Predicting the risk of Clostridium difficile infection following an outpatient visit: development and external validation of a pragmatic, prognostic risk score

Increasing morbidity related to Clostridium difficile infection (CDI) has heightened interest in the identification of patients who would most benefit from recognition of risk and intervention. We sought to develop and validate a prognostic risk score to predict CDI risk for individual patients following an outpatient healthcare visit. We assembled a cohort of Kaiser Permanente Northwest (KPNW) patients with an index outpatient visit between 2005 and 2008, and identified CDI in the year following that visit. Applying Cox regression, we synthesized a priori predictors into a CDI risk score, which we validated among a Kaiser Permanente Colorado (KPCO) cohort. We calculated and plotted the observed 1-year CDI risk for each decile of predicted risk for both cohorts. Among 356 920 KPNW patients, 608 experienced CDI, giving a 1-year incidence of 2.2 CDIs per 1000 patients. The Cox model differentiated between patients who do and do not develop CDI: there was a C-statistic of 0.83 for KPNW. The simpler points-based risk score, derived from the Cox model, was validated successfully among 296 550 KPCO patients, with no decline in the area under the receiver operating characteristic curve: 0.785 (KPNW) vs. 0.790 (KPCO). The predicted risk for CDI agreed closely with the observed risk. Our CDI risk score utilized data collected during usual care to successfully identify patients who developed CDI, discriminating them from patients at the lowest risk for CDI. Our prognostic CDI risk score provides a decision-making tool for clinicians in the outpatient setting.