684. How to use antimicrobial use data? A model to support decision-making and facilitate understanding
Session: Poster Abstract Session: Stewardship: Data and Program Planning
Thursday, October 5, 2017
Room: Poster Hall CD

Background: Antimicrobial use data are increasingly available, yet it is not clear how to use them most effectively. An understanding of how practice decisions influence antimicrobial use may aid individual knowledge development and rational policy planning. We developed a mathematical model to describe antimicrobial use and demonstrate how it could be used in a model-driven decision support system.

Methods: We developed a discrete-time Markov chain model to describe antimicrobial use as a function of the following parameters: Choice decisions to start antibiotics on admission or after, Change decisions to stop antibiotics, and Completion decisions to discharge patients whether they were on or off antimicrobials. Partial derivatives were used to predict the extent to which antimicrobial use would respond to changes in each parameter. We used Veterans Affairs Bar Code Medication Administration data from 2010 to estimate parameters, as well as antimicrobial use using National Healthcare Safety Network (NHSN) definitions. Categories of anti-methicillin-resistant Staphylococcus aureus (MRSA), broad community, broad hospital, and surgical site infection prophylaxis (SSIP) from NHSN were also used. Because of certain assumptions made when estimating parameters, we used non-linear regression to adjust them using data from year 2010. We then applied our model to predict antimicrobial use from 2013 parameters and compared with actual use with Pearson’s correlation coefficient.

Results: Correlation of predicted and actual antimicrobial use was 0.97, 0.99, 0.95, and 0.92 (using NHSN category order above; Figure 1). As a conservative estimate, the correlation of yearly changes between predicted and actual antimicrobial use for all categories was 0.75. For > 99% of all combinations of medical center, antimicrobial category, and year, decreasing the probability of starting antimicrobials had the most impact on measured antimicrobial use.

Conclusion: Our mathematical model is highly predictive of antimicrobial use and can be used to anticipate how much changes in decision points might lead to changes in antimicrobial use. Given the parameter space that most VA medical centers occupy, not starting antimicrobials appears to have greatest impact on use.

Makoto Jones, MD, MS, Department of Medicine, University of Utah School of Medicine, Salt Lake City, UT, Karim Khader, PhD, Ideas Center, VA Salt Lake City Health Care System, Salt Lake City, UT, Benedikt Huttner, MD, MSCI, Geneva University Hospitals, Geneva, Switzerland, Christopher Graber, MD, MPH, FIDSA, VA Greater Los Angeles Healthcare System, Los Angeles, CA, Yue Zhang, PhD, Division of Epidemiology, Department of Medicine, University of Utah, Salt Lake City, UT, Matthew Samore, MD, FSHEA, University of Utah School of Medicine, Division of Epidemiology, Salt Lake City, UT, Karl Madaras-Kelly, PharmD., MPH, Vet. Med. Ctr., Boise, ID, Matthew Goetz, MD, Infectious Diseases, VA Greater Los Angeles Healthcare System, Los Angeles, CA and Peter Glassman, MBBS, MSc, David Geffen School of Medicine at UCLA, Los Angeles, CA


M. Jones, None

K. Khader, None

B. Huttner, None

C. Graber, None

Y. Zhang, None

M. Samore, None

K. Madaras-Kelly, None

M. Goetz, None

P. Glassman, None

Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 4th with the exception of research findings presented at the IDWeek press conferences.