Methicillin-resistant Staphylococcus aureus (MRSA) has long been recognized as a serious pathogen affecting patients in hospitals¹. Following initial recognition of MRSA as an important source of community-acquired infections^2,3, surveillance data indicate that incidence of hospital-onset MRSA bloodstream infections has declined by 74% in the United States⁴.

Healthcare in the United States has also evolved. The average length of a hospital admission shortened during the 1990s and 2000s^5,6, and the proportion of hospitalizations resulting in discharge to post-acute-care facilities such as skilled nursing facilities for continued care has increased over time⁷. Several studies have identified a high prevalence of MRSA colonization in nursing home residents. A review reported that 8-10% of all U.S. nursing home residents are colonized with MRSA⁸. However, some studies have found colonization prevalence among residents exceeding 40% in certain facilities^9,10.

In accordance with these shifts in MRSA epidemiology and in healthcare utilization, in regions where the CDC’s Emerging Infections Program conducts surveillance for MRSA, a higher burden of nursing home-onset invasive MRSA infections compared to hospital-onset infections has been observed¹¹. That analysis found that of nursing home-onset invasive MRSA infections, the most common infection syndromes were bloodstream infection (91%, including 41% bloodstream infection alone), pneumonia (21%), urinary tract infection (9%), bone or joint infection (8%), and pressure ulcer or chronic wound infection (7%). It also found that aggregated rates of invasive MRSA infections in nursing homes varied across surveillance sites, but data were not available to determine how such rates varied across facilities. Understanding the reasons for these variations could help identify facilities that would most benefit from additional MRSA prevention efforts.

We investigated the variation in nursing home-onset invasive MRSA rates at the facility level and sought to describe which facility-level characteristics were associated with high MRSA rates.

During the project period, 3,074 invasive MRSA cases occurred among residents in long-term care facilities 3 days prior to the initial invasive MRSA culture. For 250 cases, the medical record was unavailable to determine whether the long-term care facility was a nursing home, the facility was determined not to be a nursing home in the surveillance area, or the facility was referred to as a nursing home in the medical record, but the specific facility was not identified. In total, 2,824 were nursing home-onset MRSA cases linked to a specific nursing home.

Of the 666 nursing homes located in the 7-site surveillance area during the project period, 38 closed before 2015 and two did not have data in MDS. Consequently, data from 626 facilities were eligible for inclusion in the analysis.

Of these 626 nursing homes, 82% had at least one resident with an invasive MRSA infection during the project period (Figure 1). It was found that the 20% of facilities (n=127) with the highest rates (i.e., those with “high rates”) accounted for 50% of the invasive MRSA cases. These facilities had rates of at least 3.84 cases/100,000 resident-days (Figure 2). Among all 626 nursing homes, the median invasive MRSA rate was 1.82 cases/100,000 resident-days, with a maximum of 19.3 and an interquartile range of 0.74-3.37 cases/100,000 resident-days. Across these facilities, the median number of invasive MRSA cases per facility was 3 (range: 0–36 cases, interquartile range: 1–6 cases) (Supplemental Figure 2).

A univariate assessment of the characteristics of facilities with high rates vs other facilities is displayed in Table 1. There was a significant difference in geographic distribution (P<0.0001 for the overall chi-square test of site vs nursing home-onset invasive MRSA rate). For example, only 12% of the facilities in the project were in Maryland, but 28% of the facilities with high invasive MRSA rates were in Maryland. All facility-level characteristics were significantly associated with differences in rates in univariate analysis except for registered nurse (RN) staffing (P=0.74) and CMS quality measure rating (P=0.47). The minimum, median, quartile range, and maximum value of facility characteristics are shown in Table 2.

In multivariable regression, a higher percentage of White residents in a facility and a higher number of RN hours per resident per day were associated with lower odds of having a high invasive MRSA rate (Table 3). Presence of a patient with a multidrug-resistant organism; higher hospital-onset MRSA rate in the surrounding area; and higher percentages of male residents, residents with a nasogastric or percutaneous gastrostomy tube, short stay residents (i.e., in the facility <100 days), and residents needing extensive assistance for bed repositioning were associated with higher odds of having a high invasive MRSA rate. Variables that were not significantly associated with having higher invasive MRSA rates include percentages of residents that are African American, percentage of residents with diabetes, presence of any resident with tracheostomy, percentage of residents with a wound, percentage of residents receiving antibiotics recently, percentage of residents on dialysis, CMS survey rating, and CMS quality measure rating.

Substantial variation in invasive MRSA rates was observed across the nursing homes included in this project. For the most part, facility characteristics associated with high rates reflected increased patient complexity. In addition, lower RN staffing levels were found to be significantly associated with high rates of invasive MRSA infection.

The finding that facilities with patients who have more functional needs had high invasive MRSA rates is consistent with a prior finding that functional independence is protective against MRSA colonization for nursing home residents¹⁹. Surprisingly, other than feeding tubes we did not find other medical device use or wounds to be associated with high rates of invasive MRSA, even though they have been associated with MRSA colonization in multiple studies^19–22. Since wounds and other devices were significantly associated with differences in invasive MRSA rates in univariate analysis, this is because of incorporation of other variables in the model (e.g., short-stay, limited mobility) that account for the risk associated with devices. Most of the factors associated with high rates were characteristics of the patient population and not directly modifiable by facilities. We also did not find dialysis to be associated with high rates in the model, even though hemodialysis patients are known to be at very high risk for S. aureus bloodstream infections²³. This might be because only a small proportion of facilities had a substantial proportion of dialysis patients. Indeed, the aOR for the dialysis variable in the model was large but did not reach statistical significance.

However, the finding that increased RN staffing was associated with lower rates of invasive MRSA suggests there may be actions that nursing homes can take to reduce their rates. Less infection control training and increased staff turnover have been associated with infection control citations^24,25. CMS has reported that higher staffing levels were associated with improvements in quality measures such as urinary tract infections, sepsis, and electrolyte imbalance²⁶. RN staffing might represent more than just increased availability of hours for direct resident care, given that the actual number of RN staffing hours per resident per day was relatively low. Instead, it may also represent increased availability of staff with a higher level of clinical training who might be able to provide guidance and facility-wide training on infection control policies or characteristics of facility leadership. For example, increased accessibility of managers and leadership support of staff have been associated with lower staff turnover rates and better resident safety in recent studies^27,28, and these benefits might also translate to infection control practices. Of note, the staffing was not significantly different between facilities with low vs high rates in unadjusted analysis, only after adjusting for characteristics of the resident population. This suggests that facilities with more complex resident populations might benefit from higher staffing levels.

The significance—and even lack of significance in some cases—for other factors assessed here might also be important. For example, facilities with a higher proportion of White residents had lower rates. This finding is a potential marker for issues related to health equity that should be explored in future studies. Notably, this finding was independent of resident characteristics and facility staffing levels and parallels recent findings correlating resident racial composition with differences in outcomes^29,30. We did not find CMS survey or quality ratings to be associated with differences in invasive MRSA rates. This might be because these ratings do not measure the differences in infection prevention and control quality that are most influential in affecting MRSA rates. Finally, we did find that higher hospital-onset invasive MRSA rates for the surveillance area were correlated with higher nursing home-onset invasive MRSA rates. This may reflect the differences in the likelihood that short-stay nursing home patients, expected to be at the highest risk for invasive nursing home-onset MRSA, might have acquired MRSA colonization during hospitalization before admission to the nursing home and likely explains some of the regional differences in nursing home-onset invasive MRSA rates.

This analysis is subject to limitations. The definition of residence in the surveillance area used by this EIP activity excludes cases in persons who moved to a nursing home from outside the surveillance area less than two weeks before the MRSA culture, and the impact of this nuance on the results is not known. The facilities within the surveillance area might not be representative of nursing homes nationally. The CMS MDS data are not validated, and the point prevalence data from 2015 used from MDS to characterize facilities might not be reflective of the entire surveillance period. Some data elements that would be useful, such as the presence of central venous catheters or MRSA colonization specifically, were not available in the MDS data and could be unmeasured confounders. The ecological nature of the analysis could limit our ability to precisely understand the individual-level characteristics of patients. For a small percentage of long-term care facility-onset invasive MRSA cases, we were not able to determine the specific facility, so we might have underestimated rates for some facilities. The data from this time period might not reflect more recent trends. Finally, this paper uses a relatively simple logistic regression model to describe the factors associated with differences in rates of invasive MRSA infections. In the future, for a specific application that needs to identify facilities most likely to have high rates, another type of model using more sophisticated machine learning approaches might yield even more accurate predictions³¹.

In conclusion, in this analysis of invasive MRSA infection data from >600 nursing homes, we found several factors to be associated with high rates of nursing home-onset invasive MRSA infection, including medical complexity of residents and lower RN staffing hours per resident. These findings suggest potential opportunities to improve infection prevention practices in U.S. nursing homes in general, such as increased focus on residents needing the most assistance with activities of daily living. In addition, the model used in this analysis could help to distinguish nursing homes likely to have high rates from those likely to have lower rates; this could be particularly useful for public health and healthcare organizations seeking to allocate limited resources and prioritize specific nursing homes with the highest burden of invasive MRSA infections for future prevention efforts.