Clinical prediction rule to guide diagnostic testing forShigellosisand improve antibiotic stewardship for pediatric diarrhea

ABSTRACT Background Diarrheal diseases are a leading cause of death for children under-5. Identification of etiology helps guide pathogen-specific therapy, but availability of diagnostic testing is often limited in low resource settings. Our goal is to develop a clinical prediction rule (CPR) to guide clinicians in identifying when to use a point-of-care diagnostic for Shigella in children presenting with acute diarrhea. Methods We used clinical and demographic data from the Global Enteric Multicenter Study (GEMS) study to build predictive models for diarrhea of Shigella etiology in children ≤59 months presenting with moderate-to-severe diarrhea in Africa and Asia. We screened variables using random forests, and assessed predictive performance with random forest regression and logistic regression using cross-validation. We used the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) study to externally validate our GEMS-derived CPR. Results Of the 5011 cases analyzed, 1332 (27%) had diarrhea of Shigella etiology. Our CPR had high predictive ability (AUC=0.80 (95% CI: 0.79, 0.81) using the top two predictive variables, age and caregiver reported bloody diarrhea. We show that by using our CPR to triage who receives diagnostic testing, 3 times more Shigella diarrhea cases would have been identified compared to current symptom-based guidelines, with only 27% of cases receiving a point-of-care diagnostic test. Conclusions We demonstrate how a clinical prediction rule can be used to guide use of a point-of-care diagnostic test for diarrhea management. Using our CPR, available diagnostic capacity can be optimized to improve appropriate antibiotic use. Key points Using an externally validated clinical prediction tool to triage who receives diagnostic testing, 3 times more Shigella diarrhea cases would have been identified compared to current symptom-based guidelines, with only 27% of cases receiving a point-of-care diagnostic test.