Functional gastrointestinal disorders predictors in neonates and toddlers: A machine learning approach to risk assessment

Background: Functional Gastrointestinal Disorders (FGIDs) can pose a great burden on affected children, their families, and the healthcare system. Due to the lack of knowledge about the pre cise pathophysiology of FGIDs, a proper identification of children at risk to develop FGIDs has never been attempted. The research aims to identify early-life risk factors for FGIDs such as in fantile colic, regurgitation, and functional constipation, within the first year of life. Methods: This prospective observational cohort study enrolled both term and preterm infants from a tertiary care university hospital between January 1, 2020, and December 31, 2022. The study employed both traditional statistical methods and artificial intelligence (AI) techniques, specifically a random forest classification model, to identify key risk factors associated with the development of FGIDs. Based on these findings, an AI-based predictive model will be devel oped, along with a user-friendly, web-based interface designed for practical risk assessment. Results: 6060 infants were enrolled. 8.1 % were born preterm. According to random forest clas sification model by AI, birth weight (BW), cord blood pH, and maternal age were the most rel evant variables linked to development of FGIDs in the first year of life. Some discrepancies be tween potential risk factors identified through conventional statistics and AI were detected. Conclusion: For the first time machine learning allowed to identify BW, cord blood pH and ma ternal age as important variable for risk prediction of FGIDs in the first year of life. This practi cal risk assessment tool would help clinicians to identify infants at risk of FGIDs who would ben efit from a tailored preventive approach.