A Multicenter,Prospective,Observational Cohort Study To Validate A Prediction APP For Perioperative Hypothermia

Objectives:Inadvertent perioperative hypothermia is a common but preventable complication,which may cause many adverse consequences.At present,it is known that many factors,such as patients’ condition,anesthesia,surgery,environment and warming status,can affect the risk of hypothermia.However,there have been no externally well validated tools for rapid identification of perioperative hypothermia in surgical patients.The objective of this study was to validate the accuracy of a previously established prediction model for estimating the risk of perioperative hypothermia in a multicenter,prospective cohort.Methods:In this observational study,consecutive adults scheduled for elective surgery under general anesthesia have been enrolled prospectively at 30 centers since May 25,2021.There was no additional intervention except routine practice in operation,anesthesia and perioperative temperature management.An intraoperative hypothermia risk score was calculated by a mobile application(APP)based on the prediction model.A wireless axillary thermometer was used to continuously measure perioperative core temperature as the reference standard.The discrimination and calibration of the model were assessed,using the area under the receiver operating characteristic curve(AUC),calibration plot,Hosmer-Lemeshow goodness-of-fit test,Brier score,and etc.Subgroup analyses stratified by different predicting factors were performed.The potential risk factors of perioperative hypothermia were evaluated by multivariable logistic regression,and the independent risk factors were selected for model refinement.Results:Up to the time of this data analysis,317 participants from 3 centers were included.213(67.2%)developed perioperative hypothermia,and 132(41.6%)received perioperative forced-air warming.The APP had an AUC of 0.721(95%confidence interval[CI],0.663-0.780)in the overall cohort,with the calibration plot showing adequate calibration(Hosmer-Lemeshow P=0.247;Brier score=0.20[95%CI,0.170.22]).By two cutoffs of 70 and 90,we categorized the risk scores into low-,moderate-,high-risk group,in which the incidence of perioperative hypothermia was 47.2%(95%CI,37.7%-56.7%),72.2%(95%CI,63.8%-80,7%),and 83.4%(95%CI,76.3%-90.7%),respectively(P for trend<0.001).The performance of APP for different subgroups was basically consistent with that in the overall cohort.The refined model showed significantly better discrimination(AUC=0.785;95%CI,0.735-0.829)compared to the original APP(P=0.022),with good calibration according to calibration plot(HosmerLemeshow P=0.213;Brier score=0.17[95%CI,0.14-0.19]).Conclusions:The perioperative hypothermia prediction model demonstrated clinically acceptable discrimination and calibration in this multicenter prospective validation.The model can identify patients with different risk stratification of perioperative hypothermia.The performance of APP for each subgroup remained basically stable.The discrimination of the refined prediction model is better than that of the original APP,but the refined one still needs to be further modified and validated after the multicenter data acquisition being accomplished.These findings suggest that the risk screening model could facilitate future perioperative temperature management.