Construction And Virification Of Intraoperative Hypothermia Among Lung Cancer People Risk Prediction Model Based On Machine Learning Algorithm

Research background and purpose:The aim of this study is to construct and verify the risk prediction models for intraoperative hypothermia in lung cancer patients based on machine learning algorithms,and to evaluate the prediction efficacy of each model by comparing the area under the curve(AUC)value to obtain the most suitable modeling algorithm for lung cancer patients,which lays a foundation for the subsequent establishment of intraoperative hypothermia prediction software,prediction system or App for lung cancer patients.To provide a reference for clinical medical staff to identify high-risk groups of intraoperative hypothermia in patients with lung cancer.Materials and Methods:This study was divided into two parts.Part I:Based on the understanding o f the influencing factors of hypothermia in adult patients,a self-designed informa tion questionnaire was used to collect the clinical data of 1100 valid lung cancer patients who underwent surgery in a tertiary cancer hospital in Sichuan Provinc e from June to November 2022 and met the inclusion and exclusion criteria,incl uding general demographic data,disease and surgery-related data.They were rand omly divided into training set(770 cases)and validation set(330 cases)accordin g to 7:3.Univariate analysis,Lasso regression analysis and Logistic regression a nalysis were performed on the data to identify the independent risk factors for i ntraoperative hypothermia in patients with lung cancer.Part Ⅱ:Based on the 7 i ndependent risk factors,R4.2.0 software was used to construct logistic regression model,random forest model,support vector machine model and extreme gradien t boosting model,and the prediction efficiency of each model was evaluated by comparing the AUC value and other indicators to obtain the most suitable algorit hm for the risk prediction model of intraoperative hypothermia in lung cancer pa tients.At the same time,individual models were visualized with the help of rms,ggplot2 and other packages,and finally the validation set patients were verified.Result:1.Analysis of intraoperative hypothermia and independent risk factors in patients with lung cancerThere were 410 lung cancer patients with intraoperative hypothermia in the t raining set,and the incidence of intraoperative hypothermia was 53.2%.Multivari ate analysis showed that intraoperative blood loss(OR=4.06,95%CI:2.24-3.71),int raoperative infusion volume(OR=5.46,95%CI:2.83-10.54),operation time(OR=3.28,95%CI:1.57-7.21),and intraoperative blood loss(OR=4.06,95%CI:2.24-3.71).An esthesia time(OR=8.3,95%CI:4.50-15.51),operating room temperature(OR=0.19,95%CI:0.13-0.28),core body temperature after anesthesia(OR=0.08,95%CI:0.04-0.14),Seven characteristic variables of surgical resection sites(left upper lobe,left l ower lobe,right upper lobe,right middle lobe)were independent risk factors for intraoperative hypothermia in patients with lung cancer.2.Comparison of several algorithms and performance of intraoperative hypothermia prediction models for lung cancer patientsThe results showed that the AUC of the training set of the four models ran ged from 0.928 to 0.968,among which the random forest model showed the hig hest AUC value,indicating that the recognition and accuracy of this model were the best compared with the other three models,so as to achieve the best predic tion effect.Finally,it is concluded that random forest algorithm is the optimal al gorithm for intraoperative hypothermia in lung cancer patients.Conclusion:Machine-learning algorithms offer better predictive techniques.The prediction model of intraoperative hypothermia in lung cancer patients constructed by rand om forest algorithm can better screen high-risk patients and help and improve th e intraoperative management of lung cancer patients.