Abstract: Objective To construct a risk prediction model for intraoperative hypothermia in patients undergoing radical resection of oral cancer based on various warming methods, to visualize the model and to evaluate its clinical value. Methods A total of 234 patients in a stomatological hospital of Nanjing were enrolled from January 2020 to November, and risk factors in hypothermia group (103 cases) and non-hypothermia group (131 cases) were compared and logistic regression was adopted for model construction. The discrimination of the model was evaluated by the area under the curve（AUC）. Model calibration was done by Hosmer-Lemeshow goodness-of-fit test. A total of 100 patients were selected as the validation group from December 2020 to February 2021 to evaluate the application value of the model. Results The factors entering the predictive model were age (OR=1.033), BMI (OR=0.346), basal body temperature (OR=0.338), intraoperative fluid infusion (OR=2.320) and heating blanket temperature (OR=0.236). The p-value in the Hosmer-Lemeshow goodness of fit test was 0.656; the area under the subject operating characteristic curve 0.812; Youden index 0.530; sensitivity 0.767; specificity 0.763. Conclusion The model is effective for predicting the occurrence of hypothermia during radical resection of oral cancer, which provides reference for medical staff to take preventive measures for high-risk patients in the early stage.

Key words: radical resection of oral cancer, intraoperative hypothermia, prediction model, establishment and validation