Related Experiment Video
Updated: Jan 12, 2026
06:09
Author Spotlight: Advancing Hepatic Fibrosis Diagnosis Using Magnetic Resonance Elastography and AI
Published on: July 21, 2023
1.8K
Use of Natural Language Processing to Objectively Identify Hepatic Encephalopathy in Multiple Cohorts
Scott Silvey1, Nilang Patel2, Brian C Davis2
1Department of Population Health, Virginia Commonwealth University, Richmond, Virginia, USA.
The American Journal of Gastroenterology
|November 6, 2025
Summary
Natural language processing (NLP) identified five key terms to objectively define hepatic encephalopathy (HE) in hospitalized patients. This approach can improve liver transplant prioritization and clinical trial analysis for HE.
Area of Science:
- Medical Informatics
- Hepatology
- Natural Language Processing
Background:
- Hepatic encephalopathy (HE) presents a significant burden but lacks standardized diagnostic criteria for clinical trials and transplant prioritization.
- Subjectivity in HE definition complicates multicenter studies and outcome analysis.
- Objective diagnostic markers are needed to standardize HE assessment.
Purpose of the Study:
- To develop and validate an objective method for identifying hepatic encephalopathy (HE) episodes in hospitalized patients using natural language processing (NLP).
- To establish a standardized definition of HE for improved consistency in clinical research and liver transplant evaluation.
Main Methods:
- Natural language processing (NLP) was employed to extract terms associated with HE from electronic health records.
- Terms were identified through guideline review and physician input, then validated across derivation and two independent cohorts.
- Machine learning cross-validation was used to assess the contribution of individual terms and their combinations in diagnosing HE.
Main Results:
- Five key terms (asterixis, altered mental status, confusion, lactulose initiation/continuation, rifaximin initiation/continuation) were identified with high sensitivity and negative predictive value (NPV) for HE.
- The derived NLP model demonstrated strong performance across derivation (85.8% sensitivity/82.1% NPV) and validation cohorts (100% sensitivity/NPV internally, 94.1% sensitivity/84.2% NPV externally).
- Machine learning analysis confirmed the top five features and highlighted the superior diagnostic capability of their combination.
Conclusions:
- A set of five simple phrases extracted via NLP from inpatient charts can objectively define hepatic encephalopathy (HE).
- This NLP-driven approach offers a standardized method for HE identification, potentially improving liver transplant prioritization.
- The objective criteria can enhance the reliability of HE event adjudication in clinical trials.
