Effect of meteorological factors and geographic location on methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococci colonization in the US

Background Little is known about the effect of meteorological conditions and geographical location on bacterial colonization rates particularly of antibiotic-resistant Gram-positive bacteria. We aimed to evaluate the effect of season, meteorological factors, and geographic location on methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci (VRE) colonization. Methods The prospective cohort included all adults admitted to 20 geographically-dispersed ICUs across the US from September 1, 2011 to October 4, 2012. Nasal and perianal swabs were collected at admission and tested for MRSA and VRE colonization respectively. Poisson regression models using monthly aggregated colonization counts as the outcome and mean temperature, relative humidity, total precipitation, season, and/or latitude as predictors were constructed for each pathogen. Results A total of 24,704 ICU-admitted patients were tested for MRSA and 24,468 for VRE. On admission, 10% of patients were colonized with MRSA and 12% with VRE. For MRSA and VRE, a 10% increase in relative humidity was associated with approximately a 9% increase in prevalence rate. Southerly latitudes in the US were associated with higher MRSA colonization, while northerly latitudes were associated with higher VRE colonization. In contrast to MRSA, the association between VRE colonization and latitude was observed only after adjusting for relative humidity, which demonstrates how this effect is highly driven by this meteorological factor. Conclusions To our knowledge, we are the first to study the effect of meteorological factors and geographical location/latitude on MRSA and VRE colonization in adults. Increasing humidity was associated with greater MRSA and VRE colonization. Southerly latitudes in the US were associated with greater MRSA and less VRE. The effect of these factors on MRSA and VRE rates has the potential not only to inform patient management and treatment, but also infection prevention interventions.

in prevalence rate. Southerly latitudes in the US were associated with higher MRSA colonization, while northerly latitudes were associated with higher VRE colonization. In contrast to MRSA, the association between VRE colonization and latitude was observed only after adjusting for relative humidity, which demonstrates how this effect is highly driven by this meteorological factor.

Conclusions
To our knowledge, we are the first to study the effect of meteorological factors and geographical location/latitude on MRSA and VRE colonization in adults. Increasing humidity was associated with greater MRSA and VRE colonization. Southerly latitudes in the US were associated with greater MRSA and less VRE. The effect of these factors on MRSA and VRE rates has the potential not only to inform patient management and treatment, but also infection prevention interventions.

Background
Throughout history, certain infectious diseases have been tightly correlated with seasonal, meteorological, and climatic conditions [1]. The winter peaks in influenza infections are an example of this interaction. However, the mechanisms underlying this association, particularly on pathogens transmitted from person-to-person, are not yet well understood [1]. Furthermore, despite the global public health importance of infections associated with Gram-positive bacteria [2], few studies have examined its association with seasonal and meteorological conditions and yielded inconsistent results [3][4][5][6][7][8].
Patients can be either infected or colonized with bacteria. Colonization is the presence of bacteria in an anatomic site without any symptoms of disease [9][10][11]. Colonization is detected by obtaining surveillance cultures, while bacterial infection is identified through clinical cultures after signs and symptoms are evident, often at sites other than the site of colonization. As bacterial colonization represents an earlier step in the disease pathway, research at this level can provide valuable insights. However, studies evaluating the correlation between seasonal and meteorological factors with methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci (VRE) colonization are limited and have focused on infant populations, particularly on neonates (Table 1) [12,13].
Additionally, our understanding of the effect of geographic location (latitude) on MRSA and VRE rates is limited. Latitude represents not only geographical location, but it can also act as a proxy of environmental conditions, in addition to differences in topography, access and quality of healthcare, and socio-economic conditions. Thus, latitude allows us to explore not only the effect of geographic location but also if this effect is fully explained by meteorological conditions or if some other factors should be studied. The previous effects have not been fully explored for VRE and MRSA effect as multi-site studies on this topic are rare [3,5,16].
To further explore these gaps, this study aims to address two different research questions. First, we aim to evaluate the effect of season, temperature, humidity, and precipitation on MRSA and VRE colonization among adults. Second, we aim to assess and explore the effect of geographical location on MRSA and VRE colonization. To our knowledge, we are the first to address these questions specifically on MRSA and VRE adult colonization. Understanding these associations has the potential not only to inform patient management and treatment, but -Seasonal trend decomposition method.
-CA-MRSA incidence peaked in late summer, particularly in children.
-HA-MRSA incidence peaked in the winter. (Continued)

Methods Population
We analyzed a prospective cohort of adult patients admitted to 20 geographically-dispersed intensive care units (ICUs) across the US as part of the Benefits of Universal Glove and Gown (BUGG) cluster-randomized trial during the period between September 1, 2011 and October 4, 2012 [21] (Fig 1). For the purpose of this analysis, nasal and perianal admission swabs were tested for MRSA and VRE respectively using culture and PCR to detect resistance genes (mecA or vanA/vanB).

Meteorological and geographical variables
We collected mean monthly temperature (˚F), mean monthly relative humidity (%), and monthly total precipitation (inches) by city from the National Oceanic and Atmospheric Administration website [22]. In addition, latitude from each hospital site was collected from the My NASA Data website [23]. The previous variables were included in the analysis as continuous variables ( Table 2). In addition, a categorical variable for season was created based on the swab collection date as follows: winter (December to February), spring (March to May), summer (June to August), and autumn (September to November).

Statistical analysis
Descriptive analysis. Monthly proportions of MRSA or VRE colonization on admission were initially estimated per site. We estimated correlation coefficients between these proportions and each meteorological variable. We also performed tests of proportions (Generalized linear model (GLM)) across sites and seasons.
Univariate models. To determine the effect of each meteorological and geographical variable on colonization, we first constructed a Poisson regression model for each pathogen using monthly aggregated colonization counts per site as its outcome and mean temperature, relative humidity, total precipitation, season, or latitude as the primary exposure as individual variables. The log of the total monthly number of swabs collected and processed per site was defined as the offset of the model. Initially, we detected overdispersion, usually due to higher variability among counts than would have been expected for a Poisson distribution, which affected our models' goodness of fit (deviance and Pearson Chi Square) [24]. We accounted for overdispersion by introducing a dispersion parameter to the model [24]. The statistical methods chosen were consistent with prior studies [4,25].
Multivariate models. Model 1: Combined effect of meteorological variables. This multivariate model evaluated the association between the monthly prevalence of each pathogen and meteorological variables (mean temperature, relative humidity, and precipitation).
Model 2: Effect of geographical location. This multivariate model evaluated the association between the monthly prevalence of each pathogen and the ICU's geographical location (latitude), adjusting for the meteorological variables that were statistically significant (p<0.05) in their respective univariate models.

Descriptive analysis
Our study was conducted in 20 different ICUs across 16 different US states with an average of 19 beds (range: 9-36 beds) per ICU. Fifty five percent of our ICUs were medical intensive care units (MICUs), while the rest were surgical intensive care units (SICUs) (25%) or a combination  [21]. In summary, a total of 24,704 and 24,468 patients were tested for MRSA and VRE colonization respectively upon admission to the ICU. Overall, we observed an average of 10% MRSA colonization on admission, ranging from 3% to 16% for each hospital. For VRE, we observed an average of 12% colonization rate on admission, ranging from 3% to 25%. In our dataset, mean temperature was positively correlated with total precipitation (r = 0.34, p<0.001). Similarly, relative humidity was positively correlated with total precipitation (r = 0.54, p<0.001). However, no significant association was observed between temperature and relative humidity. In contrast, latitude was negatively correlated with mean temperature (r = -0.48, p<0.001), total precipitation (r = -0.38, p<0.001), and relative humidity (r = -0.06, p = 0.379).
MRSA monthly colonization was positively correlated with mean temperature (r = 0.16, p = 0.008), relative humidity (r = 0.24, p<0.001), and total precipitation (r = 0.22, p<0.001). MRSA monthly colonization was negatively correlated with latitude (r = -0.34, p<0.001) i.e. colonization was higher at latitudes closer to the equator. For VRE, a positive correlation was only observed between VRE monthly colonization and relative humidity (r = 0.20, p = 0.004). In contrast, for VRE, no significant correlations were detected with total precipitation, mean temperature, or latitude. No significant difference on MRSA or VRE colonization was observed across seasons (p = 0.589 and p = 0.922 respectively).

Univariate models
For MRSA, positive associations were observed in our univariate Poisson models between MRSA counts and all studied meteorological conditions. For every unit (˚F) increase of mean monthly temperature, there was a 0.7% increase in MRSA prevalence (p = 0.002). Furthermore, for every 1% increase in relative humidity, there was a 1.3% increase in pFrevalence (p<0.001). Similarly, a 5.6% increase in MRSA prevalence was observed per one-inch increase in annual precipitation (p<0.001). In addition, for every one degree (˚) reduction of latitude, there was a 3.8% increase in prevalence (p<0.001). However, no significant association was observed between colonization and season (p = 0.669).
For VRE, a positive association was only observed in our univariate Poisson models between VRE counts and relative humidity. For every 1% increase in relative humidity, there was a 0.9% increase of VRE prevalence (p = 0.015). In addition, no significant association was observed between colonization and season (p = 0.669) or latitude (p = 0.672).

Multivariate models
Model 1: Combined effect of meteorological variables. For MRSA, only relative humidity remained significant in this multivariate model (Table 3). For every 1% increase of relative humidity, there was approximately a 0.9% increase in MRSA colonization when controlling for the other meteorological variables (p = 0.010). For VRE, only relative humidity approached significance in this multivariate model (Table 4). For every 1% increase of relative humidity, there was approximately a 0.8% increase in VRE colonization when controlling for the other meteorological variables (p = 0.064).
Model 2: Effect of geographical location. For MRSA, latitude was negatively associated with colonization even after adjusting for meteorological variables (p = 0.001). A 2.4% increase in MRSA colonization was observed per unit decrease of latitude (˚) ( Table 3). In other words, southern states had higher MRSA colonization rates than northern states.
For VRE, latitude was positively associated to colonization after controlling for the confounding effect of relative humidity (p = 0.019). A 1.8% increase in VRE colonization was observed per unit increase of latitude (˚). In other words, northern states had higher VRE colonization than southern states (Table 4).

Discussion
We observed a significant effect of meteorological factors and geographical location on MRSA and VRE colonization in our study population. For MRSA and VRE, a 10% increase in relative humidity led to an 8-9% increase in prevalence rate. Furthermore, we observed a stronger effect of geographical location on colonization. Interestingly, the direction of this association varied by pathogen. The closer to the equator, the higher the observed MRSA colonization but the lower the observed VRE colonization.
Literature specifically studying the effect of season on MRSA and VRE colonization is very limited. There are only three studies assessing this effect specifically on MRSA colonization. Similar to our study, Ogawa et al. also reported no significant difference in S. aureus colonization across seasons on the skin of 40 healthy individuals in Japan [14]. Likewise, Harrison et al. observed no significant difference on S. aureus colonization across seasons among 72 infants in the United Kingdom [15]. In contrast, Giuffrè et al. described an incidence density peak of MRSA nasal colonization among neonates (n = 832) admitted to the neonatal ICU (NICU) during the summer and autumn quarters [12], although no statistical analysis was done. With regard to VRE, only Hufnagel et al. have reported a significant increase in multidrug resistant enterococcus colonization in winter and spring months across 274 neonates admitted to the NICU [13]. The epidemiological differences between adult and neonatal populations, such as different levels of immunity, types of exposure and risk factors (14)(15)(16), and a different microbiome that potentially could have an effect of antibiotic-resistant Gram-positive bacteria colonization (17,18), may explain the observed differences with our results.
Unlike previous studies on these pathogens, we were particularly interested to study the effect of geographic location or latitude on colonization rates. Generally, higher temperatures and lower seasonal variation can be observed closer to the equator [26]. Fisman et al. found that Gram negative-associated bacteremia is more common in locations closest to the equator [27]. Albelnoor et al. also described higher rates of MRSA-associated mediastinitis across sites with lower latitude [19]. Similarly, we observed more MRSA colonization in the southerly sites. However, the opposite was observed for VRE. Our northerly sites had higher VRE colonization rates. The appearance of the effect of latitude on VRE colonization only after adjusting for relative humidity demonstrates this effect is highly driven by this meteorological factor. To our knowledge, there are no other studies describing the effect of latitude on VRE colonization or infection rates. Additionally in our study, the observed geographic effect particularly on MRSA colonization remains significant even after adjusting for meteorological variables. This suggests that other factors besides the analyzed meteorological variables are necessary to explain this association. Further studies are necessary to identify these specific other factors.
This study has several limitations. The most important limitation is that we only had one year of data. Hence, we were unable to perform a more rigorous time-series analysis to investigate VRE and MRSA seasonality or to further analyze the effect of meteorological factors and geographical location. We believe that future studies should include several years to confirm the observed effects on bacterial colonization. In addition, we could not adjust for important confounders. Antibiotic use/prescription has been associated with the prevalence of antibioticresistant microorganisms in a number of studies at the patient level. For instance, Sun et al. reported a correlation between prescriptions of fluoroquinolones and prevalence of ciprofloxacin-resistant MRSA [28]. Another limitation is that we were unable to collect and adjust on hospital-level confounders (i.e. age distribution, sociodemographic status) for each of the study sites.
Nevertheless, the main strength of this study was the ability to assess the effect of meteorological variables on solely MRSA and VRE colonization instead of infection (or a combination). As colonization is an earlier step in the disease pathway, research at this level may provide valuable insights to strengthen prevention strategies as colonized individuals could potentially have different risk factors than infected individuals. In addition, due to the diversity of sites across the US that made up our study, to our knowledge, we are the first to study the effect of geographical location (latitude) on MRSA and VRE adult colonization rates.
Further studies should investigate the effect of latitude using global multi-site data. Moreover, the role played by additional factors such as weather, topography, or socioeconomic factors that we were unable to collect should be investigated to help elucidate the different geographical effect observed across microorganisms. Additionally, the potential effect of global warming on bacterial colonization and infection should be analyzed.
MRSA and VRE adult colonization rates at admission to ICUs across the US are far from inconsequential and should be considered when making decisions on patient care and infection control. For example, bacterial colonization information can drive local empiric antibiotic choice and influence local infection control intervention choices. Additionally, relative humidity and geographical location appear to have an important effect on VRE and MRSA colonization rates. If our results are confirmed in future studies, this conclusion has different implications for different levels of public health. At the local level, infection preventionists could enhance surveillance and decolonization measures during humid periods regardless of season. In contrast, state and national public health officials may need to incorporate the effect of geographical location on their decision making process depending on the pathogen of interest and regardless of the time of year. For instance, they could allocate more funding for MRSA prevention to southern states than northern states year round regardless of season.