Positive Predictive Value of the Wells Score in Diagnosis of Deep Vein Thrombosis: Results of Bayesian Analysis

Objective. To evaluate the positive predictive value of the Wells score for diagnosis of deep vein thrombosis (DVT) depending on prevalence in population. Material and methods. We retrospectively analyzed the prospectively collected data of all patients who admitted to the Clinical Hospital No. 1 of the Presidential Executive Office with suspected DVT in 2012—2018. Upon admission, we assessed the risk of DVT using Wells score with subsequent Doppler ultrasound. In 2013—2014, an educational program was launched to increase the awareness of physicians about the methods of diagnosis and treatment of venous thromboembolism (VTE). We analyzed positive and negative post-test probability with 89% credible interval for Wells score ≥2 regarding the risk of DVT depending on prevalence using Bayesian analysis. Results. There were 867 patients with suspected DVT (425 (49%) men and 442 (51%) women) aged 19—101 years (mean 62.4±16.3). Of these, 467 (53.9%) ones had Wells score ≥2. DVT was confirmed in 462 (53.3%) cases. Sensitivity and specificity of the Wells score in the entire sample were 92.9% and 90.6%, respectively. In 2012—2013 (before introduction of educational program), the prevalence of DVT was 35.3%, in 2014—2018 — 64.1%. Positive predictive value increased from 9.05% (89% credible interval 7.37—11.24) to 99.47% (89% credible interval 99.34—99.59), negative predictive value — from 0.09% (89% credible interval 0.07—0.11) to 63.03% (89% credible interval 56.84—68.25) with DVT prevalence increase from 1% to 95%. Conclusion. Predictive value of the Wells score depends on population. In case of DVT prevalence > 30%, positive post-test probability of the Wells score exceeds 80%. This justifies anticoagulation before establishing the diagnosis.