Algorithm to Avoid Unneeded Antibiotics Whiffs for Kids' Diarrhea

For physicians using a smartphone-based clinical decision-support tool, incorporating an algorithm aimed at estimating the probability that diarrhea is due to a viral cause failed to reduce antibiotic prescribing, a randomized crossover trial conducted in Mali and Bangladesh found.

In the study of more than 900 children with acute diarrhea, no significant difference was observed in the proportion of children receiving antibiotics when physicians used the algorithm versus when they did not (69.8% vs 76.5%; risk difference [RD] -4.2%, 95% CI -10.7 to 1.0), reported Eric Nelson, MD, PhD, of the University of Florida in Gainesville, and colleagues in JAMA Pediatrics.

However, a post hoc analysis that assessed antibiotic prescribing when the diarrheal etiologic prediction (DEP) algorithm pointed to viral etiology showed potential benefit (RD -5.6%, 95% CI -12.8 to -10). A 14% drop in the likelihood of antibiotic prescribing for the DEP group was seen for every 10% increase in the predicted probability of viral diarrhea (OR 0.86, 95% CI 0.76-0.96).

"If replicated, the use of etiological prediction in decision support tools represents an important advancement to improve antibiotic stewardship in a clinical context prone to high rates of inappropriate antibiotic use," the researchers concluded.

Overuse of antibiotics can lead to antimicrobial resistance and put children at risk for adverse events, Nelson's group noted. But many children in lower-income nations are exposed to antibiotics for viral diarrhea despite recommendations from the World Health Organization that use of antibiotics should be reserved for pediatric diarrhea cases involving a suspicion of cholera, or for bloody diarrhea.

"Due to overuse of antibiotics, they no longer work against bacteria in many countries," said coauthor Daniel Leung, MD, MSc, of the University of Utah School of Medicine in Salt Lake City.

"One way to fix that is to reduce antibiotic use by healthcare workers," he told MedPage Today. "However, healthcare workers do not have many tools to help make antibiotic decisions."

The researchers had previously developed the DEP algorithm to predict diarrhea etiology based on statistical models from a large multicenter study of pediatric diarrhea. "DEP draws upon data from clinical history and symptoms of the patient (patient-specific sources) and location-specific sources, such as clinical presentation of prior patients, historical prevalence, and weather parameters," they explained.

For their study, Nelson and colleagues enrolled 30 physicians who treated 941 children with diarrhea at three centers in Bangladesh (November 2020 to January 2021) and at four centers in Mali (January to March 2021).

Physicians were randomized to treat their patients with the smartphone-based clinical decision-support tool alone or with DEP for 4 weeks. After 4 weeks, physicians underwent a washout period for 1 week, followed by a subsequent crossover period for another 4 weeks.

Explaining why the primary outcome was not met, the researchers suggested that the trial may have been underpowered. They also noted findings suggesting "that older physicians were less willing to change [prescribing] behavior based on the predicted [DEP] value."

Pediatric participants (median age 12 months, 57% boys) were blinded to the intervention. To be eligible, patients had to be between 2 and 59 months of age, have acute diarrhea (at least 3 stools within 24 hours), and have cell phone access for follow-up. Children with severe sepsis, pneumonia, meningitis, malnutrition, or conditions other than gastroenteritis were excluded.

Nearly all kids in the study achieved diarrhea resolution at 10 days after discharge, with similar rates among the DEP (97.9%) and no DEP (98.6%) groups. On average, diarrhea symptoms took 2 to 3 days to resolve.

Severe dehydration rates were higher among participants in Bangladesh than in Mali (11.1% vs 0.22%). Overall, one patient in each group (DEP or no DEP) died after discharge. Adverse events and severe adverse effects were uncommon in the two groups.

The authors acknowledged limitations to the data, including the small number of physicians, the lack of blinding among physicians, and that diagnostic stool tests were not performed.

Comment