Researchers at the University of California San Francisco have developed a computer model that combines AI and genetics to diagnose pneumonia in critically ill patients with 96% accuracy.
The model pairs a generative AI analysis of electronic health records with the biomarker FABP4 — a gene found in lung fluid. FABP4 is expressed at lower levels in infected lung cells, making it a useful biomarker to differentiate between infectious and noninfectious respiratory failure.
In an observational study published Dec. 16 in Nature Communications, the model outperformed ICU clinicians in identifying pneumonia. If implemented at admission, it could have reduced inappropriate antibiotic use by more than 80%, researchers said.
The study drew data from two patient cohorts: 98 individuals enrolled before the COVID-19 pandemic and 59 during it. The AI component — powered by GPT-4 and run on a privacy-protecting platform developed at UCSF — matched the diagnostic accuracy of internal medicine and infectious disease specialists.
At the Becker's 11th Annual IT + Revenue Cycle Conference: The Future of AI & Digital Health, taking place September 14–17 in Chicago, healthcare executives and digital leaders from across the country will come together to explore how AI, interoperability, cybersecurity, and revenue cycle innovation are transforming care delivery, strengthening financial performance, and driving the next era of digital health. Apply for complimentary registration now.