Benchmarking Inpatient Antimicrobial Use: A Comparison of Risk-Adjusted Observed-to-Expected Ratios. (24th April 2018)

• Record Type: Journal Article
• Authors: Yu, Kalvin C; Moisan, Elizabeth; Tartof, Sara Y; Nguyen, Hien M; Rieg, Gunter; Ramaprasad, Charulata; Jones, Jason
• Is Part Of: Clinical infectious diseases. Volume 67:Number 11(2018)
• Page Start: 1677
• Page End: 1685
• Publication Date: 2018-04-24
• Subjects: antimicrobial stewardship -- antibiotic benchmarking -- drug resistance -- public health; Communicable diseases -- Periodicals; 616.905
• DOI: 10.1093/cid/ciy354
• Languages: English
• ISSNs: 1058-4838

Abstract

Abstract : In this retrospective cohort of 2.7 million admissions, diagnosis-related groups, infection present on admission, unit, patient class, and history of methicillin-resistant Staphylococcus aureus /vancomycin-resistant Enterococcus were isolated as determinants of expected antimicrobial consumption in the most parsimonious model of observed-to-expected antibiotic use. Abstract: Background: Increasing antibiotic resistance has made benchmarking appropriate inpatient antibiotic use a worldwide priority supported by expert societies and regulatory bodies; however, standard risk adjustment for fair interfacility comparison has been elusive. We describe a risk-adjusted antibiotic exposure ratio that may help facilitate assessment of antimicrobial use. Methods: This was a retrospective cohort study of 2.7 million admissions evaluating a wide array of potential explanatory variables for correlation with expected antibiotic consumption in a 2-step approach using recursive partitioning and Poisson regression. Observed-to-expected ratios of risk-adjusted antibiotic use were calculated. Three models of varying complexity were compared: (1) a complex ratio consisting of all available antibiotic use risk factors in a hierarchical model; (2) a simplified antimicrobial stewardship program (ASP) ratio using common facility and encounter factors in a single-level model; and (3) a facility ratio using only broad hospital characteristics. Results: Diagnosis-related groups, infection present on admission, patient class, and unit type were the major predictors of expected antibiotic use. Aside from a history of gram-positive resistance in the prior 12 months for anti–methicillin-resistant Staphylococcus aureus drugs, additional clinical and comorbid history information did not improve the model. The simplified ASP ratio demonstrated higher Pearson correlation ( R 2 = 0.97–0.99) to the complex ratio than the facility ratio ( R 2 = 0.57–0.85) and provided clinical explanations when discordant. Conclusions: The simplified ASP ratio is derived from a parsimonious model that incorporates disease burden through patient-level risk adjustment and better informs stewardship assessment. This may allow for improved comparison of antibiotic use between healthcare facilities.