Document resource
Background and Aims This study aims to develop a personalized prediction tool utilizing the decision tree algorithm to assess the risk of hypothermia in patients undergoing thoracoscopic radical lung cancer surgery in the post-anesthesia care unit (PACU), thereby providing a basis for precision postoperative management.Methods This retrospective study included patients from Chongqing Songshan Hospital who underwent thoracoscopic radical lung cancer surgery between 2020 and 2024, totaling 420 cases that met the inclusion criteria, excluding those with severe underlying diseases or missing intraoperative temperature control data. Data collected included demographic characteristics, preoperative evaluations (BMI, ASA classification, basal metabolic rate), intraoperative variables (surgery duration, fluid management, anesthesia depth changes, and temperature fluctuations), along with PACU temperature records. Hypothermia was defined as a core temperature below 36°C. A decision tree algorithm was employed to construct the predictive model, which was assessed using five-fold cross-validation and confusion matrices. Key performance metrics included AUC, accuracy, sensitivity, and specificity.Results The incidence of hypothermia in the PACU was 21.7%. The decision tree model achieved an AUC of 0.90 (95% CI: 0.86–0.92), significantly higher than the logistic regression model (AUC = 0.82, P < 0.05). Key predictive factors identified were intraoperative temperature fluctuations, surgery duration, and basal metabolic rate, with the model showing a sensitivity of 84% and specificity of 88%. Targeted interventions based on the model effectively lowered the incidence of hypothermia in high-risk groups.Conclusions The decision tree-based predictive model developed in this study is accurate and reliable, effectively identifying high-risk patients for hypothermia in the PACU following thoracoscopic radical lung cancer surgery. This model serves as a scientific tool for postoperative risk assessment and can guide personalized temperature management, ultimately improving patient outcomes.