Identifying Metabolic Dysfunction-Associated Steatotic Liver Disease Using Natural Language Processing in a US...

Area of Science:

• Hepatology
• Medical Informatics
• Epidemiology

Background:

• Metabolic dysfunction-associated steatotic liver disease (MASLD) and MASLD with increased alcohol intake (MetALD) are recognized clinical entities.
• A lack of tools hinders patient identification from electronic health records (EHRs) for large-scale outcome studies.

Purpose of the Study:

• To develop and validate an algorithm for identifying and classifying steatotic liver disease (SLD) in a large veteran population using EHR data.
• To assess the algorithm's accuracy in diagnosing MASLD, MetALD, and alcohol-associated steatotic liver disease (ALD).

Main Methods:

• A retrospective study utilized a rule-based natural language processing (NLP) algorithm on EHR data from 817,657 veterans (2013-2022).
• The algorithm identified hepatic steatosis, cardiometabolic risk factors, and alcohol use.
• Algorithm-derived diagnoses were validated against a blinded chart review.

Main Results:

• Over half of the cohort (53.5%) had SLD, with MASLD in 36.5% and MetALD in 12.1%.
• The NLP algorithm demonstrated high accuracy (κ=0.86 for steatosis, κ=0.87 for classification) compared to chart review.
• The algorithm identified significantly more MASLD cases (36.5%) than traditional ICD codes (2.8%).

Conclusions:

• An algorithm combining NLP, cardiometabolic factors, and alcohol use accurately identifies and classifies SLD in EHRs.
• This validated algorithm is suitable for large epidemiologic studies of SLD within the Veterans Health Administration.