Inspired by sentiment analysis used in online marketplaces, researchers at UC San Francisco are exploring how artificial intelligence can improve the diagnosis of hepatorenal syndrome (HRS), a complex and often challenging liver-related condition. In a recent study published in Gastro Hep Advances, the team investigated whether large language models (LLMs) could analyze clinical notes from multiple healthcare providers to enhance diagnostic accuracy and support better patient care. HRS, a serious complication of advanced liver disease, is difficult to diagnose during hospitalization due to overlapping symptoms and inconsistent clinical assessments. Traditional diagnostic methods rely heavily on lab results and established clinical criteria, but these can sometimes fall short in the face of clinical uncertainty. The study introduced a novel approach: using AI to perform sentiment analysis on the unstructured text of clinical notes written by various members of the care team—doctors, nurses, and other specialists. By evaluating the collective tone and content of these notes, the AI model generated a sentiment score reflecting the team’s overall clinical impression of a patient’s condition. Jin Ge, MD, MBA, assistant professor of medicine at UCSF and a gastroenterologist who led the research, explained the concept: “The idea is similar to how online platforms use AI to summarize customer reviews. We asked whether the collective sentiment from multiple providers’ notes could help predict an HRS diagnosis more accurately.” The results showed that combining traditional clinical data with AI-derived sentiment scores significantly improved the model’s ability to predict HRS at the time of patient discharge. The AI not only enhanced accuracy but also provided a unified, data-driven summary of the care team’s collective assessment—especially valuable in cases where providers offer conflicting opinions. “This isn’t just about predicting outcomes,” said Ge. “It’s about capturing the ‘wisdom of the crowd’—giving clinicians a clearer picture of what the team collectively believes about a patient’s condition. In uncertain cases, this can help align treatment decisions and speed up care.” While the technology remains in the research phase and has not yet been implemented in clinical settings, the findings open a promising path for future applications. The team plans to conduct further studies to assess how these AI-generated insights influence real-world decision-making, care coordination, and patient outcomes. The research, led by Mason Lai and colleagues, was published in Gastro Hep Advances under the title Clinical Sentiment Analysis by Large Language Models Enhances Prediction of Hepatorenal Syndrome. The study highlights the potential of AI to transform complex clinical decision-making by turning fragmented, human-generated notes into actionable, consensus-driven insights.