Unrecognized Fibrosis Risk in MASLD: A Real-World Analysis and the Case for AI-Augmented Stratification

Current fibrosis risk stratification in metabolic dysfunction-associated steatotic liver disease (MASLD) relies on provider-initiated noninvasive testing and referral, making it vulnerable to variability in awareness, documentation, and follow-through. We aimed to quantify care gaps associated with this provider-dependent approach and explore opportunities for artificial intelligence to improve MASLD detection and management.

We performed a retrospective analysis of all adults undergoing abdominal ultrasound in 2024 at Ochsner Health's South Shore campuses. Natural language processing identified reports with hepatic steatosis, and patients with at least 1 cardiometabolic risk factor were included. Fibrosis-4 index (FIB-4) scores were calculated from recent laboratory data (within 6 months of ultrasound) using age-adjusted thresholds to classify patients as low, indeterminate, or high fibrosis risk. Management was defined as hepatology referral for high- or indeterminate-risk patients and documentation of a primary care provider for low-risk patients requiring reassessment.

Among 14,814 adults with ultrasound in 2024, 3052 (20.6%) met the MASLD criteria. Based on age-adjusted FIB-4, 15.2% were high risk, 18.0% indeterminate, and 66.0% low risk for advanced fibrosis. Of 465 high-risk patients, only 33.5% had hepatology referrals, leaving 309 (10.1% of the MASLD cohort) without appropriate specialty evaluation. Among 549 indeterminate-risk patients, 58.7% lacked referral for secondary assessment. In the low-risk group, 224 (7.3%) had no documented primary care provider for follow-up, and 24 (0.8%) lacked sufficient laboratory data for FIB-4 calculation. Overall, 28.0% of the MASLD cohort had a critical, moderate, or monitoring care gap.

Significant gaps persist in MASLD fibrosis risk stratification and management, largely reflecting system-level coordination failures rather than access barriers. Artificial intelligence-driven workflows integrated into the electronic health record could automate steatosis detection, calculate FIB-4 scores, flag care gaps, and prompt risk-stratified referrals or reassessments, offering a scalable solution to standardize MASLD management and improve outcomes.