Diagnosing pulmonary embolism: a comparison of clinical probability scores.

Pulmonary embolism (PE) is a major cause of community and in-hospital mortality. This study aimed to compare the performance of the British Thoracic Society (BTS) score to the Wells' score in diagnosing PE. Data from two separate prospective diagnostic PE studies were analysed. All patients underwent gold standard investigation to determine the presence or absence of PE, together with a 3-month follow-up. The posttest prevalence of PE was compared using both scores and the receiver operating characteristic (ROC) curves. Seven hundred and seventy-nine patients were consented and investigated for PE. In patients with pleuritic chest pain, respiratory rate <20 breaths/min and absence of dyspnoea, 4·0% [95% confidence interval (CI) 1·9-7·9%] had PE. The BTS score allocated 463/779 patients as low probability, compared to 565/779 according to the Wells' score. Both scores identified a low risk group in the Manchester Investigation of Pulmonary Embolism Diagnosis cohort, however the BTS low probability group in the Thromboembolism Assessment and Diagnosis study had a prevalence of 9·7% (95% CI 5·8-15·9%). For the BTS score, the areas under the ROC curves were 0·67 (95% CI 0·61-0·72) and 0·71 (95% CI 0·61-0·75). For the Wells' score these were 0·76 (95%CI 0·71-0·81) and 0·68 (95%CI 0·64-0·73). Given the lack of BTS validation studies to date, the Wells' score appears to be the safer assessment option.