Machine learning for estimating and comparing clinical rules for treating diarrheal illness with antibiotics.

Acute diarrheal disease is one of the leading causes of death in children under age 5, disproportionately impacting children in low-resource settings. Many of these cases are caused by bacteria and therefore could respond to antibiotic treatment; however, the benefits of widely prescribing antibiotics must be weighed against the risks for the emergence of microbial resistance. These challenges present the opportunity for developing individualized treatment guidelines for diarrheal disease. In this study, we utilize a framework for the creation and evaluation of individualized treatment rules that leverage diagnostic and other clinical information to recommend antibiotic treatment to children with watery diarrhea. In contrast to many applications of pipelines for creating and evaluating treatment rules, we (i) explicitly consider creating rules that limit the proportion of children treated under a rule, to limit risks for overtreatment and the emergence of microbial resistance and (ii) propose methods to compare the performance of rules based on different sets of input covariates, which allows for quantification of the impact of measuring additional diagnostic biomarkers in clinical settings. We use a nested cross validation procedure that makes use of ensemble machine learning and doubly-robust estimation approach to derive, evaluate, and compare rules. We demonstrate that our proposed method yields appropriate inference in a realistic simulation study and apply our method to a real-data analysis of the AntiBiotics for Children with severe Diarrhea (ABCD) trial.