Derivation of a risk assessment model for hospital-acquired venous thrombosis: the NAVAL score

Abstract

Venous thrombosis (VT) is a preventable cause of death in hospitalized patients. The main strategy to decrease VT incidence is timely thromboprophylaxis in at-risk patients. We sought to evaluate the reliability of risk assessment model (RAM) data, the incremental usefulness of additional variables and the modelling of an adjusted score (the NAVAL score). We used the RAM proposed by Caprini for initial assessment. A 5 % systematic sample of data was independently reviewed for reliability. We evaluated the incremental usefulness of six variables for VT during the score modelling by logistic regression. We then assessed the NAVAL score for calibration, reclassification and discrimination performances. We observed 11,091 patients with 37 (0.3 %) VT events. Using the Caprini RAM, high-risk and moderate-risk patients were respectively associated with a 17.4 (95 % confidence interval [CI] 6.1-49.9) and 4.2 (95 % CI 1.6-11.0) increased VT risk compared with low-risk patients. Four independent variables were selected for the NAVAL score: "Age", "Admission clinic", "History of previous VT event" and "History of thrombophilia". The area under the receiver-operating-characteristic curve for the NAVAL score was 0.72 (95 % CI 0.63-0.81). The Net Reclassification Index (NRI) for the NAVAL score compared with the Caprini RAM was -0.1 (95 % CI -0.3 to 0.1; p = 0.28). We conclude that the NAVAL score is a simplified tool for the stratification of VT risk in hospitalized patients. With only four variables, it demonstrated good performance and discrimination, but requires external validation before clinical application. We also confirm that the Caprini RAM can effectively stratify VT risk in hospitalized patients in our population.

Validation of risk assessment models for venous thrombosis in hospitalized medical patients

“…The literature search produced 415 articles on December 17, 2017 (Figure ). After reviewing titles and abstracts, 12 RAMs from 11 developers were obtained to validate; the Padua prediction score, the electronic alert system by Kucher, the electronic alert system by Lecumberri, the NICE guideline, the PRETEMED guideline, the 4‐factor and 7‐factor predicted IMPROVE models, the Geneva risk score, model 2 as published by Zakai et al, the Intermountain risk score, the NAVAL score, and the Caprini risk assessment score . Almost all scores include the presence of active cancer and a previous venous thrombosis as predictors whereas, for example, predictors like ischemic stroke, lower‐limb paralysis or recent travel are included in only three RAMs.…”

“…In this study we externally validated 12 RAMs for venous thrombosis in medical patients . The best performing models in this study were the IMPROVE 7‐factor model, with an AUC of 0.66 (95% CI 0.56‐0.75), and the NICE score (AUC of 0.66; 95% CI 0.56‐0.75).…”

“…In this study we externally validated 12 RAMs for venous thrombosis in medical patients. [4][5][6][7][8][9][10][12][13][14][15] The best performing models in this study were the IMPROVE 7-factor model, with an AUC of 0.66 (95% CI 0.56-0.75), and the NICE score (AUC of 0.66; 95% CI 0.56-0.75). The negative predictive value was high for every RAM (range 98.6%-99.4%), meaning that all RAMs were able to correctly predict who would not develop venous thrombosis.…”

“…Patients hospitalized for acute medical illness have an eightfold increased risk of venous thrombosis as opposed to the general population . As routine pharmacological thromboprophylaxis for all patients is not recommended due to the associated bleeding risk, risk assessment models (RAMs) have been developed to identify high‐risk patients . International guidelines recommend to prescribe thromboprophylaxis in selected high‐risk medical patients by use of these models…”

“…3 As routine pharmacological thromboprophylaxis for all patients is not recommended due to the associated bleeding risk, risk assessment models (RAMs) have been developed to identify high-risk patients. [4][5][6][7][8][9][10][11][12][13][14][15] International guidelines recommend to prescribe thromboprophylaxis in selected high-risk medical patients by use of these models. 2,7 However, for high-risk medical patients, thromboprophylaxis remains underused as shown in the ENDORSE study where medical patients at risk of venous thrombosis actually received thromboprophylaxis in only 48% of cases.…”