Objectives Pulmonary embolism is believed to be a common cause of death of hospital inpatients. The aims of this study were to estimate the number of deaths caused by pulmonary embolism and the potential to reduce this by the use of caval filters according to accepted indications. Design Review of autopsy reports and death notification records from 2007 and 2008. When pulmonary embolism was given as cause of death (in the autopsy report or in section 1 a-c or part 2 of the Medical Certificate of the Cause of Death), hospital records were reviewed for evidence of pre-mortem diagnosis of pulmonary embolism or deep vein thrombosis (DVT) and for evidence of accepted indications for caval filter placement. Setting Large UK teaching hospital. Participants Hospital inpatients whose deaths were attributed to pulmonary embolism. Main outcome measures Proportion of deaths adjudged at autopsy to be due to pulmonary embolism; evidence of pre-mortem diagnosis of DVT or pulmonary embolism; total number of hospital admission and deaths. Results From a total of 186,517 adult inpatient admissions there were 2583 (1.4%) adult inpatient deaths of which 696 (27%) underwent autopsy. Of those undergoing autopsy, 14 (2.0%, 95% CI 1.2–3.3%) deaths were caused by pulmonary embolism. Pulmonary embolism was recorded as a cause of death in a further 12 (0.7%) of 1773 patients who did not undergo autopsy. Of these, five had a pre-mortem diagnosis of DVT or pulmonary embolism. Conclusions The proportion of deaths caused by pulmonary embolism appears to be considerably lower than the widely published rate, and of this small number, few have a pre-mortem diagnosis of DVT or pulmonary embolism. There is little scope for further reduction of pulmonary embolism mortality through use of caval filters according to guidelines. Current policy on pulmonary embolism risk prevention appears to be based on an over-estimate of the level of risk.