Hepatic steatosis, or nonalcoholic fatty liver disease (NAFLD), affects a significant portion of the global population and can lead to more severe liver conditions, including hepatic fibrosis. Early and accurate risk prediction of fibrosis is crucial for timely intervention. Traditional diagnostic methods are invasive and carry risks, while imaging techniques and blood-based biomarkers have limitations in routine general practice. This study presents a machine learning-based clinical decision support system designed to assess the risk of hepatic fibrosis in patients with NAFLD using routine laboratory tests. The framework is developed using electronic health record data collected over 15 years, initially encompassing 1,272,572 patients from general practice. After applying clinical selection criteria, two cohorts of 12,960 and 25,478 patients were used for model development and evaluation. The proposed approach provides a robust foundation for monitoring fibrosis risk by implementing a novel screening method , which preprocesses predictors by leveraging well-established clinical indicators (e.g., hepatic steatosis index, fibrosis-4 index), alongside a selected minimal number of predictors, making it practical and cost-effective for widespread clinical use. The study's findings indicate promising results for screening and monitoring fibrosis risk in NAFLD patients, achieving the best AUC of 92.97%, PRAUC of 75.44%, and Sensitivity of 79.63%.
Bernardini et al. (Mon,) studied this question.