Predictors of Total Antibiotic Use among a National Network of Academic Hospitals

Background The Centers for Disease Control and Prevention National Healthcare Safety Network (NHSN) provides hospitals a mechanism to report antibiotic use (AU) data to benchmark against peer institutions and direct antibiotic stewardship efforts. Differences in patient populations need to be adjusted for to ensure unbiased comparisons across hospitals. Our objective was to identify predictors of total AU across a nationwide network of hospitals. Methods Data from 126 academic hospitals were extracted from the Vizient Clinical Data Base Resource Manager for adult inpatients (age ≥ 18 years) in 2015. AU was expressed as total antibiotic days of therapy/patient-days. We constructed a negative binomial regression model to explore potential predictors of AU including age, race, sex, case mix index, hospital bed size, length of stay, geographic region, transfer cases, service line, and illness severity. A backwards stepwise approach based on likelihood ratio test was used to identify significant (P < 0.05) predictors and construct the final, parsimonious model. We calculated dispersion-based R2 to assess the percent variability explained by the full and final models. Results A total of 3,076,394 total admissions, representing 17,544,763 patient days, were included. Factors identified as significant predictors in the final model are shown in the Table. The percent variance explained by the full and final models was 90.3% and 89.6%, respectively. Table: Independent predictors of total facility antibiotic use per patient days Relative Risk 95% Confidence Interval Case Mix Index 1.36 1.16, 1.60 Region West Ref – Midwest 1.05 0.92, 1.20 Northeast 0.92 0.81, 1.04 South 1.07 0.94, 1.23 Transfer cases 0.31 0.15, 0.63 Surgery service line 0.45 0.25, 0.81 Major illness severity 3.24 1.04, 10.09 Conclusion The current NHSN AU risk adjustment metric, the standardized antimicrobial administration ratio (SAAR), has been developed separately for different antibiotic groupings and adjusts for a limited set of facility characteristics. Further work is needed to assess if the independent predictors identified in this model can improve upon the performance of existing SAAR metrics and aid in directing stewardship strategies. Disclosures All authors: No reported disclosures.