Comparison of four clinical prediction scores for the diagnosis of lower limb deep venous thrombosis in outpatients

PURPOSE: We compared three scores for the prediction of deep venous thrombosis with a new score designed specifically for outpatients. METHODS: Patients referred for evaluation because of suspected deep venous thrombosis were examined by ultrasonography. Sensitivity and specificity were calculated for three clinical scores (Wells [nine components], Kahn [four components], and St. André [six components]). We developed a new score by multivariate analysis, and then compared this score with the others in a new sample. RESULTS: Four hundred and forty-four outpatients were included in the first sample, of whom 126 (28%) had deep venous thrombosis. The Wells score was a better predictor of deep venous thrombosis than the Kahn and St. André scores. According to the Wells score, 73 patients had a high probability of deep venous thrombosis (of whom 51 [70%] actually had a thrombosis) and 178 had a low probability of deep venous thrombosis (of whom 19 [11%] had a thrombosis). A new score was developed as follows: male sex (+1), lower limb palsy or immobilization (+1), confinement to bed >3 days (+1), lower limb enlargement (+1), unilateral lower limb pain (+1), and other plausible diagnosis (-1). In a validation sample of 282 outpatients, this score identified 31 patients who had a high probability of deep venous thrombosis (score > or =3), of whom 18 (58%) had a thrombosis, and 70 patients who had a low probability (score < or =0), of whom 3 (4%) had a thrombosis. The Wells score and this ambulatory score had similar test operating characteristics in the validation sample. CONCLUSION: Our new six-component score had similar diagnostic utility as the nine-component Wells score among outpatients being evaluated for deep venous thrombosis.