Epidemiology and recurrence risk prediction of Clostridium difficile Infections: A retrospective cohort study of the United States Veterans Health Care System
Reveles, Kelly Renee
MetadataShow full item record
Clostridium difficile infection (CDI) is the leading cause of bacterial infectious diarrhea in nosocomial settings and approximately 25% of patients with CDI experience disease recurrence. Prior CDI epidemiological investigations are limited though. They do not reflect the burden of CDI in federal facilities, nor do they capture recent estimates on patient health outcomes. Furthermore, few studies have integrated CDI recurrence risk factors into a tool that clinicians can use to identify patients at risk for CDI recurrence. This study 1) described the epidemiology of CDI in the national Veterans Health Administration (VHA), 2) derived and validated a clinical prediction rule for 60-day CDI first recurrence, and 3) derived and validated a clinical prediction rule for 60-day CDI second recurrence. This was a retrospective cohort study of VHA beneficiaries with CDI between October 1, 2001 and September 30, 2012. VHA clinical and pharmacy data were integrated to develop several independent variables, including patient baseline demographics, CDI characteristics, comorbidities, concomitant medications, prior medications, prior hospitalization, hospital length of stay (LOS), and CDI severity. The dependent variables included 30/60/90-day mortality, and 30/60/90-day CDI recurrence. CDI incidence and outcomes were presented descriptively and compared using generalized linear regression models. CDI recurrence prediction rules were derived using multivariable logistic regression models and validated using the area under the receiver-operating-characteristic curve (AUROC), sensitivity, specificity, positive predictive value, and negative predictive value. Our study demonstrated that CDI first episodes, recurrences, and severity increased over the study period, while mortality decreased. Our CDI first recurrence prediction rule included the following predictor variables: dyslipidemia, CDI type, renal disease, hospital LOS <7 days, principal CDI diagnosis, concomitant gastric acid suppressors, and concomitant antibiotics. This model demonstrated moderate 60-day first recurrence discrimination (AUROC=0.62). Our CDI second recurrence prediction rule was similar in predictor variables and validity. In conclusion, CDI is an important, rapidly-emerging public health problem in the VHA. A clinical prediction rule might aid clinicians in directing preventative therapies to patients at high risk for CDI recurrence.