Objective To develop a risk assessment tool for venous thromboembolism (VTE) in Chinese hospitalized cancer patients. Methods Totally, 1,195 cancer patients were divided into a VTE group (n=595) and a non-VTE group (n=600). Based on the Khorana score and relevant guidelines in China, 24 potential risk factors were screened for further analysis. Then 18 significant predictors were identified via univariate analysis, of which 14 remained significant in binary logistic regression analysis. A new risk assessment was developed based on the 14 risk factors (15 items), and the effectiveness to predict VTE was compared with the Khorana score. Results The total score of the new risk assessment was 16. Area under the receiver operating characteristic curve (AUC) for prediction of VTE was 0.826 (95%CI 0.804-0.847). A cut-off value of ≥3.5 was determined, with the sensitivity of 75.8% and specificity of 74.2%, indicating superior performance over the Khorana score. Conclusion The modified Khorana score can be used to assess VTE risk in hospitalized cancer patients in China, and the performance is superior to that of the Khorana score.