Risk assessment models for venous thromboembolism in acutely ill medical patients. A systematic review.

Although the use of thromboprophylaxis is recommended for acutely ill medical patients at increased risk of venous thromboembolism (VTE), it remains unclear which risk assessment model (RAM) should be routinely used to identify at-risk patients requiring thromboprophylaxis. We therefore aimed to describe existing RAMs, and to compare these tools in terms of validity and applicability for clinical decision-making. We performed a comprehensive systematic search in MEDLINE from the date of initiation until May 2016 for studies in acutely ill medical patients investigating validity of RAMs for VTE. Two reviewers independently screened the title, abstract, and full text, and evaluated the characteristics of studies, and the composition, evidence of validation, and results on validity of the RAMs. We included 11 studies assessing eight RAMs: 4-Element RAM, Caprini RAM, a full logistic model, Geneva risk score, IMPROVE-RAM, Kucher Model, a "Multivariable Model", and Padua Prediction Score. The 4-Element RAM, IMPROVE-RAM, Multivariable Model, and full logistic model had derivation by identifying factors with predictive power. The other four RAMs were empirically generated based on consensus guidelines, published data, and clinical expertise. The Kucher Model, the Padua Prediction Score, the Geneva Risk Score and the IMPROVE-RAM underwent multicenter external validation. The Kucher Model, the Padua Prediction Score, and the Geneva Risk Score improved rates of thromboprophylaxis or clinical outcomes. In conclusion, existing RAMs to evaluate the need of thromboprophylaxis in acutely ill medical patients are difficult to compare and none fulfills the criteria of an ideal RAM. Nevertheless, the adequacy of thromboprophylaxis may be improved by implementing one of the validated RAMs.