A comprehensive review of venous thromboembolism risk assessment models for hospitalized medical patients: comparative evidence, implementation challenges, and future directions.

Venous thromboembolism (VTE) is a leading cause of preventable hospital-acquired morbidity and mortality. Despite the availability of effective prophylaxis, its application in clinical practice remains inconsistent, often due to uncertainty in risk stratification. This review evaluates the validity and implementation of VTE risk assessment models (RAMs) in medical inpatients. Seven widely used RAMs, Caprini, Padua, IMPROVE, IMPROVEDD, Wells, Geneva, and Kucher e-alert, are critically examined alongside emerging digital and biomarker-enhanced tools. The Padua and IMPROVE scores show consistent reliability across various medical populations, while the Caprini RAM remains the most accurate in surgical contexts. The Wells deep vein thrombosis (DVT) and revised Geneva scores are preferred for diagnosing suspected thrombosis and pulmonary embolism, respectively. Electronic alerts, such as the Kucher and Woller models, have shown promise in increasing prophylaxis adherence and reducing symptomatic VTE events. Nonetheless, challenges like limited external validation, gaps in clinician training, and inconsistent local protocols hinder their real-world application. Future research should recalibrate RAMs for underrepresented groups, incorporate biomarkers and mobility data, and create user-friendly AI tools that can optimize the balance between thrombosis and bleeding risks. The adoption of validated, user-friendly RAMs is essential for improving thromboprophylaxis, enhancing patient safety, and reducing the burden of hospital-associated VTE.