OBJECTIVE--To determine the incidence of venous thromboembolism in all necropsy reports over 30 years. DESIGN--Study of all necropsies in one hospital in 1987 and longitudinal analysis of results of necropsy in 1957, 1964, 1975, and 1987. SETTING--Departments of general surgery, infectious diseases, internal medicine, oncology, and orthopaedics in a Swedish general hospital. MAIN OUTCOME MEASURE--Number of cases of venous thromboembolism. RESULTS--About a third of all necropsies showed venous thromboembolism. In 1987, 260 of 347 necropsies showing venous thromboembolism found pulmonary embolism, of which 93 were classified as fatal, 90 as contributory, and 77 as incidental. Only 21 contributory or fatal postoperative pulmonary emboli were seen. In only 106 of 260 patients with pulmonary embolism did routine necropsy not show the source of embolism. Calf veins were not routinely examined. Deep venous thrombosis was seen in 239 patients in 1987. The incidence of venous thromboembolism in the four years studied was remarkably stable except in the department of orthopaedics, where the proportion fell from 60.7% in 1975 to 32.2% in 1987 (p less than 0.05), although there were only a few patients. CONCLUSIONS--The overall incidence of venous thromboembolism has not changed over 30 years. During this period the proportion of the population aged over 65 has doubled, and this may have masked the beneficial effects of prophylaxis and early mobilisation. Prevalences corrected for age are needed.