Abstract
OBJECTIVE: We aimed to analyze the factors related to delay in transfer of patients in the post-anesthesia care unit (PACU) and to develop and validate a prediction model for understanding these factors to guide precise clinical intervention. METHODS: We collected data from two cohorts of 1153 and 297 patients who underwent surgery and were treated in the PACU at two time points. We examined their clinical features and anesthesia care data using analytical methods such as logistic regression, Random Forest, and eXtreme Gradient Boosting (Xgboost) to screen out variables and establish a prediction model. We then validated and simplified the model and plotted a nomogram. Using LASSO regression, we reduced the dimensionality of the data. We developed multiple models and plotted receiver operating characteristic (ROC) and calibration curves. We then constructed a simplified model by pooling the identified variables, which included hemoglobin (HB), alanine transaminase (ALT), glucose levels, duration of anesthesia, and the minimum bispectral index value (BIS_min). RESULTS: The model had good prediction performance parameters in the training and validation sets, with an AUC of 0.909 (0.887-0.932) in the training set and 0.939 (0.919-0.959) in the validation set. When we compared model 6 with other models, the net reclassification index (NRI) and the integrated discriminant improvement (IDI) index indicated that it did not differ significantly from the other models. We developed a scoring system, and it showed good prediction performance when verified with the training and validation sets as well as external data. Additionally, both the decision curve analysis (DCA) and clinical impact curve (CIC) demonstrated the potential clinical efficacy of the model in guiding patient interventions. CONCLUSION: Predicting transfer delays in the post-anesthesia care unit using predictive models is feasible; however, this merits further exploration.