Risk-assessment models for predicting venous thromboembolism among hospitalized non-surgical patients: a systematic review.

Venous thromboembolism (VTE) prophylaxis is suboptimal in American hospitals despite long-standing evidence-based recommendations. Data from observational studies indicate a lower uptake of effective prophylaxis in patients hospitalized with medical versus surgical conditions. Reluctance to use prophylaxis in medical patients has been attributed to difficulty in identifying at-risk patients and balancing risks of bleeding against occurrence of VTE. Several risk-assessment models (RAMs) have been proposed to assist physicians in identifying non-surgical patients who need prophylaxis. We conducted a systematic review of published RAMs, based on objective criteria, to determine whether any RAM is validated sufficiently to be employed in clinical practice. We identified 11 RAMs, six derived from primary data and five based on expert opinion. The number, types, and strength of association of VTE risk predictors were highly variable. The variability in methods and outcome measurement precluded pooled estimates of these different models. Published RAMs for VTE lack generalizability and adequate validation. As electronic health records become more ubiquitous, validated dynamic RAMs are needed to assess VTE risk at the point-of-care in real time.