Diagnostic classification in patients with suspected deep venous thrombosis: physicians' judgement or a decision rule?

Background: Clinical decision rules can aid in referral decisions for ultrasonography in patients suspected of having deep venous thrombosis (DVT), but physicians are not always convinced of their usefulness and rely on their own judgement.

Aim: To compare the performance of a clinical decision rule with the probability of DVT presence as estimated by GPs.

Design Of Study: Cross-sectional survey.

Setting: Primary care practices in The Netherlands.

Method: GPs (n = 300) estimated the probability of the presence of DVT (range 0-100%) and calculated the score for the clinical decision rule in 1028 consecutive patients with suspected DVT. The clinical decision rule uses a threshold of three points and so, for the GP estimates, thresholds were introduced at 10% and 20%. If scores were below these estimates, it was not considered necessary to refer patients for further examination. Differences between the clinical decision rule and the GP estimates were calculated; this is discrimination (c-statistic) and classification of patients.

Results: Data of 1002 patients were eligible for analysis. DVT was observed in 136 (14%) patients. Both the clinical decision rule and GP estimates had good discriminative power (c-statistic of 0.80 and 0.82 respectively). Fewer patients were referred when using the clinical decision rule compared with a referral decision based on GP estimates: 51% versus 79% and 65% (thresholds at 10% and 20% respectively). Both strategies missed a similar and low proportion of patients who did have DVT (range 1.4-2.0%).

Conclusion: In patients suspected of DVT both GP estimates and a clinical decision rule can safely discriminate in patients with and without DVT. However, fewer patients are referred for ultrasonography when GPs rely on a clinical decision rule to guide their decision making.