Predicting Post-Anesthesia Care Unit (PACU) Length of Stay (LOS) Using Machine Learning for Patients Undergoing Lumbar Spinal Stenosis Surgery

Background Lumbar spinal stenosis surgery is commonly performed to address conditions such as spinal canal narrowing and degenerative changes. The duration of a patient's stay in the Post-Anesthesia Care Unit (PACU) following surgery is influenced by a variety of factors including potential complications, anesthetic management, and the patient's overall health status. This study aims to analyze clinical data to identify the key factors that affect the length of stay in the PACU. By doing so, the study seeks to provide valuable insights that can lead to improvements in postoperative recovery and overall patient outcomes. Methods We collected data on 539 cases of patients undergoing lumbar spinal stenosis surgery under general anesthesia from August 2018 to December 2022. The cases were divided into three groups: Group A with 377 cases, Group B with 82 cases, and Group C with 80 cases. Univariate logistic regression analysis was conducted on Group A using SPSS software to identify factors significantly associated with postoperative retention in the PACU. Multivariate logistic regression was then applied to these selected factors to determine independent risk factors. The independent risk factors were used to construct a Nomogram predictive model using R software. Group B was utilized to externally validate the predictive model. Group C data was used for the evaluation of the predictive model. The model's consistency was assessed by calculating the C-index, constructing calibration plots, and generating the Receiver Operating Characteristic (ROC) curve to evaluate the model's predictive accuracy and discrimination ability. Results Univariate logistic regression analysis was conducted to identify factors associated with prolonged stay in the Post-Anesthesia Care Unit (PACU), revealing age, Body Mass Index (BMI), coronary heart disease, surgery duration, and creatinine clearance rate as significant predictors. Subsequently, a multivariate logistic regression analysis was performed on these identified factors, yielding age, BMI, and muscle strength as independent risk factors for extended PACU stay. A Nomogram predictive model was constructed using the R programming language. The model's consistency was assessed across Groups A, B, and C by calculating the C-index and generating Receiver Operating Characteristic (ROC) curves, demonstrating good consistency across the three groups. Conclusions The Nomogram predictive model demonstrates acceptable performance in certain groups (Groups A and C) but requires improvement in Group B. The model exhibits satisfactory calibration in Groups A and C, with notable deviations in high probability regions. Group B shows adequate calibration across most ranges but similar deviations in high probability regions. Discrimination is good in Groups A and C but is suboptimal in Group B. The independent risk factors for postoperative retention in the Post-Anesthesia Care Unit (PACU) following surgery for lumbar spinal stenosis were identified as Body Mass Index (BMI), surgery duration, and muscle strength. The Nomogram predictive model demonstrated good predictive performance and consistency, effectively forecasting the likelihood of postoperative PACU retention in patients undergoing lumbar spinal stenosis under general anesthesia. This model serves as a reference for personalized anesthetic management.