Explainable AI to predict a complex multifactorial outcome, childhood obesity: Application to clinical epidemiology

Background Childhood obesity, driven by genetic and epidemiological factors, poses significant health risks, yet traditional machine learning models lack interpretability for clinical use.

This study aims to apply Kolmogorov-Arnold Networks (KAN), an explainable machine learning model, to predict body mass index (BMI) at age 8 as an indicator of obesity risk and to develop a publicly accessible prediction tool.

We utilized the Raine Study Gen2 cohort (n=2,868) to train KAN and traditional models (such as Random Forest, Gradient Boosting, Lasso, and Multi-Layer Perceptron) using perinatal, early-life, and polygenic risk score (PGS) data collected before age 5. Feature importance was analyzed across all the models. A publicly accessible online calculator was developed for practical use.

KAN achieved an R2 of 0.81, outperforming traditional models. Key predictors included Year 5 BMI z-score, mid-arm circumference, occupation of mother, and PGS. The online calculator supports predictions without PGS, maintaining an R2 of 0.81.

KAN’s transparent formulas enhance interpretability, offering a practical approach to predicting childhood obesity. The freely accessible tool enables clinicians to implement personalized prevention strategies, advancing precision medicine. Competing Interest Statement The authors have declared no competing interest. Funding Statement This study did not receive any funding. Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Our Ref: 2025/ET000396 Monday, 28 April, 2025 Dr Fuling Chen The University of Western Australia Dear Dr Chen HUMAN RESEARCH ETHICS EXEMPTION FROM REVIEW. Early Life Origins of Cardiovascular Health: A Machine Learning-Driven Risk Prediction Algorithm for Pediatric Populations Based on the information you have provided to the Human Ethics office in relation to the above project, the described activity has been assessed as exempt from ethics review at the University of Western Australia. However, should there be any significant changes to the project, you must contact the HREO to determine whether your exempt status remains valid or whether you will be required to submit an application for ethics approval. If you have any queries please contact the Human Ethics office at humanethics{at}uwa.edu.au. Please ensure that you quote the file reference 2025/ET000396 and the associated project title in all future correspondence. Yours sincerely Senior Officer Human Ethics & Clinical Trials I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Footnotes Update the comparative models and results. Add the online accessible tool. Data Availability The datasets generated and analyzed during this study are not available. The Raine study is committed to a high level of confidentiality of the data in line with the informed consent provided by participants. Requests for data should be directed to the Raine Study Executive. Abbreviations - EN - Elastic Net - ERF - Extreme Random Forest - GBM - Gradient Boosting Machine - KAN - Kolmogorov Arnold Network - Lasso - Least Absolute Shrinkage and Selection Operator - MLP - Multi-Layer Perceptron - PGS - Polygenic Score - RFE - Recursive Feature Elimination - XGB - Extreme Gradient Boosting