University of Maryland School of Medicine Study: Harnessing AI to Enhance H5N1 Surveillance
In a groundbreaking study from the University of Maryland School of Medicine, researchers have unveiled a novel application of artificial intelligence (AI) designed to fortify the surveillance of H5N1 avian influenza, commonly known as bird flu. As the virus continues its unsettling spread among U.S. animal populations, the research highlights a crucial advancement in public health monitoring. The findings were recently published in the prestigious journal Clinical Infectious Diseases.
The researchers employed a state-of-the-art generative AI large language model (LLM) to analyze 13,494 emergency department visits from adult patients across the University of Maryland Medical System (UMMS). These visits, spanning urban, suburban, and rural environments in 2024, included patients presenting with acute respiratory symptoms and conjunctivitis—common indicators of potential H5N1 infection. The central objective was to determine how effectively generative AI could identify patients at high risk who might have been overlooked during initial assessments.
With a broad scan of all emergency department notes, the AI model flagged 76 cases where patients reported high-risk exposures to bird flu (like working in a slaughterhouse or on a farm with livestock). Notably, these exposures were often documented incidentally, revealing a systemic gap in monitoring for potential avian influenza infections during routine care.
Upon a thorough review by the research team, 14 patients were identified as having had relevant exposure to H5N1-carrying animals, including poultry and livestock. While these patients had not been specifically tested for the avian virus, the AI’s capabilities made it possible to unearth critical cases that would typically get lost among the noise of everyday healthcare.
“This study demonstrates how generative AI can bridge an urgent gap in our public health infrastructure by spotting high-risk patients who otherwise might remain undetected,” states Katherine E. Goodman, PhD, JD, the study’s corresponding author and a faculty member at the University of Maryland Institute for Health Computing (UM-IHC). “Given the ongoing circulation of H5N1 in U.S. animal populations, it is vital that we track symptomatic patients for potential avian flu exposure.”
The urgency of this research is underscored by the alarming statistics: since early 2024, over 1,075 dairy herds across 17 states have been reported infected, and more than 175 million poultry and wild birds tested positive during the outbreak. Although human cases remain rare—with only 70 confirmed infections and a singular fatality recorded in the U.S. by mid-2025 according to the Centers for Disease Control and Prevention (CDC)—the researchers express concern that undetected infections may be more widespread due to insufficient testing methods.
“The AI review required just 26 minutes of human time and a mere 3 cents per patient note, showcasing impressive scalability and efficiency,” notes co-author Anthony Harris, MD, MPH, Professor and Acting Chair of Epidemiology & Public Health at UMSOM. “This method offers the possibility of establishing a nationwide network of clinical sentinel sites for emerging infectious disease surveillance.”
The performance metrics of the LLM (GPT-4 Turbo) were compelling, achieving a 90% positive predictive value and a 98% negative predictive value. These results were evaluated against a historical sample of 10,000 emergency department visits from 2022 to 2023, prior to the onset of H5N1 in domestic livestock. However, the model exhibited a tendency to err on the side of caution when identifying relevant avian flu exposures, leading to a necessity for human oversight in cases it flagged.
“We stand at the brink of a transformative revolution in big data and artificial intelligence,” declares Mark T. Gladwin, MD, Dean of UMSOM and Vice President for Medical Affairs at University of Maryland, Baltimore. “The collaborative efforts of our engineers and clinician-researchers at the Institute for Health Computing exemplify how we can utilize AI and extensive data to identify early signs of emerging illnesses like bird flu and take timely action.”
This study also benefited from contributions by various faculty members at UMSOM, including Laurence S. Magder, PhD, Jonathan D. Baghdadi, PhD, MD, and Daniel J. Morgan, MD, MS, who are recognized for their expertise in Epidemiology & Public Health.
University of Maryland, College Park, the University of Maryland, Baltimore, and the University of Maryland Medical System. This institute consolidates the computational acumen, clinical knowledge, biomedical innovation, and health data resources provided by these institutions.
“As an academic health system, our duty is to prepare for future medical breakthroughs while providing care today. Our long history of using data to inform health research positions us as national leaders in this domain,” comments Mohan Suntha, MD, MBA, President and CEO of the University of Maryland Medical System. “The data across our system also reflects the diverse communities we serve, enhancing our ability to address public health challenges effectively.”
Funding for this significant research endeavor was provided by the federal Agency for Healthcare Research and Quality. Additionally, costs associated with computing and data storage necessary for the LLM analyses were supported by the UM Institute for Health Computing.
About the University of Maryland School of Medicine
The University of Maryland School of Medicine, established in 1807 as the first public medical school in the U.S., is recognized as one of the leading biomedical research institutions worldwide. With nearly $500 million in total research funding, 46 departments, centers, and institutes, and a network of over 2,200 student trainees and more than 3,000 faculty members, the School is at the forefront of addressing critical health challenges. The collaboration with the University of Maryland Medical Center serves nearly 2 million patients annually, while the School’s global reach encompasses research and treatment facilities in 33 countries. To explore more about its groundbreaking work, visit medschool.umaryland.edu.