare in-hospital mortality rates adjusted for differences in profile. data were obtained from the national hospital database of the unified national health system (sih/sus). the logistic model included the variables age and primary diagnosis to calculate risk-adjusted in-hospital mortality rates. hospital admissions of elderly patients (n = 7,584) represented 29.3% of a total of 25,928 hospitalizations that took place in these units. senile cataract (7.8%) was the most frequent cause of hospitalization, followed by prostate hyperplasia (4.7%), congestive heart failure (2.9%), and complete atrioventricular block (2.8%). non-teaching hospitals presented in-hospital mortality rates higher than teaching hospitals even after adjusting for case profile differences with regard to age and primary diagnosis. the use of sih/sus databases and the risk adjustment methodology represent an alternative for exploratory analysis of healthcare outcomes.