New Study: AI Could Boost Detection of Bird Flu in Human Patients
AI tool finds possible bird flu infections overlooked in emergency departments.
Researchers at the University of Maryland School of Medicine (UMSOM) have demonstrated a powerful new use of artificial intelligence (AI) to detect potential bird flu infections that may otherwise go unnoticed in hospital emergency departments. Their findings, published in Clinical Infectious Diseases, show that generative AI can rapidly scan electronic medical records for high-risk exposures to H5N1 avian influenza.
A press release explained that the team applied a large language model (LLM) to 13,494 emergency visits across the University of Maryland Medical System in 2024. All patients had acute respiratory symptoms—such as cough, fever, or congestion—or conjunctivitis, which can mirror early signs of bird flu. The model flagged 76 notes that mentioned possible high-risk exposures, such as working as a butcher or on a farm with livestock.
Often these exposures were “incidentally” documented—for example, as part of a patient’s occupation—rather than because clinicians suspected avian influenza. After review, 14 patients were confirmed to have had recent, relevant exposure to poultry, wild birds, or livestock. None were specifically tested for H5N1, but the model was able to surface “needle in a haystack” cases buried among thousands of routine respiratory illnesses.
“This study shows how generative AI can fill a critical gap in our public health infrastructure by detecting high-risk patients that would otherwise go unnoticed,” said corresponding author Katherine E. Goodman, PhD, JD, assistant professor of Epidemiology & Public Health at UMSOM.
“With H5N1 continuing to circulate in U.S. animals, our biggest danger nationwide is that we don’t know what we don’t know,” noted Katherine E. Goodman, PhD, JD, assistant professor of Epidemiology & Public Health at UMSOM. (Photo credit: UMSOM)
Goodman emphasized that without consistent tracking of potential exposures, infections may be slipping through the cracks.
“Because we are not tracking how many symptomatic patients have potential bird flu exposures, and how many of those patients are being tested, infections could be going undetected,” she said. “It’s vital for healthcare systems to monitor for potential human exposure and to act quickly on that information.”
Bird Flu’s Spread in U.S. Livestock and Birds
The urgency stems from the current spread of H5N1 among U.S. livestock and birds. Since early 2024, the virus has been detected in more than 1,075 dairy herds across 17 states, with over 175 million poultry and wild birds testing positive. Human cases remain rare—70 confirmed with one fatality as of mid-2025, according to the CDC—but experts caution that many more may have gone undetected due to limited testing. The emergence of strains capable of human-to-human spread remains a concern.
Efficiency and scalability were key strengths of the AI approach. “The AI review required only 26 minutes of human time and cost just 3 cents per patient note, demonstrating high scalability and efficiency,” said co-author Anthony Harris, MD, MPH, acting chair of Epidemiology & Public Health at UMSOM. “This method has the potential to create a national network of clinical sentinel sites for emerging infectious disease surveillance.”
The LLM used, GPT-4 Turbo, proved highly accurate when tested on 10,000 historical emergency visits from 2022–2023, showing a 90% positive predictive value and a 98% negative predictive value. Still, researchers noted the model occasionally flagged lower-risk exposures, such as to dogs, underscoring the need for human review.
Next Steps for Real-Time Monitoring
Looking ahead, the team hopes to integrate the tool directly into electronic health records for real-time alerts. This would allow clinicians to ask about exposures, order targeted tests, and isolate patients if needed. Currently, CDC surveillance depends on mandated lab reports, with no system to track whether clinicians are documenting animal exposure in symptomatic patients.
UMSOM dean Mark T. Gladwin, MD, called the work a glimpse of what’s possible with big data and AI. “We are at the forefront of a disruptive but incredibly promising revolution around big data and artificial intelligence,” he said. “As this study demonstrates, [we] can use AI and big data to identify early signals of emerging infectious diseases like bird flu to enable us to take action sooner to test for these diseases and keep them from spreading.”
The project relied on the University of Maryland Institute for Health Computing (UM-IHC), a collaboration established in 2022 to combine computational, clinical, and data resources across the state’s medical and academic institutions.
The research was supported by the federal Agency for Healthcare Research and Quality, with additional data storage and computing resources provided by UM-IHC.
—Janette Wider
s
