Autopsy confirmed deaths due to miliary tuberculosis in Finland were analysed in order to improve the diagnosis of the disease. Tuberculosis deaths from mortality statistics were examined in order to identify miliary tuberculosis deaths, and the medical records of the autopsied cases were studied. The deceased were divided into 2 groups, 'overt' disease and 'cryptic' disease, on the basis of chest X-ray findings. There were 114 overt (mean age 79 y) and 140 cryptic (mean age 78 y) miliary tuberculosis cases. The majority of patients in both groups were females. There was no difference between the groups in history of previous tuberculosis, in predisposing factors or in symptoms. Suspicion of tuberculosis was recorded before death in 86% in overt form and in 53% in cryptic form. In overt disease 50% of the patients received chemotherapy, but in cryptic form only a quarter were treated. In one third of cases autopsy had been carried out without suspicion of tuberculosis. Suspicion of tuberculosis had arisen too seldom, especially in the cryptic group. On the other hand, those suspected to have tuberculosis were not promptly treated with the appropriate chemotherapy. Absence of suspicion and delayed diagnosis mean increased risk in health care and at autopsy.