From January to May 1996, 156 inpatients died in a clinic for elderly people in Rio de Janeiro, Brazil. The highest mortality rate was observed in May: 143/1,000 inpatients. As a result, the clinic was closed by the Ministry of Health. This study investigated whether the excessive number of deaths observed in the clinic in early 1996 was unexpected or reflected prevailing conditions. The investigation used the Public Health System database (SIH-SUS). The study period was 01/1993 to 05/1996. The investigation was based on: 1) a time-series analysis of the number of deaths and crude mortality rates and 2) comparison of the mortality rates observed in that clinic with those calculated for 15 area hospitals, defined as the reference rates. Risk of death in the clinic was higher than expected in 28 of the 41 months considered in the study. Highest risks were observed in January 1993 (RRcrude = 2.23; 95% CI 1.56-3.14) and May 1996 (RRadjusted = 2.73; 95% CI 1.88-3.95). The high mortality rates observed in the clinic in 1996 were already present in 1993. Thus, adequate use of the SIH-SUS could have anticipated and avoided the excess mortality identified in early 1996.