Predicting High-Risk Colorectal Polyps in African Americans Using Pre-Colonoscopy Clinical Features: Machine Learning Model Development and Temporal Validation

The study developed and validated machine learning models to predict high-risk colorectal polyps using non-invasive pre-colonoscopy features in a predominantly African American cohort. Various algorithms, including neural networks, random forests, SVM, and XGBoost, were evaluated on a dataset of 4,681 patients for internal validation and 1,562 patients for external validation. This approach aims to improve risk stratification and equitable access to surveillance, potentially optimizing resource allocation in healthcare settings with limited colonoscopy availability.