Evaluating the Population-Level Impact of Predictive Machine Learning Frameworks on Community-Based Interventions for Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD)

• Billy Elly

  Lautech

  Author

Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) affects nearly 40% of adults worldwide, yet remains significantly underdiagnosed in community settings, with population-based prevalence studies indicating that approximately 70% of at-risk individuals have undetected steatotic liver disease. Despite the availability of evidence-based lifestyle interventions, current risk stratification approaches rely on retrospective cohort identification rather than prospective, predictive targeting, limiting the efficiency and population-level impact of community-based programmes. This study addresses this gap by developing and evaluating a predictive machine learning framework designed to identify high-risk MASLD individuals for targeted community-based intervention. Using a hybrid model combining gradient boosting with regularised logistic regression, applied to multi-source data from UK Biobank (n=3,123) and simulation-based prospective validation, the framework achieved an area under the curve (AUC) of 0.89 (95% CI: 0.87–0.91) for predicting MASLD diagnosis within 24 months, representing a significant improvement over traditional risk factor-based methods (AUC=0.76, p<0.001). Key predictors included HbA1c, Controlled Attenuation Parameter (CAP) values, obesity metrics, and dietary factors, with model explainability achieved through SHAP analysis. The framework demonstrated a positive predictive value of 24% in the top 1,000 high-risk individuals, compared to 12% for baseline methods, suggesting that ML-based targeting could improve intervention efficiency by approximately 100%. The study contributes a validated, replicable predictive framework for community-based MASLD intervention targeting, with practical implications for public health administrators seeking cost-effective population screening strategies. The findings support the integration of ML-driven risk stratification into routine community health assessments, potentially transforming MASLD from a condition identified late in its progression to one detected early enough for meaningful lifestyle intervention.

• pdf

Copyright (c) 2026 Billy Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.