Staphylococcus aureus Infections in US Veterans, Maryland, USA, 1999–20081

Trends in Staphylococcus aureus infections are not well described. To calculate incidence in overall S. aureus infection and invasive and noninvasive infections according to methicillin susceptibility and location, we conducted a 10-year population-based retrospective cohort study (1999-2008) using patient-level data in the Veterans Affairs Maryland Health Care System. We found 3,674 S. aureus infections: 2,816 (77%) were noninvasive; 2,256 (61%) were methicillin-resistant S. aureus (MRSA); 2,517 (69%) were community onset, and 1,157 (31%) were hospital onset. Sixty-one percent of noninvasive infections were skin and soft tissue infections; 1,112 (65%) of these were MRSA. Ten-year averaged incidence per 100,000 veterans was 749 (± 132 SD, range 549-954) overall, 178 (± 41 SD, range 114-259) invasive, and 571 (± 152 SD, range 364-801) noninvasive S. aureus infections. Incidence of all S. aureus infections significantly increased (p<0.001), driven by noninvasive, MRSA, and community-onset infections (p<0.001); incidence of invasive S. aureus infection significantly decreased (p<0.001).

S taphylococcus aureus exists as a commensal organism living on the human body in equilibrium with other bacteria and as a common agent associated with a spectrum of diseases ranging from mild, noninvasive skin and soft tissue infections (SSTIs) to invasive, life-threatening bloodstream infections. Increasing incidence of infections caused by methicillin-resistant S. aureus (MRSA) has complicated treatment of S. aureus infection. Previously MRSA infections were problematic primarily among hospitalized persons or persons exposed to the health care settings. However, since the 1990s, MRSA infections have become more prevalent in healthy, younger persons who have little to no exposure to health care settings. Of particular concern is the rapid increase in MRSA SSTIs reportedly driven by emergence of a new MRSA strain, USA300 (1,2).
Despite these changes, the epidemiology of S. aureus infection, particularly the total effect of infection in the United States, is not well described. Several populationbased studies on S. aureus infections exist; however, these studies focused on hospital-based populations (3)(4)(5)(6), MRSA infection (7)(8)(9), non-US populations (10)(11)(12), or only estimated the impact of invasive S. aureus disease (10,(13)(14)(15)). Additionally, population-level changes in incidence, particularly before and after USA300 MRSA emerged, are largely unknown. To describe overall trends and recent changes in the incidence of S. aureus infection while differentiating between invasive and noninvasive, community-and hospital-onset, and methicillin-susceptible and -resistant S. aureus infections, we conducted a retrospective population-based study.

Data Source
Our study used data from the Veterans Affairs Maryland Health Care System (VAMHCS) over a 10-year period (1999-2008). VAMHCS, a large, integrated health care system, comprises 3 medical centers (Baltimore VA Medical Center, Perry Point VA Medical Center, and the Baltimore VA Rehabilitation and Extended Care Center), ≈730 inpatient beds, and 5 community-based outpatient clinics. VAMHCS uses an electronic health information system known as the Veterans Health Information Systems and Technology Architecture (VistA). This system is used to collect and maintain all health information at each VA medical facility, including the VAMHCS. VAMHCS's electronic medical records and administrative data, including codes from the International Classifi cation of Diseases, 9th Revision, Clinical Modifi cation (ICD-9-CM), and microbiology culture results collected from the VistA served as the primary data source for this study. The VAMHCS Research and Development Committee and the University of Maryland, Baltimore Institutional Review Board approved this study. Given that data were retrospectively obtained from VistA, informed consent requirements were waived.

S. aureus Culture Collection and Classifi cation
We collected data during October 1, 1998-September 30, 2008, on all S. aureus-positive blood and clinical cultures (excluding surveillance and fecal and nasal cultures) identifi ed through microbiologic accession number, date, time, and specimen type (e.g., blood, skin). Each positive culture was classifi ed as originating from a sterile or a nonsterile body site. We defi ned a sterile body site according to Centers for Disease Control and Prevention Active Bacterial Core surveillance criteria, and all other body sites were classifi ed as nonsterile (13). We defi ned a unique culture as the fi rst S. aureus-positive culture obtained from a patient during a 6-month period. If cultures were obtained from a sterile and a nonsterile site from the patient during the same period, we chose the culture from the sterile site.
For S. aureus-positive cultures obtained during an outpatient visit for which the patient was not subsequently hospitalized within a 72-hour period after culture, we obtained all ICD-9-CM codes associated with all of the patient's outpatient visits on the day of culture. For S. aureus-positive cultures obtained during an outpatient visit for which the patient was subsequently hospitalized within a 72-hour period after culture, we collected the outpatient and hospital discharge ICD-9-CM codes. Finally, we obtained all ICD-9-CM discharge codes for S. aureus-positive cultures obtained during a hospitalization. Using information from previous studies, we developed a comprehensive list of ICD-9-CM codes for S. aureus-related infections and categorized them by the site of infection most consistent with the associated code (16).
We determined an invasive S. aureus infection on the basis of S. aureus isolation from a clinical or blood culture from a normally sterile body site, such as blood, cerebrospinal fl uid, pleural fl uid, pericardial fl uid, peritoneal fl uid, joint/synovial fl uid, bone, internal body site, and muscle. Identifi cation of noninvasive S. aureus infection was based on isolation of S. aureus from a clinical culture of a nonsterile site, without a concurrent culture from a sterile site obtained during the same 6-month period, and at least 1 ICD-9-CM code for S. aureus-related infection from the outpatient visit or hospitalization associated with the positive culture. For a noninvasive infection, we based the requirement of a matching ICD-9-CM code along with positive culture on concerns that S. aureus obtained from a sample from a nonsterile body site can represent either infection or colonization. In a substudy that used a random sample of cases, we estimated that an ICD-9-CM code for S. aureus-related infection plus positive clinical culture (from a nonsterile site) increases the probability of a true noninvasive S. aureus infection by ≈23.8% over positive clinical culture alone (16).
Positive cultures obtained after the fi rst 48 hours of hospitalization, rehabilitation, or long-term stay were classifi ed as hospital onset and all others as community onset. All S. aureus infections were classifi ed according to methicillin susceptibility on the basis of in vitro susceptibility to oxacillin. All MRSA infections were grouped into the following epidemiologic categories: health care-associated community onset, defi ned as cases in persons with at least 1 listed risk factor in the past 12 months; health care-associated hospital onset, defi ned as cases in persons who have a positive culture within 48 hours after hospitalization; or community-associated, defi ned as cases in persons with no documented health careassociated community-onset risk factor (13). To determine a risk factor for health care-associated community-onset infection, we obtained history of hospitalization, surgery, residence in a long-term care facility, or prior MRSApositive culture in the past 12 months before the date and time of each index positive MRSA culture. Site of infection (bone or joint, skin or soft tissue, endovascular, respiratory, intraabdominal/pelvic, central nervous system, urinary tract, S. aureus-nonspecifi c site, bacteremia without focus, and other or site not specifi ed) was determined for each S. aureus infection on the basis of matching ICD-9-CM code. When cultures matched with multiple ICD-9-CM codes, we chose the highest ranking site of infection on the basis of the likelihood that a culture represented true S. aureus infection (16).

Statistical Analysis
Information about annual number of unique veterans, admissions, and total inpatient days for hospitalization, long-term care, and residential rehabilitation programs were obtained from the VAMHCS Medical Administrative Service fi scal year (FY) databases. Annual and 10-year averaged incidence rates per 100,000 veterans were estimated overall and for community-onset infections and per 100,000 inpatient days for hospital-onset S. aureus infections. To account for the fi rst 48 hours in the defi nition for hospital-onset infection, we adjusted inpatient days (number of annual inpatient days minus 2× the number of annual admissions) (13). Because the average inpatient length of stay in the VAMHCS is >48 hours, this adjustment provides a consistent and reasonably accurate estimate of inpatient days past 48 hours.
Trends in all S. aureus infections were initially assessed by plotting natural and cubic spline smoothers to the observed data plotted as a function of time. The formal analysis of trends of all S. aureus infections was based on generalized linear models, assuming a Poisson distribution with a log link function, including FY as a predictor variable and log total number of unique veterans or logadjusted inpatient days as an offset variable (17). Model fi t was assessed by evaluating the deviance and Akaike information criterion, and regression coeffi cients for trend were assessed by the partial Wald test (18,19). Additional models were fi t for each stratum of interest, i.e., invasive, noninvasive, onset, and methicillin susceptibility. Analyses were performed by using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA) and R version 2.7.0 (2008) software (www.r-project.org).
The incidence of invasive hospital-onset S. aureus infections decreased from 150 to 85 cases per 100,000 adjusted acute-care inpatient days (Table 2). Invasive hospital-onset MSSA and MRSA infections signifi cantly decreased during the 10-year period (p = 0.01 and p<0.001, respectively). The incidence of invasive community-onset S. aureus infections decreased 1.9-fold from 119 to 64 per 100,000 veterans and was driven by decreases in incidence of invasive community-onset MSSA. Incidence of invasive community-onset MRSA remained relatively unchanged (46 to 41/100,000 veterans).
The incidence of noninvasive hospital-onset infections did not follow any apparent increasing or decreasing trend (p = 0.08; Table 2). However, driven by noninvasive MRSA infections, the overall incidence of noninvasive communityonset S. aureus infections increased signifi cantly (p<0.001). Incidence after FY 2003 rapidly increased from 218 to 546, peaking at 644 cases per 100,000 veterans in FY 2005. After FY 2000, incidence of noninvasive community-onset MRSA infections increased 4-fold from 100 to 397 cases per 100,000 veterans. Incidence of noninvasive communityonset MSSA infections did not change (p = 0.83; Table 2).
A total of 1,703 (61%) noninvasive infections were classifi ed as SSTIs of which 1,112 (65%) were caused by MRSA. Changes in incidence of overall noninvasive MRSA infections were driven by increases in MRSA SSTIs ( Figure 2). Incidence per 100,000 veterans signifi cantly increased from 90 cases in FY 1999 to 345 in FY 2008 (p<0.0001); the largest increase began in FY 2003 and incidence peaked at 440 cases in FY 2005.

Discussion
We have described the incidence of all S. aureus infections during a 10-year period in a large US-based   (20,21). However, attributing the decrease to a single practice is diffi cult, if not impossible, and few published studies exist with which we can compare our invasive S. aureus results.
Laupland et al. estimated an annual incidence of invasive S. aureus of 28 cases per 100,000 population, but their results may not be comparable because they were based primarily on MSSA infections (12). Klevens et al. estimated an incidence of invasive MRSA infection of 32 cases per 100,000 population, 75% were bacteremias, and 27% were hospital-acquired on the basis of 2005 data from 9 US cities (13). Their study reported a noticeably higher incidence of invasive MRSA in Baltimore, Maryland, USA, of 117 per 100,000 compared with estimates of 20-50 per 100,000 population at other sites. Our estimated rate of invasive MRSA infection (160 cases/100,000 veterans) is commensurate during the same year (2005), and our calculated proportion of invasive hospital-onset MRSA   (14). However, few MRSA cases were identifi ed in this study, thereby making it diffi cult to compare with our results for which MRSA caused almost half of infections. For several reasons, our incidence rates of invasive infections are higher than those previously reported. Black race has previously been reported as a marker for increased risk for invasive MRSA infections (13,22). Therefore, we would expect a higher incidence of all S. aureus infections, given that in our study 49% of patients were black. Additionally, given that our study was performed on data from a primarily urban population located in or around Baltimore, we could attribute the higher incidence of infections to suspected risk factors that are more prevalent in this location, including intravenous drug use.
Our study identifi ed increases in noninvasive S. aureus infections, particularly around 2003, which most likely are associated with the emergence of the USA300 MRSA clone that has led to increases in community-associated MRSA, specifi cally in SSTIs (1,22). Also, despite dramatic increases in noninvasive community-onset MRSA infections, we did not observe a proportionate increase in invasive community-onset MRSA as might be expected if USA300 MRSA had the same propensity as non-USA300 MRSA to invade the bloodstream. No population-based studies have been published with which to compare overall noninvasive S. aureus infections, and few exist for comparison of noninvasive MRSA infections. For instance, Liu et al. recently reported annual (2004Liu et al. recently reported annual ( -2005 incidence rates of community-acquired and hospital-acquired MRSA among residents in San Francisco, California, USA, community-associated of 316 and 31 cases per 100,000 population, respectively, for which most cultures were from skin and soft tissue (8). The results of our study for community-onset MRSA are slightly lower but similar. Crum et al. reported a dramatic increase during 2002-2004 in community-associated MRSA infections, of which most were classifi ed as SSTIs, and an incidence rate of 155 cases per 100,000 persons from 2004 data (7). The FY 2004 incidence of noninvasive community-associated MRSA in our study was 188 cases per 100,000 veterans, which is similar to that reported by Crum et al.
Our study adds new information to the existing literature and has several strengths. A major strength is its calculation of annual incidence of all S. aureus infections for a 10-year period by using actual numbers of patients at risk, admissions, and inpatient days. Our estimates of incidence are more accurate than those in previous studies, which were based on census-level data (11)(12)(13)23). Noninvasive infections were identifi ed by using an automated approach that required both positive clinical culture and confi rmed ICD-9-CM code for infection. This defi nition is more rigorous, thereby producing higher positive predictive values than clinical cultures alone, particularly for infections of bone and skin or soft tissue (16). In addition, this automated approach enabled us to identify and classify types of S. aureus infection, which is useful for understanding the overall population distribution of infection. Access to comprehensive, patientlevel information, including prior hospitalizations, prior MRSA infections, and surgeries, allowed us determine the epidemiologic class for all MRSA infections. This study was performed in a population receiving standardized health care; therefore, fi ndings should be free of bias associated with access to care or duration and type of treatment received.
Our study also has several limitations. First, the VAMHCS population of adult, mostly male patients living in the mid-Atlantic region does not fully represent the overall US population. We are unable to extrapolate these fi ndings to children, a population for which an increase of community-onset MRSA SSTIs has been reported (24). Previous reports suggest that men are at higher risk than women for S. aureus infections; therefore, our estimates may overestimate true rates for women (8,12,13). Second, although we did not perform molecular typing on the S. aureus-positive isolates, we expect that a signifi cant proportion (>80%) of noninvasive MRSA infections were caused by the USA300 MRSA strain (1). Given  incidence because our defi nition of S. aureus infection required a positive clinical or blood culture. However, given the standardized access to care in the VAMHCS, we expect that cultures were uniformly collected in patients who had clinical signs or symptoms of infection. Fourth, the clinical culturing rate may have increased during our study period, which would contribute to overestimates in the incidence of S. aureus infection, particularly noninvasive infections. Johnston et al. observed an increase in the absolute number of SSTI cultures obtained in the VAMHCS Emergency Care Service (1). However, they determined that the proportion of MRSA infections increased, even though the proportion of MSSA remained the same, which suggests a true increase in MRSA. We observed similar patterns: whereas MRSA infections increased, MSSA infections remained relatively stable, and the annual number of S. aureus cultures did not signifi cantly change.
In conclusion, this large, population-based study demonstrated an increase in the overall incidence of S. aureus infections during FY 1999-FY 2008, which was driven by a rapid increase in noninvasive, communityonset, MRSA skin and soft tissue infections. This increase was most striking during and after 2003, which is coincident with the time during which the USA300 clone became a major contributor to noninvasive S. aureus infections. Despite this increase, incidence of invasive community-onset MRSA infections did not signifi cantly increase, and the overall MSSA infections and noninvasive MSSA infections remained generally stable. These results suggest a shift in the distribution of S. aureus infections to more noninvasive community-onset MRSA infections. This information is useful for interpreting changes in the epidemiology of S. aureus infections, which may help guide additional prevention strategies focused on reducing community-onset S. aureus infections. To further understand these trends, additional studies are warranted to identify risk factors for S. aureus infection and to describe the epidemiology of S. aureus infections across the entire population.