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

Researchers developed machine learning models to predict high-risk colorectal polyps in African American patients using only pre-colonoscopy clinical features, potentially improving equitable access to preventive care. The study analyzed 4,681 patients for internal validation and 1,562 for external validation, employing multiple algorithms including neural networks, random forests, and XGBoost to stratify risk without invasive procedures.

This research addresses a significant healthcare equity gap by leveraging machine learning to improve early detection of advanced colorectal polyps in underserved populations. The study's focus on African Americans is particularly relevant, as this demographic experiences disproportionately higher colorectal cancer incidence and mortality rates, often due to delayed diagnosis and limited access to preventive screening. By developing predictive models based on non-invasive pre-colonoscopy data, researchers aim to optimize resource allocation in settings where colonoscopy capacity is constrained. This approach could identify high-risk patients who benefit most from immediate screening while deferring procedures for lower-risk individuals, reducing unnecessary healthcare utilization. The temporal validation methodology—using data from 2015-2022 for model training and 2023-2024 for external testing—demonstrates rigorous validation practices essential for clinical adoption. The multi-algorithm approach, comparing neural networks against traditional methods like logistic regression and XGBoost, allows identification of the most reliable predictive framework. Implementation of such models could enhance clinical decision-making in community hospitals and urban medical centers serving predominantly African American populations. For healthcare systems facing resource constraints, this represents a practical pathway to more equitable risk stratification. The research underscores how artificial intelligence can address healthcare disparities when specifically designed for underrepresented populations rather than applied post-hoc to existing disparate systems.

• →Machine learning models can predict high-risk colorectal polyps using only pre-colonoscopy clinical features without invasive procedures.
• →The study prioritizes healthcare equity by specifically analyzing a predominantly African American cohort experiencing disproportionate colorectal cancer burden.
• →Non-invasive risk stratification enables better resource allocation when colonoscopy capacity is limited in underserved settings.
• →Temporal validation using 2023-2024 data ensures model reliability and real-world applicability beyond the training period.
• →Multiple algorithms were compared to identify optimal predictive performance for clinical implementation.