Development and internal validation of a nomogram to predict perioperative hypothermia in patients undergoing...

Perioperative hypothermia remains a common complication among patients undergoing laparoscopic gynecologic surgery under general anesthesia and is associated with adverse postoperative outcomes. Early identification of patients estimated to be at high risk may facilitate targeted preventive strategies. However, practical and individualized prediction tools for perioperative hypothermia in laparoscopic gynecologic surgery remain limited.

This single-center, retrospective observational cohort study included adult patients who underwent laparoscopic gynecologic surgery under general anesthesia between January 2019 and February 2024. The prediction model was developed and internally validated using a split-sample design within this gynecologic laparoscopic surgery cohort. Perioperative hypothermia was defined as a core body temperature <36.0 °C during the intraoperative or immediate postoperative period. Patients were randomly divided into a training cohort for model development (n = 294) and a split-sample internal validation cohort for model assessment (n = 130). Candidate predictors were selected a priori based on clinical relevance. Feature selection was performed using least absolute shrinkage and selection operator regression and the Boruta algorithm. A multivariable logistic regression model was developed and visualized as a nomogram. Model discrimination, calibration, and clinical utility were evaluated using the area under the receiver operating characteristic curve (AUC), calibration plots, and decision curve analysis, respectively. Split-sample validation was supplemented by bootstrap optimism-corrected internal validation using 500 resamples.

A total of 424 patients were included, of whom 279 (65.8%) developed perioperative hypothermia, a relatively high incidence that may reflect the outcome definition, routine temperature surveillance, and the characteristics of laparoscopic gynecologic surgery. Feature selection consistently retained age, body mass index, and operative time as the final model predictors. These variables were incorporated into the final prediction model and nomogram. In the training cohort, the model demonstrated good discrimination with an AUC of 0.788 (95% CI, 0.732-0.843), good calibration, and favorable clinical utility. In the split-sample internal validation cohort, discrimination remained stable with an AUC of 0.810 (95% CI, 0.727-0.893), and calibration and decision curve analyses showed consistent model performance. Bootstrap internal validation yielded an optimism-corrected AUC of 0.776 and an optimism-corrected calibration slope of 0.91. At the training-derived Youden cutoff, the model achieved a sensitivity of 0.623 and specificity of 0.853 in the training cohort, and a sensitivity of 0.925 and specificity of 0.580 in the validation cohort.

A simple nomogram based on age, body mass index, and operative time was developed and internally validated to estimate the predicted probability of perioperative hypothermia specifically in adult patients undergoing laparoscopic gynecologic surgery under general anesthesia. Within this surgical setting, the tool may assist clinicians in perioperative risk stratification, triage for intensified temperature monitoring or active warming, and individualized temperature-management planning. External validation in independent laparoscopic gynecologic cohorts is required before broader clinical implementation.