Usefulness of clinical prediction rules for the diagnosis of venous thromboembolism: a systematic review.

PURPOSE

To summarize the evidence on the predictive value of clinical prediction rules for the diagnosis of venous thromboembolism.

METHODS

We selected all studies in the English literature in which a clinical prediction rule was prospectively validated against a reference standard, and calculated likelihood ratios, predictive values, and the area under the receiver operating characteristic (ROC) curve for each prediction rule.

RESULTS

Twenty-three studies met our eligibility criteria: 17 evaluated prediction rules for the diagnosis of deep venous thrombosis and six evaluated rules for pulmonary embolism. The most frequently evaluated prediction rule for deep vein thrombosis was the Wells rule, which had median positive likelihood ratios of 6.62 for patients with a high pretest probability, 1 for moderate pretest probability, and 0.22 for low pretest probability. The median area under the ROC curve was 0.82. Addition of the D-dimer test to the prediction rule increased the median area under the curve to 0.90. The Wells prediction rule was the most commonly studied for pulmonary embolus and had median positive likelihood ratios of 6.75 for those with high pretest probability, 1.82 for moderate pretest probability, and 0.13 for low pretest probability. The median area under the ROC curve was 0.82.

CONCLUSION

The Wells prediction rule is useful in identifying patients at low risk of being diagnosed with venous thromboembolism. The addition of a rapid latex D-dimer assay improved the overall performance of the prediction rule.