Although recent advances have been made in understanding its epidemiology, diagnosis and treatment, pulmonary embolism (PE) is still largely undetected and untreated, and the mortality rate has not appreciably changed in the last decades. The aim of this study was to: compare the postmortem frequency of massive and sub-massive PE during two different time periods in the same general hospital; ascertain whether the percentage of correct clinical diagnosis of PE has changed; identify factors which might contribute to the inaccuracy of the clinical diagnosis of PE. Altogether, 288 patients with autopsy-proven PE and adequate clinical data were collected in the first period; 182 subjects with the same characteristics were found in the second period. Cases observed from 1989 through 1994 were evaluated in terms of frequency of false negatives and false positives, predictive value of the clinical diagnosis of PE, and correlations between clinical and post-mortem diagnosis of PE on one side and several independent variables such as age, gender, associated diseases, recent surgery on the other. In our hospital the frequency of massive and submassive PE at autopsy was 8.6% from 1966 through 1974, 12.6% from 1989 through 1994 (p < 0.01). The percentage of correct clinical diagnosis of PE was 19.6% in the former period, 21.6% in the latter (NS) with 78.57% of false negatives and only 1.73% of false positives. Altogether the true positives were 21.42%, most of them being patients with massive PE. Clinical findings showed the coexistence of heart disease in 51.6% of the cases, congestive heart failure in 20.15%, metabolic disease in 7%, stroke in 12.5%, recent surgery in 12.5%. Autopsy revealed the presence of pulmonary infarction in 22% of cases, malignancy in 24.0%, pneumonia in 17.05%, acute myocardial infarction in 14.8%. Seventy percent of the cases in whom the point of origin of thromboemboli could be demonstrated had one or more thrombus in the district of inferior vena cava, more frequently at the level of the femoral and iliac veins. The positive predictive value of the clinical diagnosis of PE was 0.60, the negative predictive value 0.84. Multivariate logistic regression analysis showed that the clinical diagnosis of PE was hindered by the presence of pneumonia, facilitated by admission to the Cardiological Department. Age, duration of hospitalization, presence of pulmonary infarction, cancer, obesity, stroke, heart failure and recent surgery did not influence the clinical diagnosis of PE in this series. A positive correlation (p < 0.05) was found between autopsy rate and the percentage of correct clinical diagnosis of PE in the various hospital departments. This relationship needs further investigation, all the more so as in most countries the autopsy rate has been dramatically declining in recent times, especially in late life. In conclusion, at least in some institutions, the autopsy frequency of PE has increased during the last decades, and this increase has not been paralleled by a significant improvement in clinical diagnosis.