Clostridium difficile classification overestimates hospital acquired infections

BACKGROUND Clostridium difficile infections are common among hospitalised patients, with some infections acquired in hospital and others in the community. International guidelines classify cases as hospital-acquired if symptom onset occurs >2 days after admission. This classification informs surveillance and infection control, but has not been verified by empirical or modelling studies. AIMS To assess current classification of C. difficile acquisition using a simulation model as a gold standard. METHODS We simulated C. difficile transmission in a range of hospital scenarios. We calculated the sensitivity, specificity and precision of classifications that use cut-offs ranging from 0.25 hours to 40 days. We identified the optimal cut-off that correctly estimated the proportion of cases that were hospital acquired and the balanced cut-off that had equal sensitivity and specificity. FINDINGS The recommended two-day cut-off overestimated the incidence of hospital-acquired cases in all scenarios and by >100% in the base scenario. The two-day cut-off had good sensitivity (96%) but poor specificity (48%) and precision (52%) to identify cases acquired during the current hospitalisation. A five-day cut-off was balanced and a six-day cut-off was optimal in the base scenario. The optimal and balanced cut-offs were more than two days for nearly all scenarios considered (ranges four to nine days and two to eight days). CONCLUSIONS Current guidelines for classifying C. difficile infections overestimate the proportion of cases acquired in hospital in all model scenarios. To reduce misclassification bias, an infection should be classified as being acquired prior to admission if symptoms begin within five days of admission