Multicentre derivation and validation of a simple predictive index for healthcare-associated Clostridium difficile infection

Objectives: Clostridium difficile infection (CDI) is the most common cause of healthcare-associated infections in the United States. Despite well-established risk factors, little research has focused on use of these variables to identify a patient population at high risk for CDI to target with primary prevention strategies. A predictive index for healthcare-associated CDI could improve clinical care and guide research for primary prevention trials. Our objective was to develop a predictive index to identify patients at high risk for healthcare-associated CDI. Methods: We performed a secondary database analysis in a five-hospital health system in Houston, Texas. Our cohort consisted of 97 130 hospitalized patients admitted for more than 48 hours between October 2014 and September 2016. The derivation cohort consisted of the initial 80% of admissions (75 545 patients), with the remainder being used in the validation cohort. Results: CDI rates in the derivation and validation cohorts were 1.55% and 1.43%, respectively. Thirty-day predictors of CDI were increased number of high-risk antibiotics, Charlson comorbidity index score, age and receipt of a proton pump inhibitor. These variables were incorporated into a simple risk index with a score range of 0 to 10. The final model demonstrated good discrimination and calibration with the observed CDI incidence ranging from 0.1% to 20.4%. Conclusions: We developed a predictive index for 30-day risk of healthcare-associated CDI using readily available and clinically useful variables. This simple predictive risk index may be used to improve clinical decision making and resource allocation for CDI stewardship initiatives, and guide research design.