Comparison between the Khorana prediction score and Caprini risk assessment models for assessing the risk of venous thromboembolism in hospitalized patients with cancer: a retrospective case control study

This study aimed to evaluate the optimal risk assessment model (RAM) to stratify the risk of venous thromboembolism (VTE) in hospitalized patients with cancer. We examined and compared the VTE predictive ability of the Khorana score (KS) and the Caprini RAM in hospitalized cancer patients.We performed a retrospective case-control study among hospitalized cancer patients admitted to a comprehensive hospital in China from January 2015 to December 2016. A total of 221 cases were confirmed to have VTE during hospitalization and 221 controls were selected randomly. The Caprini RAM and KS were implemented and the individual scores of each risk factor were summed to generate a cumulative risk score. Meanwhile, the sensitivity, specificity, areas under curve of the receiver operating characteristic curve and calibration of these 2 models were analysed.Significant differences were observed in risk factors between VTE and non-VTE hospitalized cancer patients and the VTE risk increased significantly with an increase in the cumulative KS or Caprini RAM score. A classification of 'high risk' according to KS and Caprini RAM was associated with 2.272-fold and 3.825-fold increases in VTE risk, respectively. However, the Caprini RAM could identify 82.4% of the VTE cases that required preventive anticoagulant therapy according to American College of Chest Physicians guidelines, whereas the KS could only identify 35.3% of the VTE cases. In addition, the areas under curve of Caprini RAM were significantly higher than those of the KS (0.705 ± 0.024 vs 0.581 ± 0.025, P < 0.001), with a best cut-off value of 5 score, which happened to be the cut-off value for high risk of VTE in Caprini RAM. Both Caprini RAM and KS showed an excellent calibration curve (0.612 vs 0.141, P > 0.05), but the risk of VTE events predicted by Caprini seemed closer to the observed risk of VTE events.The Caprini RAM was found to be more effective than the KS in identifying hospitalized patients with cancer at risk of VTE.