PURPOSE: Two prediction rules for pulmonary embolism have been described recently: the Wells' rule, which was derived from both outpatients and inpatients, and which includes a subjective element; and the Geneva rule, which is entirely standardized and is suitable only for emergency department patients. We compared the predictive accuracy and the concordance of the two methods, as well as the Geneva score overridden by implicit clinical judgment. SUBJECTS AND METHODS: We studied 277 consecutive patients admitted to the emergency departments of three teaching hospitals. Clinical probability was assessed prospectively with the Geneva score and the Geneva score overridden by implicit clinical judgment in case of a disagreement. The Wells' score was calculated retrospectively. RESULTS: The three methods classified similar proportions of patients as having a low (53% to 58% of patients), intermediate (37% to 41% of patients), or high (4% to 10% of patients) probability of pulmonary embolism. The actual frequencies of pulmonary embolism in each category were also similar (5% to 13% in the low, 38% to 40% in the intermediate, and 67% to 91% in the high clinical probability categories). Receiver operating characteristic curve analysis showed no difference between the two prediction rules, but the Geneva score overridden by implicit evaluation had a marginally higher accuracy. Concordance between the two prediction rules was fair (kappa coefficient = 0.43). Clinicians disagreed with the Geneva score in 21% of patients (n = 57). CONCLUSIONS: The two prediction rules had a similar predictive accuracy for pulmonary embolism among emergency department patients. The Geneva rule appears to be more accurate when combined with clinical judgment, although it does not apply to inpatients.