Background: In 1998, tuberculosis incidence and mortality rates in the state of Rio de Janeiro (RJ) were the highest in Brazil. However, the RJ tuberculosis database (SINAN-TBRJ) has proven unreliable. Objective: To evaluate the current tuberculosis control program by analyzing tuberculosis-attributed deaths. Methods: Descriptive studies of the SINAN-TB-RJ and tuberculosis mortality (SIM-TB-RJ) databases were carried out. Both databases were linked using the Reclink program. A study based on medical records was performed in the five hospitals where the greatest numbers of tuberculosis deaths occurred. Results: In the SINAN-TB-RJ database, 16,567 cases were registered in adults (> 14 years of age). Pulmonary disease was present in 13,989 (84.5%), of whom 8223 (56.8%) presented sputum smears that were positive for acid fast bacilli. Anti-HIV testing, recommended for all patients with tuberculosis, was performed in only 4141 (25%) of tuberculosis cases. The SIM-TB-RJ database showed 1146 deaths that were attributed to tuberculosis. Only 478 (41.7%) of those had been reported to the health care system (SINAN-TB-RJ). Among the 302 medical records analyzed, 154 (50.9%)recorded hospitalizations of up to 10 days and 143 (47.3%) had respiratory symptoms for more than 60 days before diagnosis. Among 125 cases of retreatment, the RHZE regimen recommended by the Brazilian Ministry of Health was prescribed for only 43 (34.4%). Conclusion: This study demonstrates weakness in the RJ tuberculosis control program, characterized by delayed diagnosis, limited use of the recommended tests, poor reporting, and non-compliance with the Ministry of Health guidelines.

Keywords: Tuberculosis. Epidemiology. Program Evaluation