Predictive Modeling 0f Length 0f Stay in General Surgery Patients Using Artificial Intelligence

Abstract

For effective resource allocation, patient management, and discharge planning, it is crucial to accurately forecast the length of stay (LOS) for patients undergoing general surgery. In this study, we suggest a predictive modeling strategy utilizing Artificial Intelligence (AI) methods to calculate the LOS for patients with adult spinal deformity (ASD).  LOS following ASD surgery denoted a crucial phase to enable the best possible recovery. The categorization of high-risk patients is made possible by predictive algorithms that estimate LOS. Patients with ASD were found in a multicenter database that was prospectively gathered. Patients who had staged surgery or a LOS for more than 30 days were not included. Redundancy and collinearity tests, as well as univariable predictor importance of 0.90, were used to choose the variables for the model. Using a dataset created from a bootstrap sample, the Gradient Ascent Decision Tree Model (GADTM) was suggested for prediction; patients who were not by chance chosen for the bootstrap sample were selected for the dataset. To determine an accuracy percentage, LOS forecasts, and actual LOS were compared. 653 patients complied with the inclusion requirements. 893 patients were modeled using bootstrapping. Accuracy of the prediction within two days of the actual LOS. Our approach accurately predicted LOS after ASD surgery within two days. Rehab accommodation and social assistance services are not included in large projected databases. Predictive analytics will become more important in ASD surgery as future models improve accuracy