Inadvertent intraoperative hypothermia (IOH) significantly increases the risk of complications in elderly patients undergoing general anesthesia. This study aimed to develop and validate a predictive model for IOH specifically for elderly patients in Southwest China. A total of 443 patients aged ≥ 60 years from 24 hospitals were enrolled and randomly assigned to a derivation cohort (n = 310) and an internal validation cohort (n = 133). Logistic regression, LASSO regression, and random forest models were developed, with internal validation used to select the optimal approach. External validation was performed on 334 patients from 4 participating and 4 non-participating hospitals. In the internal validation, the Logistic Regression model outperformed machine learning algorithms, demonstrating an area under the receiver operating characteristic curve (AUC) of 0.841 (95% CI: 0.772–0.910), with a specificity of 81.48%, sensitivity of 70.21%, and optimal risk threshold of 48%. In the external validation cohort (n = 334), the model maintained robust discriminative ability (AUC: 0.760; 95% CI: 0.703–0.817). Crucially, in a subgroup analysis of patients with normal temperature (≥ 36.5℃) at 10 min before anesthesia induction, the model successfully identified 34 high-risk individuals who would likely be overlooked by standard clinical assessment. The Logistic Regression model effectively predicts IOH risk in elderly patients, and a risk probability ≥ 48% serves as a critical threshold to guide stratified temperature management.
Wei et al. (Fri,) studied this question.