Objectives: We tested a previously published model for the analysis of the temporal relationship between antibiotic use and the incidence of Clostridium difficile infection in a hospital with stable incidence of infection at >1 case per 1000 admissions per month.
Methods: The study period was from April 2004 to June 2008 and used data from Infection Control and Hospital Pharmacy. We first described the monthly variation in C. difficile infection and then constructed a multivariate transfer function model that included lag time (cases of C. difficile infection in previous months and delays between changes in antibiotic use and changes in C. difficile infection).
Results: The average incidence of C. difficile infection was 1.5 cases per 1000 patients per month with no significant increase over 3 years. The number of cases of C. difficile infection in 1 month was dependent on the average number of cases of C. difficile infection in the previous 2 months. The models with data from the whole hospital showed a statistically significant relationship between the number of both hospital-acquired C. difficile infections and total C. difficile infections and consumption of piperacillin/tazobactam, ciprofloxacin and cefuroxime. The association between C. difficile infection and consumption of co-amoxiclav was only significant for hospital-acquired C. difficile infection. The model for hospital-acquired C. difficile infections explained 61% of the variance in C. difficile infections.
Conclusions: These results provide support for antibiotic policies that minimize the use of broad-spectrum penicillins (co-amoxiclav and piperacillin/tazobactam), cephalosporins and fluoroquinolones.