腹腔镜袖状胃切除术患者术后肺部并发症预测模型的构建和验证
打开文本图片集
【Abstract】ObjectiveTo investigate risk factorsand develop/validate clinical prediction models for postoperative pulmonary complications （PPCs） in patients undergoing laparoscopic sleeve gastrectomy （LSG） using machine learning technology.MethodsA retrospective cohort of186 patients who underwent LSG at Shanxi Bethune Hospital from 2019 to 2O22 was established.These patients were divided into a modeling cohort and a validationcohortina7：3ratio.BasedonthediagnosticcriteriaforPPCs，thepatientswerecategorizedinto the PPCsgroup（Group P）and non-PPCs group （Group N）.Variableswere screened through univariate-Logistic analysisand Lasso regresion.Prediction models for PPCs in LSG patients were developed using the random forest algorithmandmultivariate-Logisticregressionanalysis.Therisk factors forPPCsinLSGpatientswereanalyzed， andboth models werevalidated in the validationcohort.The clasification performance of theprediction models was evaluaed by the area under the ROC curve （AUC）.ResultsAmong the 186 patients，the incidence of PPCs was 21% .BMI，smoking history，preoperative chest X-ray abnormalities，and total sufentanil dosage were identified as independent risk factors for PPCs，while nerve block for postoperative analgesia served as a protective factor.In thetraining cohort，therandom forest model demonstratedanAUCof O.985and theLogisticregressionmodel of 0.789.Inthevalidationcohort，therandom forest modelachievedAUCvaluesof O.811and theLogisticregression modelof O.767，respectively.Therandom forest-basedpredictionmodel outperformed theLogisticregressionmodel， demonstrating superior predictive performance and was recommended for clinical prediction of PPCs．Conclusion The BMI，smoking history，and soon，was independent risk factors for PPCs.The random forest-based prediction model for PPCs in LSG patients has demonstrated significant predictive value.
【Key words】Gastrectomy；Perioperative period; Machine learning; Random forest; Clinical prediction model
随着生活条件的改善和饮食结构的改变，我国肥胖人口数量日益增长，超重和肥胖逐渐成为威胁健康的重大公共卫生问题1。（剩余9856字）
