肿瘤预防与治疗2026，Vol.39Issue(3)：179-189,11.DOI:10.3969/j.issn.1674-0904.2026.03.002
基于机器学习算法的胸科肿瘤术后急性疼痛预测模型的构建
Construction of an Acute Postoperative Pain Prediction Model for Thorac-ic Oncology Surgery Based on Machine Learning Algorithms
摘要
Abstract
Objective:To determine the incidence and significant risk factors for acute pain following thoracic oncology surgery,to develop multiple risk prediction models,and to evaluate and identify the optimal model,thereby providing evi-dence for targeted clinical pain management.Methods:A prospective study was conducted on 647 participants who underwent thoracic oncologic surgery at a tertiary cancer hospi-tal in Sichuan Province from November 2022 to December 2023 and met the inclusion criteria.LASSO regression was used to identify risk factors for acute postoperative pain in thoracic tumor patients.Based on these factors,six machine learn-ing models were developed using R software.The predictors in the optimal model were visualized,and the key factors influ-encing acute postoperative pain in thoracic tumor patients were identified and ranked by their importance.Results:Among the thoracic tumor patients included in the study,the incidence of acute postoperative pain was 21.63%.Based on the influ-encing factors identified by LASSO regression,we constructed multiple machine learning prediction models.After compre-hensive consideration of several performance metrics,including the AUC and sensitivity,the XGBoost model was ultimately identified as the optimal predictive model.Furthermore,the SHAP(SHapley Additive exPlanation)method was employed for interpretability analysis of the XGBoost model.The visualization results revealed that the primary factors influencing acute postoperative pain in patients undergoing thoracic tumor surgery,ranked from highest to lowest importance,were:duration of chest tube placement,number of chest tubes placed,surgical procedure(surgical resection site),postoperative analgesia method,use of hypnotics the night before surgery,single-port or multi-port approach,postoperative diagnosis,long-term use of hypnotics,C-reactive protein,and ethnicity.Conclusion:The XGBoost model demonstrated a superior ability to predict acute postoperative pain following thoracic tumor surgery,thereby improving screening accuracy and providing a valuable ref-erence for the development of preventive and interventional strategies.关键词
胸科肿瘤/手术/术后急性疼痛/影响因素/预测模型Key words
Thoracic oncology/Surgery/Acute postoperative pain/Risk factors/Predictive model分类
医药卫生引用本文复制引用
杨青,张瑜,张甜,王禛..基于机器学习算法的胸科肿瘤术后急性疼痛预测模型的构建[J].肿瘤预防与治疗,2026,39(3):179-189,11.基金项目
This study was supported by grants from Sichuan Provincial Health Care Committee(No.Sichuan Cadre Research 2023-807). 四川省保健委员会办公室普及应用项目(编号:川干研 2023-807) （No.Sichuan Cadre Research 2023-807）
