Prediction of significant bleeding during vitamin K antagonist treatment for venous thromboembolism in outpatients.

Bleeding is the most concerning complication associated with anticoagulant therapy but poorly characterized and important for risk/benefit assessment. We developed a risk stratification score to predict vitamin K antagonist (VKA)-associated bleeding in venous thromboembolism (VTE) using the UK Clinical Practice Research Datalink. Significant bleeding events in outpatients consisted of major bleeding and clinically relevant non-major bleeding requiring hospitalisation (CRNMB-H) within 90 days of VKA initiation. A scoring scheme for predicting bleeding was developed from subhazard ratios, validated using cross-validation and expressed by the C-statistic. The study cohort consisted of 10,010 patients with first VTE receiving initial VKA treatment, mean age 62·2 years. Between 2008 and 2016, 167 significant bleeding events were recorded (1·7%), i.e. incidence rate was 7·4/100 person-years. Independent predictors for community-acquired significant bleeding included active cancer, trauma/surgical procedure, male gender, dementia, liver disease, anaemia, history of bleeding, cerebrovascular, renal and chronic pulmonary disease, VTE presenting as pulmonary embolism and age over 75. The overall C-statistic was 0·68 (95% CI, 0·60-0·76), 0·75 (0·60-0·88) for major bleeding and 0·65 (0·55-0·75) for CRNMB-H, and higher than in other risk schemes applied to our study population. The developed risk score may identify patients having a significant bleeding risk, in particular major bleeding events, in outpatients.