Clinical predictors for etiology of acute diarrhea in children in resource-limited settings

Abstract

Background Diarrhea is one of the leading causes of childhood morbidity and mortality in lower- and middle-income countries. In such settings, access to laboratory diagnostics are often limited, and decisions for use of antimicrobials often empiric. Clinical predictors are a potential non-laboratory method to more accurately assess diarrheal etiology, the knowledge of which could improve management of pediatric diarrhea. Methods We used clinical and quantitative molecular etiologic data from the Global Enteric Multicenter Study (GEMS), a prospective, case-control study, to develop predictive models for the etiology of diarrhea. Using random forests, we screened the available variables and then assessed the performance of predictions from random forest regression models and logistic regression models using 5-fold cross-validation. Results We identified 1049 cases where a virus was the only etiology, and developed predictive models against 2317 cases where the etiology was known but non-viral (bacterial, protozoal, or mixed). 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. Cross-validation suggests an AUC of 0.825 can be achieved with a parsimonious model of 5 variables, achieving a specificity of 0.85, a sensitivity of 0.59, a NPV of 0.82 and a PPV of 0.64. Conclusion Predictors of the etiology of pediatric diarrhea can be used by providers in low-resource settings to inform clinical decision-making. The use of non-laboratory methods to diagnose viral causes of diarrhea could be a step towards reducing inappropriate antibiotic prescription worldwide.

Etiology and Incidence of Moderate-to-Severe Diarrhea in Young Children in Niger

“…The combined high burden of diarrhea due to bacterial and protozoal pathogens in this study of predominantly watery diarrhea raises the question as to whether current guidelines which defer antimicrobial therapy in the absence of bloody diarrhea are sufficient in this setting [ 6 ] and suggests the potential value of clinical prediction tools or point-of-care diagnostics to better target antimicrobial therapy [ 21 ]. Evidence from a multisite clinical trial is forthcoming about the role of azithromycin in treating watery diarrhea [ 22 ] which may help further direct the value of point-of-care identification of bacterial enteropathogens in the absence of dysentery.…”

Identifying weather variables as important clinical predictors of bacterial diarrhea among international travelers to Nepal and Thailand

“…We found that a 14-day average of environmental temperature and precipitation increased model performance. While prior studies have documented complex seasonal patterns among diarrheal diseases in tropical climates 2,8,22 , studies examining weather data for individual-level clinical prediction are lacking. Climate is thought to aid pathogen transmission through mechanisms such as contamination of food or water supplies, pathogen survival on fomite surfaces, or facilitating vector life cycles.…”

“…Recent advances in machine learning methods such as random forests offer new tools for developing clinical prediction models. [7][8][9] Prior studies using CPRs to predict bacterial diarrhea among children, 8,[10][11][12] and travelers, 13 have demonstrated promising results but have been limited by small sample sizes, sub-optimal performance characteristics of predictors, and limited pathogen identification. Development of new diagnostic tools such as the TaqMan® array card have recently improved identification of diarrheal pathogens.…”

A modular approach to integrating multiple data sources into real-time clinical prediction for pediatric diarrhea

“…We defined other known etiologies as having a majority attribution of diarrhea episode by at least one other non-viral pathogen. We exclude patients with unknown etiologies when fitting the model, though it has been previously shown that these cases have a similar distribution of viral predictions using a model with presenting patient information as those cases with known etiologies ( Brintz et al, 2020 ).…”

“…The patient model derived from the GEMS data treats each enrolled patient as an observation and uses their available patient data at presentation to predict viral only versus other etiology of their infectious diarrhea. In order to make a parsimonious model, we used the previously published random forests variable importance screening ( Brintz et al, 2020 ). Using the screened variables ( Table 1 ), we fit a logistic regression including the top five variables that would be accessible to providers at the time of presentation.…”