Development of a prediction model for deep vein thrombosis in a retrospective cohort of patients with suspected deep vein thrombosis in primary care

OBJECTIVES: Early accurate prediction and diagnosis of deep-vein thrombosis (DVT) is essential to allow immediate treatment and reduce potential complications. However, all potential strong risk factors are not included in pretest probability assessments such as the Wells score. In addition, the Wells score may not be suitable in primary care since it was developed for secondary care. We hypothesized that by adding more risk factors for DVT, existing diagnostic approaches may improve DVT prediction.

METHODS: All consecutive patients suspected of DVT (2004-2016) in a primary care setting were included in this retrospective study. All patients received a Wells score, D-dimer, and duplex. Available recorded data of patients were used to develop a model to predict DVT.

RESULTS: Among 3381 eligible patients, 489 (14.5%) had confirmed DVT. The developed model based on D-dimer, Wells score, gender, anticoagulation use, age, and family history of venous thrombosis (VTE) was able to distinguish DVT patients from all suspected DVTs with a sensitivity of 82% (confidence interval [CI]: 78-86) and a specificity of 82% (CI: 80-83).

CONCLUSIONS: The proposed model is able to predict the presence of DVT among all suspected DVT patients in a primary care setting with reasonable accuracy. Further validation in prospective studies is required.