In developing countries, inappropriate antibiotic use is a persistent, difficult challenge. An international group of scientists that includes investigators from Bangladesh, Mali, and the United States have a solution: an artificial intelligence-driven tool that uses an algorithm to estimate the probability a case of diarrhea is caused by a virus alone.
Antibiotics are not only ineffective against viral diarrhea; their misuse is a major contributor to global antibiotic resistance. To combat antibiotic misuse, the researchers developed a decision-making tool that tells doctors the probability that the culprit is solely a virus. Findings published today in the journal JAMA Pediatrics show the algorithm works: As the predicted probability of viral diarrhea increased, doctors participating in the study prescribed significantly fewer antibiotics to children.
In developing nations, the vast majority of pediatric diarrhea cases are treated with antibiotics.
“That means the vast majority of cases are treated inappropriately,” said Eric J. Nelson, M.D., Ph.D., a University of Florida Health physician, a tenured associate professor in the UF College of Public Health and Health Professions’ environmental and global health department and one of the study’s lead authors.
The Diarrheal Etiology Prediction algorithm, or DEP, works by harvesting a trove of data, including clinical history, patient-specific symptoms and local weather information. The DEP was developed by Daniel Leung, M.D., Ben Brintz, Ph.D., and their team at the University of Utah. The algorithm was then built into a smartphone-based clinical decision-support tool by Nelson. When seeing children with diarrhea, their hypothesis was that it would help doctors make better-informed decisions about antibiotic use.
To test that idea, Nelson and the U.S. team worked with scientists in Bangladesh and Mali in 2020 and 2021. Thirty doctors and nearly 1,000 patients between 2 months and 5 years old were enrolled at seven sites in the two countries.
When analyzing antibiotic use independent of the likelihood a patient had viral disease alone, the study found that the DEP was not associated with a reduction in antibiotic use. However, when the predicted probability of viral-only diarrhea increased by 10%, antibiotic prescriptions fell by 14% overall, the researchers found. Individually, most doctors prescribed fewer antibiotics as the predicted probability of viral diarrhea rose.
The DEP is particularly useful to doctors because it takes in relevant data about other recent patients in the same area, Nelson said. In Bangladesh, January typically brings more viral illnesses while bacterial diseases arise in the spring and fall. The algorithm collects localized weather data, including temperature spikes and rainfall, and adds it to regionalized epidemiologic models for viral diseases. These elements help to refine the DEP’s predictions, Nelson said
The DEP algorithm builds on the researchers’ earlier work to develop digital decision-support tools for treating diarrhea-related dehydration. Nelson and his collaborators are part of a group of researchers using technology to enable better medical decisions in developing countries.
“We’re trying to figure out how to build clinical decision-support software that is fast, easy to use, accurate and meets the needs of health care providers in global health settings,” said Nelson, a hospitalist in the UF College of Medicine’s department of pediatrics and a faculty member affiliated with UF’s Emerging Pathogens Institute.
While antibiotic misuse and resistance in other countries might seem like a faraway problem to the average American citizen, Nelson takes the longer view. Tourists who visit an overseas destination may bring back more than photos and souvenirs, such as multi-drug resistant enteric bacteria. Treating a resulting infection, such as a urinary tract infection for E. coli, then becomes a lot more problematic.
“That’s a very common problem and it’s really well-documented,” Nelson said.
Using an algorithm to enhance physicians’ decision-making also has other uses in both developing and developed countries, Nelson said. Nearly any disease that has seasonality and geographic localization, such as malaria and different respiratory infections, could potentially be
diagnosed and treated more accurately with algorithms similar to DEP, he added. Algorithms could also help give physicians more information about whether and when to use point-of-care diagnostic tests.
While algorithms and other technology help drive solutions, Nelson said it’s just one part of solving complex global health issues.
“At the end of the day, you have to have a really important problem and a really outstanding approach to address that problem. And that requires an outstanding clinical research design and an outstanding team and partnership with patients to get that study delivered,” he said.
In addition to researchers from the University of Utah, the team included collaborators from the International Centre for Diarrhoeal Disease Research (Bangladesh), the Center for Vaccine Development (Mali), the University of Maryland, Brown University and the Bill & Melinda Gates Foundation’s Institute for Disease Modeling. Research funding was provided by the Gates Foundation the National Institute of Allergy and Infectious Diseases and the National Institutes of Health.
Media contact: Doug Bennett at email@example.com or 352-265-9400.