Use of artificial intelligence to study the hospitalization of women undergoing caesarean section

Abstract

Abstract Objective The incidence of caesarean sections (CSs) has increased significantly in recent years, especially in developed countries. This study aimed to identify the factors that most influence the length of hospital stay (LOS) after a CS, using data from 9,900 women who underwent CS at the “Federico II” University Hospital of Naples between 2014 and 2021. Methods Various artificial intelligence models were employed to analyze the relationships between the LOS and a set of independent variables, including maternal and foetal characteristics. The analysis focused on identifying the model with the best predictive performance and specific comorbidities impacting LOS. Results A multiple linear regression model determined the highest R-value (0.815), indicating a strong correlation between the identified variables and LOS. Significant predictors of LOS included abnormal foetuses, cardiovascular disease, respiratory disorders, hypertension, haemorrhage, multiple births, preeclampsia, previous delivery complications, surgical complications, and preoperative LOS. In terms of classification models, the decision tree yielded the highest accuracy (75%). Conclusions The study concluded that certain comorbidities, such as cardiovascular disease and preeclampsia, significantly impact LOS following a CS. These findings can assist hospital management in optimizing resource allocation and reducing costs by focusing on the most influential factors.

Keywords

• Caesarean section
• Machine learning
• Regression model
• Public health
• Length of stay