Validation of risk assessment models predicting venous thromboembolism in acutely ill medical inpatients: A cohort study

Background: Because hospital-acquired venous thromboembolism (VTE) represents a frequent cause of preventable deaths in medical inpatients, identifying at-risk patients requiring thromboprophylaxis is critical. We aimed to externally assess the Caprini, IMPROVE, and Padua VTE risk scores and to compare their performance to advanced age as a stand-alone predictor.

Methods: We performed a retrospective analysis of patients prospectively enrolled in the PREVENU trial. Patients aged 40 years and older, hospitalized for at least 2 days on a medical ward were consecutively enrolled and followed for 3 months. Critical ill patients were not recruited. Patients diagnosed with VTE within 48 hours from admission, or receiving full dose anticoagulant treatment or who underwent surgery were excluded. All suspected VTE and deaths occurring during the 3-month follow-up were adjudicated by an independent committee. The three scores were retrospectively assessed. Body mass index, needed for the Padua and Caprini scores, was missing in 44% of patients.

Results: Among 14 910 eligible patients, 14 660 were evaluable, of which 1.8% experienced symptomatic VTE or sudden unexplained death during the 3-month follow-up. The area under the receiver operating characteristic curves (AUC) were 0.60 (95% confidence interval [CI] 0.57-0.63), 0.63 (95% CI 0.60-0.66) and 0.64 (95% CI 0.61-0.67) for Caprini, IMPROVE, and Padua scores, respectively. None of these scores performed significantly better than advanced age as a single predictor (AUC 0.61, 95% CI 0.58-0.64).

Conclusion: In our study, Caprini, IMPROVE, and Padua VTE risk scores have poor discriminative ability to identify not critically ill medical inpatients at risk of VTE, and do not perform better than a risk evaluation based on patient's age alone.