Causation Prediction Tool May Lower Rates of Pediatric Antibiotic Prescriptions for Diarrhea
When physicians were told that a case was likely viral they were less likely to prescribe antibiotics, a new study shows.
A tool that helps predict the likelihood that a particular case of diarrhea has a viral cause could help reduce the rate of inappropriate antibiotics prescriptions, according to a new study.
World Health Organization (WHO) recommendations indicate that children should only be prescribed antibiotics for diarrhea if they have bloody diarrhea or are suspected of having cholera. Yet, previous research has shown that many healthcare providers do not follow those guidelines, a problem that contributes to early exposure to antibiotics among children in low- and middle-income countries.
In a new study published in JAMA Pediatrics, corresponding author Eric J. Nelson, MD, PhD, of the University of Florida, and colleagues, described the impact of a diarrheal etiology prediction (DEP) algorithm that can help providers better understand the likely cause of particular cases of diarrhea. The DEP tool uses a variety of data, including patient-specific factors like medical history and symptoms, as well as location-specific data, such as the clinical presentation of previous patients and local weather patterns, to predict the probability that a particular case is viral in nature. Nelson and colleagues hoped that such a tool might help curb the overuse of antibiotics by highlighting cases for which an antibiotics prescription would probably be ineffective.
The investigators took their tool to 7 clinics in Mali and Bangladesh. There, they provided training to physicians on WHO treatment guidelines and introduced them to a clinical decision support tool. Next, physicians were randomly assigned to use the tool without the DEP algorithm (control arm) or with the DEP algorithm. The study was conducted with a crossover design; after 4 weeks and a one-week washout period, clinicians switched tools for the remaining 4 weeks of the study.
Altogether, 30 physicians and 941 patients with diarrhea (with a median age of 12 months) were enrolled in the study. The investigators found that 69.8% of children whose physicians had the DEP tool were given antibiotics, while 76.5% of children whose physicians did not have the DEP tool were prescribed antibiotics. Though the rate of antibiotic prescriptions was lower in the DEP group, the difference did not reach statistical significance.
However, when the investigators conducted a post-hoc analysis to account for the predicted probability of a viral-only etiology, they found that a 10% increase in the predicted probability of viral-only diarrhea resulted in a 14% drop in the likelihood that a patient received antibiotics.
The investigators noted that both groups had lower-than-expected rates of diarrhea prescription. Based on previous literature, Nelson and colleagues said they expected that around 90% of patients would be given an antibiotics prescription.
“This lower rate may be due to inclusion criteria that did not discriminate based on dehydration levels and/or a Hawthorne effect on physicians in the context of research cognizant that indications for antibiotic use are few for diarrheal illnesses,” Nelson and colleagues wrote.
On a physician level, the data showed that some providers did not change their behavior at all when they had access to the DEP prediction tool, and those physicians had the highest rates of antibiotic use during their time in the control arm. This suggests some physicians are less receptive to clinical decision tools, they said.
“This finding reveals a need to delineate these physician types, and to identify other modifiable factors that influence antibiotic prescribing,” Nelson and colleagues wrote.
The investigators concluded that the study served as a meaningful proof of concept for probability-based decision-support tools, and offered evidence that such tools can reduce unnecessary antibiotics use, particularly in under-resourced countries.
“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,” they wrote.
