Project Details

Early Career

Status: Funded - Open

Establishing clinical predictors for improved management of pediatric diarrhea in the developing world

Joel Howard, MD


BACKGROUND: Diarrheal illness is a leading cause of morbidity and mortality among children worldwide, particularly in low and middle income countries (LMICs). Shigella infection is, among other pathogens, an important infection in this population and is treatable with antibiotics. As delayed treatment of this pathogen is associated with growth failure, accurate and cost-effective determination of diarrhea etiology is important for proper case management in children and for public health. With no testing readily available, clinicians often treat children in these settings with antibiotics in case these pathogens are present. GAP: The development of molecular diagnostic assays to detect diarrheal pathogens in stool has increased the speed and accuracy of pathogen detection, but at significant costs. Cost constraints in LMICs limit etiological diagnosis, and up to 70% of patients with acute diarrhea are prescribed empiric antibiotics, unnecessarily subjecting many patients to potential side effects. Electronic clinical decision support contained on a personal smartphone is one method to influence provider prescribing behavior. HYPOTHESIS: We hypothesize individual findings classically thought to be predictive of bacterial pathogens, such as bloody diarrhea, will under perform while incorporation additional clinical elements into a prediction rule will be more suggestive of a bacterial process than previously recognized. METHODS: We will use a combination of traditional and machine learning statistical methods to evaluate for predictors of bacterial etiology contained within clinical elements of the GEMS (Global Enteric Multicenter Study) dataset, a large multicenter, international cohort of 15,283 children with diarrhea presenting for care. RESULTS: Variables predictive of a viral etiology included lower age, a dry and cold season, increased height-for-age z-score (HAZ), lack of bloody diarrhea, and presence of vomiting. We created a clinical decision support tool using elements of these 5 variables which had an AUC of 0.825, achieving a specificity of 0.85, a sensitivity of 0.59, a negative predictive value of 0.82 and a positive predictive value of 0.64. IMPACT: With improved understanding of pretest probabilities for Shigellosis and other treatable bacterial causes, expensive testing can be limited to those who stand to benefit from treatment. This will enhance diagnostic stewardship in settings with already limited resources, while simultaneously improving patient outcomes by limiting unnecessary antibiotic use. Future studies are needed to assess if the diagnostic accuracy of the clinical decision support tool is sufficient to improve antimicrobial prescribing while maintaining desirable patient outcomes. Website Link: