ABSTRACTRecurrentClostridium difficileinfection (rCDI) frequently complicates recovery from CDI. Accurately predicting rCDI would allow judicious allocation of limited resources, but published models have met with limited success. Thus, biomarkers predictive of recurrence have been sought. This study tested whether PCR ribotype independently predicted rCDI. Stool samples from nonpregnant inpatients ≥18 years of age with diarrhea were included from October 2010 to January 2013 after the patients tested positive forC. difficilein the clinical microbiology laboratory. Per guidelines, the rCDI was defined as a positive test forC. difficileat >2 weeks but ≤8 weeks from the index episode. For each sample, a single colony ofC. difficilewas isolated by anaerobic culture, confirmed to be toxigenic by PCR, and ribotyped. Simple logistic regression and multiple logistic regression were used to model the primary outcome of rCDI, incorporating a wide range of clinical parameters. In total, 927 patients with 968 index episodes of CDI were included, with 110 (11.4%) developing rCDI. Age and use of proton pump inhibitors or concurrent antibiotics did not increase the risk of rCDI. Low serum bilirubin levels and ribotype 027 were associated with increased risk of rCDI on unadjusted analysis, with health care-associated CDI being inversely associated. In the final multivariable model, ribotype 027 was the strongest independent predictor of rCDI (odds ratio, 2.17; 95% confidence interval, 1.33 to 3.56;P= 0.002). Ribotype 027 is an independent predictor of rCDI.IMPORTANCECDI is a major public health issue, with over 400,000 cases per year in the United States alone. Recurrent CDI is common, occurring in approximately one in five individuals after a primary episode. Although interventions exist that could reduce the risk of recurrence, deployment in all patients is limited by cost, invasiveness, and/or an undetermined long-term safety profile. Thus, clinicians need risk stratification tools to properly allocate treatments. Because prior research on clinical predictors has failed to yield a reliable, reproducible, and effective predictive model to assist treatment decisions, accurate biomarkers of recurrence would be of great value. This study tested whether PCR ribotype independently predicted rCDI, and the data build upon prior research in showing that ribotype 027 is associated with rCDI.