Vital Signs

Trends in Staphylococcus Aureus Infections in Veterans Affairs Medical Centers — United States, 2005–2017

Makoto Jones, MD; John A. Jernigan, MD; Martin E. Evans, MD; Gary A. Roselle, MD; Kelly M. Hatfield, MSPH; Matthew H. Samore, MD


Morbidity and Mortality Weekly Report. 2019;68(9):220-224. 

In This Article


Clinical data from any patient admitted to VAMCs in the United States from January 1, 2005 through December 31, 2017 were analyzed. Facilities were excluded from the study if they did not provide acute care or if they did not report data to VA's periodic complexity assessment (e.g., the level and type of care provided)[2] in all eligible years during the study period. Clinical diagnostic culture and MRSA surveillance test results were extracted from electronic health records as described elsewhere.[3] Bloodstream infections were defined as isolation of S. aureus from blood samples. Nonblood infections were defined as isolation of S. aureus from any other sample type, excluding those obtained for surveillance purposes and those obtained within 14 days of a positive blood culture. MRSA isolates from samples collected from the same patient within 365 days were considered duplicates and excluded; MSSA duplicates were defined in the same manner. Infections were classified as hospital-onset when the specimen was obtained >3 days after admission, and as community-onset when the specimen was obtained ≤3 days after admission. Community-onset infection rates and total (combined community-onset and hospital-onset) infection rates were calculated per admission. Hospital-onset infection rates were expressed per 1,000 patient-days-at-risk, excluding days after the patient had met one of the infection definitions. MRSA colonization status at admission was considered positive if the last test within 24 hours after admission was positive. Patients were considered to have acquired MRSA if they had at least one MRSA-positive test (clinical or surveillance) after a negative admission surveillance test. Fluoroquinolone use was measured and defined according to National Healthcare Safety Network methods to assess for potential changes in antimicrobial pressure exerted on S. aureus.[4]

To model rates, trend analyses were performed with generalized estimating equation models clustering by VAMC and using a negative binomial distribution, patient days at risk as the exposure, an autoregressive correlation structure, and robust error estimation. Models were adjusted for major hospital characteristics, including Medicare Relative Risk score, patient volume, resident slots, intensive care unit level, and number of advanced specialty clinical programs.[2] Proportions were modeled similarly but with a binomial distribution. All percentage changes are based on modeled rates. Statistical analyses were performed using Stata statistical software (release 15; StataCorp, LLC). This study was performed with approval from the University of Utah Institutional Review Board and the VA Salt Lake City Health Care System Research and Development Office.