DOI
https://doi.org/10.25772/XBK8-MS40
Defense Date
2012
Document Type
Dissertation
Degree Name
Doctor of Philosophy
Department
Pharmaceutical Sciences
First Advisor
Polk Ronald
Abstract
Background. Infectious diseases societies recommend that hospitals risk-adjust their antimicrobial use before comparing it to their peers, a process called benchmarking. The purpose of this investigation is to apply and compare 3 risk-adjustment procedures for benchmarking hospital antibacterial consumption (AbC). Two standardization of rates procedures, direct and indirect standardization, are compared with one another as well as with regression modeling. Methods. Total aggregate adult AbC for 52 systemic antibacterial agents was measured in 70 hospitals that subscribed to the University HealthSystem Consortium Clinical Resource Manager database in 2009 and expressed as days of therapy (DOTs) per either 1000 patients days (PDs) or 1000 discharges. The two AbC rates served the role of the outcome while several known risk factors for AbC served the role of potential predictor variables in the linear regression models. Selection criteria were applied to select a model that represented the first rate (Model I) and another that represented the second (Model II), respectively, and outliers were identified. Adult discharges in each hospital were then stratified into 35 clinical service lines based upon their Medicare Severity-Diagnosis Related Group (MS-DRG) assignment. Direct and indirect standardization were applied to this set and the expected-to-observed (E/O) and observed-to-expected (O/E) ratios, respectively, for AbC were determined. The agreement of the different methods in ranking hospitals according to their risk-adjusted rates and in identifying outliers was determined. Results. The mean total AbC rate was 821.2 DOTs/1000 PDs or 4487.6 DOTs/1000 discharges. Model I explained 31% of the variability in AbC measured in DOTs/1000 PDs while Model II explained 64% of the variability in AbC measured in DOTs/1000 discharges. The E/O ratios ranged from 0.76-1.44 while the O/E ratios ranged from 0.73-1.45. The comparison of the risk-adjustment methods revealed a very good agreement between the two regression models as well as between the two standardization methods whereas the agreement of Model II with either standardization method was moderate. Conclusion. Standardization provides a viable alternative to regression for benchmarking hospital AbC rates. Direct standardization appears to be especially useful for benchmarking purposes since it allows the direct comparison of risk-adjusted rates.
Rights
© The Author
Is Part Of
VCU University Archives
Is Part Of
VCU Theses and Dissertations
Date of Submission
August 2012