Cause of death of patients on treatment for tuberculosis: a study in a rural South African hospital

SETTING: Tintswalo Hospital, a rural district hospital in South Africa. OBJECTIVES: To assess the quality of data relating to deaths from tuberculo sis; to determine the actual cause of death of patients who die whilst on t uberculosis treatment. DESIGN: A descriptive study of all patients who were treated for tuberculos is between January 1992 and June 1996, and who died during their course of treatment. A review of records, X-rays and death certificates of patients w hose cause of death was recorded as tuberculosis. Verbal autopsies were don e for patients who died at home. Deaths were classified into one of five ca tegories: definitely, likely or possibly due to tuberculosis, unlikely or n ot due to tuberculosis, based on criteria such as bacteriological confirmat ion of tuberculosis and the presence of other illnesses. RESULTS: There were 927 cases of tuberculosis and 80 deaths recorded as due to tuberculosis. Valid criteria or detailed information were available for only 68 patients. Only 56% of deaths were confirmed or likely to be due to tuberculosis, and 12% of patients probably did not have tuberculosis at al l. The remaining 32% had tuberculosis but died of other illnesses. CONCLUSIONS: The recorded tuberculosis mortality data were clearly inaccura te. There is a need to improve tuberculosis diagnosis, documentation on dea th certificates, and tracing of treatment interrupters in order to determin e whether or not they have died.