Use of machine learning for early prediction of short-term mortality in veterans with metabolic dysfunction-associated steatotic liver disease.

Frey LJ, Fuchs M, Ward RC, Gebregziabher M, Nasir AB, Natarajan Y, Schreiner A, Rockey DC, Syn WK. Use of machine learning for early prediction of short-term mortality in veterans with metabolic dysfunction-associated steatotic liver disease. PLoS ONE. 2025 Oct 27; 20(10):e0334715, DOI: 10.1371/journal.pone.0334715.

Related HSR&D Project(s)

• SDR 20-368 – Data-Driven Methods to Identify Social Determinants of Health

Dimensions for VA is a web-based tool available to VA staff that enables detailed searches of published research and research projects.

If you have VA-Intranet access, click here for more information vaww.hsrd.research.va.gov/dimensions/

VA staff not currently on the VA network can access Dimensions by registering for an account using their VA email address.    Search Dimensions for VA for this citation
* Don't have VA-internal network access or a VA email address? Try searching the free-to-the-public version of Dimensions

Search for Abstract from PubMed

---

Abstract:

BACKGROUND: Metabolic dysfunction associated steatotic liver disease (MASLD) is a leading cause of chronic liver disease worldwide and affects > 25% in the United States population. We hypothesized that clinical features present in electronic health records (EHR) could be extracted early to characterize patients with MASLD who are at high risk of early mortality and that machine learning models would predict mortality better than noninvasive assessments of liver disease/fibrosis. METHODS: Using previously published criteria for MASLD, applied to data from the US Veterans Affairs EHR, we identified a cohort of 13,071 patients between 2000 and 2018 who had an initial diagnosis of MASLD without clinical evidence of cirrhosis. We subsequently used machine-learning and conducted analysis of variance and logistic regression to identify clinical variables to characterize cirrhosis risk and predict mortality within the ensuing 5-years. RESULTS: The average age of the cohort was 60 years, had a BMI of 31, and 34% diabetes prevalence. Patients who progressed to cirrhosis were younger when first diagnosed with MASLD (56), had a higher BMI (33), and had significantly higher noninvasive fibrosis scores. Having diabetes at index MASLD diagnosis significantly increased the risk of developing cirrhosis and doubled the risk cirrhosis plus HCC (2.09 CI:1.217-3.63). Our machine-learning model performed significantly better than FIB-4 at predicting mortality within 5-years of being diagnosed with MASLD (AUC 83% vs 68%). CONCLUSION: Our data suggest that machine learning models based on data extracted from the EHR early during MASLD can identify patients likely to develop cirrhosis and predict short term mortality.