Related Experiment Video
Updated: May 5, 2026
08:58
Optimized Analysis of In Vivo and In Vitro Hepatic Steatosis
Published on: March 11, 2017
16.1K
Risk Prediction in Patients With Metabolic Dysfunction-Associated Steatohepatitis Using Natural Language Processing
Jordan Guillot1, Christopher Y K Williams1, Shadera Azzam1
1Bakar Computational Health Sciences Institute, University of California, San Francisco, California.
Gastro Hep Advances
|July 21, 2025
Summary
A novel natural language processing algorithm identified a real-world cohort of patients with metabolic dysfunction-associated steatohepatitis (MASH). This study highlights key comorbidities and lab values predicting MASH progression and mortality.
Area of Science:
- Hepatology
- Medical Informatics
- Public Health
Background:
- Metabolic dysfunction-associated steatohepatitis (MASH) is a heterogeneous liver disease with significant implications for end-stage liver disease.
- Understanding MASH progression in real-world populations is crucial for effective management and treatment development.
- Current real-world data on MASH progression and associated mortality remains limited.
Purpose of the Study:
- To define a real-world MASH cohort using natural language processing (NLP).
- To identify significant associations between MASH and all-cause mortality.
- To investigate predictors of progression to cirrhosis and liver transplantation.
Main Methods:
- Development and validation of a novel NLP algorithm, "NASHDetection," to identify MASH patients.
- Application of the algorithm to a patient cohort diagnosed between 2012 and 2022.
- Utilized Cox regression with stepwise variable selection to analyze outcomes.
Main Results:
- The NASHDetection algorithm achieved 86% accuracy in identifying 2695 MASH patients.
- At diagnosis, 55.4% had cirrhosis, 34.0% decompensation, and 10.8% hepatocellular carcinoma.
- Comorbidities like type 2 diabetes mellitus, heart failure, and peripheral artery disease, along with elevated LDL cholesterol and alkaline phosphatase, were associated with increased all-cause mortality.
Conclusions:
- A real-world MASH cohort was characterized using a novel NLP approach.
- Several comorbidities and laboratory markers were identified as potential predictors of MASH progression and mortality.
- NLP-driven patient characterization can facilitate future interventional trials for MASH.
