Microbiome Data Distinguish Patients with Clostridium difficile Infection and Non-C. difficile-Associated Diarrhea from Healthy Controls

ABSTRACT Antibiotic usage is the most commonly cited risk factor for hospital-acquired Clostridium difficile infections (CDI). The increased risk is due to disruption of the indigenous microbiome and a subsequent decrease in colonization resistance by the perturbed bacterial community; however, the specific changes in the microbiome that lead to increased risk are poorly un-derstood.We developed statistical models that incorporated microbiome data with clinical and demographic data to better un-derstand why individuals develop CDI. The 16S rRNA genes were sequenced from the feces of 338 individuals, including cases, diarrheal controls, and nondiarrheal controls. Wemodeled CDI and diarrheal status using multiple clinical variables, including age, antibiotic use, antacid use, and other known risk factors using logit regression. This base model was compared to models that incorporated microbiome data, using diversity metrics, community types, or specific bacterial populations, to identify char-acteristics of the microbiome associated with CDI susceptibility or resistance. The addition of microbiome data significantly improved our ability to distinguish CDI status when comparing cases or diarrheal controls to nondiarrheal controls. However, only when we assigned samples to community types was it possible to differentiate cases from diarrheal controls. Several bacte-rial species within the Ruminococcaceae, Lachnospiraceae, Bacteroides, and Porphyromonadaceaewere largely absent in cases and highly associated with nondiarrheal controls. The improved discriminatory ability of our microbiome-based models con