Taking a page from marketplace investigation tactics, UC San Francisco experts are studying whether artificial intelligence (AI) tin amended test of a analyzable liver information by utilizing nan objective notes of aggregate providers.
Their caller study, published successful Gastro Hep Advances, focused connected hepatorenal syndrome (HRS), a analyzable information associated pinch liver illness that is often difficult to diagnose during hospitalization. The researchers sought to study if ample connection models could analyse nan objective notes of aggregate physicians and different providers to amended diagnostic accuracy and streamline diligent care.
"The conception is inspired by sentiment study exertion commonly utilized pinch reviews successful online shopping platforms, wherever AI summarizes corporate opinions," said Jin Ge, MD, MBA, UCSF adjunct professor of medicine and gastroenterologist, who led nan study.
We utilized this attack to find if corporate sentiment could foretell an HRS diagnosis."
Jin Ge, MD, MBA, Assistant Professor, University of California - San Francisco
The study compared accepted diagnostic methods based connected objective variables, specified arsenic laboratory results, pinch an AI-enhanced exemplary that incorporated sentiment study derived from objective notes. Incorporating AI-generated sentiment scores importantly improved predictive accuracy for HRS test upon diligent discharge.
The exertion offers clarity successful situations wherever conflicting recommendations among wellness attraction professionals whitethorn arise, providing a unified summary of nan attraction team's statement for clinicians and patients alike. While still successful nan investigation phase, this exertion has nan imaginable to toggle shape decision-making successful hospitals and heighten diligent outcomes.
"Using nan 'wisdom of nan crowd' doesn't conscionable foretell outcomes, it offers a directional penetration into what nan objective attraction squad collectively thinks astir a patient's condition," said Ge. "For cases pinch mixed opinions aliases uncertainty, AI-generated summaries could thief align attraction decisions and expedite curen plans."
The study has not yet been implemented successful objective believe but could pave nan measurement for early trials. The researchers purpose to measure really this accusation mightiness power real-world decision-making and diligent care.
Source:
Journal reference:
Lai, M., et al. (2025). Clinical Sentiment Analysis by Large Language Models Enhances Prediction of Hepatorenal Syndrome. Gastro Hep Advances. DOI:10.1016/j.gastha.2025.100797. https://www.sciencedirect.com/science/article/pii/S2772572325001840.
English (US) ·
Indonesian (ID) ·