Methods: We used adult discharge and antibiotic use data from 552 acute care hospitals participating in the Truven Health MarketScan Hospital Database from January 1, 2006, to December 31, 2012 to determine facility-level CDI rates and AU. HO-CDI was defined as a discharge with a secondary ICD-9-CM diagnosis code for CDI (008.45) and inpatient treatment with metronidazole or oral vancomycin. The relationship between facility-level HO-CDI (HO-CDI per 10,000 patient days (PD)) and AU (days of therapy (DOT) per 1,000 PD) was examined through multivariate general estimating equation models that accounted for the correlation between annual HO-CDI rates within a hospital. The models controlled for hospital characteristics and a facility-level rate of community-onset CDI (CO-CDI), defined as a discharge with a primary ICD-9-CM code for CDI and inpatient treatment.
Results: During 2006 to 2012, the mean HO-CDI rate was 11 per 10,000 PD (interquartile range (IQR): 5.7-14.7) and mean AU was 811 DOT/1,000 PD (IQR: 710-932). After controlling for facility-level CO-CDI and other hospital characteristics, overall AU was significantly associated with facility-level HO-CDI rate; for every 50 DOT/1,000 PD increase in AU, there was a 4.4% increase in HO-CDI rate. Similarly, the only antibiotic classes significantly associated with HO-CDI were third- and fourth-generation cephalosporins (P<0.0001) and carbapenems (P=0.0011) with respective increases of 2.1% and 2.4% of HO-CDI per 10 DOT/1,000 PD increase. Fluoroquinolones and β-lactam/β-lactamase inhibitor combinations were not significantly associated with HO-CDI.
Conclusion: In this ecologic analysis of over 500 hospitals, overall antibiotic use was associated with increased rates of HO-CDI. In contrast to recent patient-level analyses in the United States and national observations in England, only third- and fourth-generation cephalosporins and carbapenems were associated with HO-CDI.
L. C. McDonald, None
S. Yi, None
K. Hatfield, None
A. Guh, None
S. Reddy, None
J. A. Jernigan, None