Development of machine learning-based prediction models of inadequate postoperative analgesia after noncardiac surgery: a retrospective cohort study

Abstract Background Early identification of high-risk patients with inadequate postoperative analgesia is crucial in noncardiac surgery. This study aimed to develop prediction models for inadequate postoperative analgesia in noncardiac surgery using machine learning algorithms. Methods This article reports on a study that enrolled 199,517 adult patients who underwent noncardiac surgery. The discovery cohort included patients who had surgical procedures performed between June 2018 and April 2022, while the validation cohort included those who had surgeries between May 2022 and October 2022. Machine learning algorithms such as logistic regression (LR), random forest (RF), eXtreme gradient boosting tree (XGBoost), artificial neural network (ANN), and gradient boosting machine (GBM) were used to develop predictive models of inadequate postoperative analgesia based on perioperative variables. Results The overall prevalence of inadequate postoperative analgesia was 18.1%. Among the five machine learning algorithms we considered, GBM performed the best performance, with an AUROC value of 0.831 [95% CI, 0.827–0.834]. The SHAP analysis showed that surgery duration, type of surgery, anesthesia method, intraoperative fluid management, and use of hormones were the strongest five predictors. In addition, we observed that as the number of predicted features gradually decreased from all to 20, the performance of the prediction model exhibited only a marginal decline from 0.83 to 0.81. Conclusions This study demonstrates that the GBM algorithm demonstrated superior predictive performance in comparison to all other algorithms utilized. Screening for inadequate postoperative analgesia based on the prediction model could improve postoperative pain management.