Examining an Impact of Peer Comparison Intervention on Vancomycin Utilization Using Interrupted Time Series Model, Negative Binomial Random Intercept Model, and Linear Mixed Model with AR(1) Covariance Structure Open Access

Yoo, Young Moo (Spring 2018)

Permanent URL: https://etd.library.emory.edu/concern/etds/ns064606n?locale=en



Introduction: Irresponsible use of antibiotics has led to concern about antibiotic resistance. The Antibiotic Stewardship Program (ASP) was introduced nationally to control the antibiotic usage. At Grady Health System, an intervention was implemented by sending out an email to the internal medicine physicians with the report of their historical and current prescription of intravenous vancomycin. The objective of this quality improvement project was to evaluate whether the report card intervention was associated with the provider’s vancomycin prescription behavior. This thesis examines various modeling strategies to assess this association.


Methods: As of December 1, 2016, the intervention was implemented by sending out the bi-weekly report on physicians’ intravenous vancomycin use. In this analysis, the outcome measure is the days of therapy per 100 patient days, which is the vancomycin use rate. An Interrupted Time Series model, Negative Binomial repeated measures model with offset term, and linear mixed model with AR(1) covariance structure were fitted and compared to examine the impact of the intervention.


Results: A total of 64 physicians were included in this two-year period quality improvement project. The estimated baseline vancomycin prescription rate was 10.35 (95% confidence interval, 9.22 to 11.47). When physicians started to receive a report card, the vancomycin prescription rate declined by 2.33 (95% confidence interval, -3.79, to 0.87). Throughout the post-intervention year, the rate decreased by 0.13 (95% confidence interval, -0.24, -0.03) every two weeks.


Conclusion: During the historical year (pre-intervention period), no significant temporal trend of the vancomycin prescription rate was measured. However, once the intervention was introduced on December 1, 2016, there was an immediate drop of the vancomycin use rate. During the post-intervention period, constant decline of the vancomycin use rate was captured. These results were consistent over three modeling strategies that provides the audience options to choose based on their research questions. 

Table of Contents


Table of Contents


1. Introduction


2. Methods


2.1 Study Design and Intervention


2.2 Data Collection and Cleaning


2.3 Outcome of Interest


2.4 Statistical Analysis


2.4.1 Interrupted Time Series


2.4.2 Nonlinear mixed model


2.4.3 Linear mixed model with AR(1) covariance structure


3. Result


4. Discussion




Appendix: SAS Programming Code

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Partnering Agencies
Last modified

Primary PDF

Supplemental Files