2147. Sample Size Estimates for Cluster Randomized Trials in Infection Control and Antimicrobial Stewardship

BACKGROUND: Cluster randomized control trials (CRCTs) are used frequently in the field of infection control and antimicrobial stewardship because randomization at the patient level is often not feasible due to contamination, ethical, or logistical issues. The correlation and thus non-independence that exists among individual patients in a cluster must be accounted for when estimating sample size for such trials, yet many studies neglect to consider or report the intracluster correlation coefficient (ICC) and the resulting coefficient of variation (CV) in rates between hospitals. The aim of this study was to estimate the sample sizes needed to adequately power studies of hospital-level interventions to reduce rates of healthcare-associated infections. METHODS: We calculated the minimum number of clusters or hospitals that would need to be included in a study to have good power to detect an impact of the intervention given a range of different assumptions. We estimated parameters needed for these calculations using national rates from the National Healthcare Safety Network (NHSN) for methicillin-resistant Staphylococcus aureus (MRSA) bacteremia, central-line associated bloodstream infections (CLABSI), catheter-associated urinary tract infections (CAUTI), C. difficile infections (CDI) and variation between hospitals in these rates. These calculations were based on the assumption that hospitals were uniform and moderate in size and were studied for 1 year. RESULTS: To study an intervention leading to a 50% decrease in daily rates and using the C vs. calculated from NHSN, 22 average-sized hospitals for MRSA bacteremia are needed, 34 for CAUTI, 9 for CDI, and 27 for CLABSI to have a statistically significant decrease with a type I error rate of 0.05 and a type II error rate of 0.8. If a 10% decrease in rates is expected instead, 709, 1205, 279, and 866 hospitals, respectively, are needed. CONCLUSION: Sample size estimates for CRCTs are most influenced by the CV and the expected effect size. Given the large sample size requirements, it is likely that many CRCTs in hospital epidemiology are under-powered. We hope that these findings lead to more definitive CRCTs in the field of hospital epidemiology that are properly powered and more studies reporting their ICC or CV. DISCLOSURES: All authors: No reported disclosures.