RESUMO
Objective To validate the performance of the model for predicting the risk of intraoperative hypothermia.Methods This observational prospective study enrolled the adult patients who were of American Society of Anesthesiologists â -â ¢ and underwent elective surgery with general anesthesia in Peking Union Medical College Hospital,Beijing Hospital,and Xuanwu Hospital of Capital Medical University from October 2019 to August 2021.The risk prediction model of intraoperative hypothermia was used to calculate the predictors score of each patient.The body temperature of each patient was monitored throughout the perioperative period,and perioperative temperature management were entirely at the discretion of the anesthesiologists.The area under the receiver operating characteristic curve(AUC),Hosmer-Lemeshow goodness-of-fit test,and Brier score were employed to evaluate the prediction performance of the model.Results Of the 472 participants included in this study,141(29.9%)developed intraoperative hypothermia and 124(26.3%)received intraoperative active warming.For predicting intraoperative hypothermia in the overall cohort,the model demonstrated good discrimination capacity with an AUC of 0.729(95% CI=0.680-0.777),adequate calibration(Hosmer-Lemeshow χ2=3.143,P=0.925),and good overall performance(Brier score of 0.34).For the patients with passive warming only,the model showed good discrimination(AUC=0.756;95% CI=0.704-0.808),good calibration(Hosmer-Lemeshow χ2=7.457,P=0.488),and the Brier score of 0.29.For the patients with active warming,the model presented the AUC of 0.747(95% CI=0.632-0.863),Hosmer-Lemeshow χ2 of 4.754(P=0.783)and the Brier score of 0.47.Furthermore,we stratified the risk scores as low,moderate and high risk groups,in which the incidence of intraoperative hypothermia was 14.4%(95% CI=9.6%-19.1%),36.7%(95% CI=29.9%-43.5%),and 58.2%(95% CI=46.1%-70.3%),respectively.The differences between the three groups were statistically significant(χ2=54.112,P<0.001).Conclusion The intraoperative hypothermia prediction model demonstrates good overall differentiation capacity and has good prediction performance for the patients with or without active warming.