We used data from the 2001–2004 National Health and Nutrition Examination Survey (NHANES) and the NHANES 2001–2004 Mortality-Linked File. NHANES is a survey of the US noninstitutionalized civilian population conducted by the Centers for Disease Control and Prevention’s National Center for Health Statistics. We selected 10,598 NHANES participants using a complex multistage sampling design and followed them through December 31, 2011, to track mortality. NHANES protocols were approved by the institutional review boards of NCHS and CDC, and all participants provided informed consent.

Nasal swabs collected from participants were tested for S. aureus and methicillin resistance during 2001–2004 following the National Committee for Clinical Laboratory Standards disk-diffusion method. We additionally tested isolates resistant to oxacillin (MRSA) and intermediately resistant to oxacillin and every tenth isolate sensitive to oxacillin (MSSA) for staphylococcal cassette chromosome mec (SCCmec) typing, genes encoding enterotoxins, toxic shock syndrome toxin-1, and Panton-Valentine leukocidin (PVL) toxin (details on the methods are available at http://www.cdc.gov/nchs/data/nhanes/nhanes_01_02/l35_b_doc.pdf). We collected data on age, sex, race/ethnicity, family income, smoking, and medical conditions using questionnaires.

Univariate and multivariate Cox proportional hazards regressions were used to estimate the hazard ratio (HR) with corresponding 95% CIs for MSSA and MRSA association with mortality. Analyses were performed in STATA version 11 (StataCorp LP, College Station, TX, USA), taking into account the survey design and sampling weights, to produce nationally representative estimates. We considered p values <0.05 as statistically significant.

A total of 27.3% of participants were colonized with MSSA and 1.3% with MRSA (Table 1). During an average of 8.5 years of follow-up, the total death rate per 1,000 person-years was 10.3: 10.8 for noncolonized participants, 8.8 for those colonized with MSSA, and 25.6%for those colonized with MRSA. MRSA-colonized participants tended to be older and non-Hispanic white, have low poverty income ratio, smoke cigarettes, have co-morbidities, or have been be hospitalized and/or have had dialysis in the previous 12 months (Table 1). Approximately 58.6% of MRSA isolates were of SCCmec type II, suggestive of hospital-associated MRSA, and 41.4% had MRSA isolates of SCCmec type IV, indicative of community-associated MRSA.

In Cox regression, MSSA-colonized participants had a significantly lower risk for death than did noncolonized participants in univariate analysis (hazard ratio [HR] 0.82, 95% CI 0.68–0.98) but not after adjustment for co-variates. MRSA colonization was associated with higher risk for death than no S. aureus colonization in univariate analysis (HR 2.40, 95% CI 1.56–3.69) and after adjustment for sociodemographic factors (HR 1.54, 95% CI 1.06–2.34); MRSA colonization also carried higher risk for death than MSSA colonization (HR 2.90, 95% CI 1.92–4.40). However, after additional adjustment for smoking, body mass index, and co-morbidities (asthma, diabetes, hypertension, cardiovascular diseases, stroke, liver disease, and cancer), these associations were no longer significant (Table 2).

Likewise, colonization with SCCmec type II MRSA more strongly predicted long-term mortality than did no S. aureus ((HR 3.38, 95% CI 2.00–5.72) or MSSA colonization (HR 4.05, 95% CI 2.42–6.79) in univariate analyses but not after adjustment for sociodemographic characteristics. Colonization with SCCmec type IV MRSA was not associated with death in univariate or adjusted analysis, unlike no S. aureus or MSSA colonization (Table 2). However, the small number of participants with MRSA SCCmec types II (83) and IV (53) limited our analysis.

In this population-based study, we found that MSSA is not associated with long-term mortality. MRSA is associated with risk for death in crude analysis but not after adjustment for socioeconomic factors, co-morbidities, or both.

Regarding MSSA, our results are consistent with studies conducted in inpatient and nursing home participants that also failed to find an association (4–7). Many of these studies were underpowered, calling for larger scale investigations. In line with the inverse association in our unadjusted analysis for MSSA and mortality, Wertheim et al. found that MSSA-colonized participants had longer survival than did noncolonized participants among patients with S. aureus bacteremia (3). They suggested that colonized patients might be more immunologically adapted to the endogenous S. aureus strains that caused the bacteremia or that exogenous strains in noncolonized patients were more virulent. In our case, we hypothesized that the negative association was imputable to some confounders, such as race/ethnicity, smoking, and co-morbidities (hypertension, cardiovascular diseases, stroke), instead of an independent association. We found that persons with MSSA were more likely to be non-Hispanic white and have asthma or diabetes than were persons without S. aureus nasal carriage, who were more likely be non-Hispanic black; have a lower poverty income ratio; be a smoker; and have hypertension, cardiovascular diseases, stroke, liver disease, or cancer. Concerning MRSA, Chan et al. observed that MRSA colonization was associated with higher mortality than non-MRSA colonization in nursing home residents in China followed for 2 years (8). Previous research also found that methicillin resistance in S. aureus bacteremia did not predict long-term mortality after adjustment for age, comorbidities, severity of acute illness, metastatic infections, and long-term facility resident status (1).

Delayed treatment initiation, virulence factors such as PVL, or both have been hypothesized to be responsible for excess mortality in MRSA infections and bacteremia. Subsequent research found that PVL was associated with decreased time from colonization to bacteremia but not to mortality (9). Again in confirmation of our results, Han et al. studied the effect of SCCmec type on the 30-day in-hospital mortality in 184 patients with MRSA and found SCCmec II significant in univariate analysis, but not after co-variates adjustment (10). That study also concluded that SCCmec II was a marker for disease severity instead of a predictor of mortality.

Our study had several limitations. The carriage of S. aureus could be transient in some cases and not always reflect the carriage pattern over time. Although the nares are the most common site for S. aureus carriage, extranasal sites can also harbor the organism (11). S. aureus carriage was tested in 2001–2004 and might not reflect the current epidemiology and MRSA strain types. Data on co-morbidities, such as cardiovascular or liver diseases and cancer, were not available for participants 18–19 years of age and were coded as missing in the analyses. We did not compare death for the genes encoding enterotoxins because of their limited numbers. However, these NHANES data are the only national data available, and the major strengths of our study include the sample size representative of the US population with generalizable findings and higher power for statistical inferences.

Our findings are of public health and clinical importance because a lack of independent association between colonization by S. aureus and mortality implies that decolonization might not prevent death. Some studies indicate that decolonization of patients at high risk for infections could reduce the number of subsequent S. aureus infections, but evidence remains limited that it prevents death (5).