World-Wide Variation in Incidence of Staphylococcus aureus Associated Ventilator-Associated Pneumonia: A Meta-Regression

Staphylococcus aureus (S. aureus) is a common Ventilator-Associated Pneumonia (VAP) isolate. The objective here is to define the extent and possible reasons for geographic variation in the incidences of S. aureus-associated VAP, MRSA-VAP and overall VAP. A meta-regression model of S. aureus-associated VAP incidence per 1000 Mechanical Ventilation Days (MVD) was undertaken using random effects methods among publications obtained from a search of the English language literature. This model incorporated group level factors such as admission to a trauma ICU, year of publication and use of bronchoscopic sampling towards VAP diagnosis. The search identified 133 publications from seven worldwide regions published over three decades. The summary S. aureus-associated VAP incidence was 4.5 (3.9–5.3) per 1000 MVD. The highest S. aureus-associated VAP incidence is amongst reports from the Mediterranean (mean; 95% confidence interval; 6.1; 4.1–8.5) versus that from Asian ICUs (2.1; 1.5–3.0). The incidence of S. aureus-associated VAP varies by up to three-fold (for the lowest versus highest incidence) among seven geographic regions worldwide, whereas the incidence of VAP varies by less than two-fold. Admission to a trauma unit is the most important group level correlate for S. aureus-associated VAP.


Introduction
Ventilator-Associated Pneumonia (VAP) in association with Staphylococcus aureus has been reported from over 100 intensive care units (ICU) worldwide . In two series drawn predominantly from ICUs in The United States of America and Europe, S. aureus accounted for 20% [134] and 22% [135] of bronchoscopically-documented cases of VAP.
Whether VAP is associated with an increase in attributable mortality may depend on the infecting organism [8,12,115,135]. There may be a specific mortality risk for ventilator-associated pneumonia in association with S. aureus infections in the ICU, although this may be influenced by associated resistance to methicillin [136,137].
The objective here is to define the extent of geographic variation in the incidence of VAP associated with S. aureus within the published literature versus the variation in the incidences associated with VAP overall and with MRSA-VAP. An additional aim is to explore the degree to which any variation may be explainable by other study-related factors, such as mode of VAP diagnosis or admission for trauma [67], using meta-regression methods. classified by geographic region as detailed in Table 1. There were 17 multinational ICU surveys from six publications that were derived from ICUs that had been anonymized by originating country in these publications.
Microorganisms 2018, 6, x FOR PEER REVIEW 3 of 22 than one study. The studies are detailed in Table S1 (see the additional file). The studies were classified by geographic region as detailed in Table 1. There were 17 multinational ICU surveys from six publications that were derived from ICUs that had been anonymized by originating country in these publications.   There were 21 studies that reported for trauma ICU populations. While none of the studies were undertaken in an ICU subject to a known outbreak, there were nine studies undertaken in the context of an infection control intervention. The use of bronchoscopic sampling in the diagnosis of VAP was unequal among the seven regions being used in more than half of the studies from Northern Europe and the Mediterranean versus less than half of studies elsewhere (p < 0.001; chi-square = 27.13, 6 df).
The study-specific S. aureus-associated VAP incidence is displayed by regions (Figures 2-6) and collectively with all studies together (Figures 7 and 8). The incidence of MRSA-VAP are displayed by year of publication ( Figure S1) and collectively by country ( Figure 9). Over all 162 groups, the summary incidence of VAP was 21.3 (18.9-23.8) per 1000 MVD ( Figure S2 There were 21 studies that reported for trauma ICU populations. While none of the studies were undertaken in an ICU subject to a known outbreak, there were nine studies undertaken in the context of an infection control intervention. The use of bronchoscopic sampling in the diagnosis of VAP was unequal among the seven regions being used in more than half of the studies from Northern Europe and the Mediterranean versus less than half of studies elsewhere (p < 0.001; chi-square = 27.13, 6 df).
The study-specific S. aureus-associated VAP incidence is displayed by regions (Figures 2-6) and collectively with all studies together (Figures 7 and 8). The incidence of MRSA-VAP are displayed by year of publication ( Figure S1) and collectively by country ( Figure 9). Over all 162 groups, the summary incidence of VAP was 21.3 (18.9-23.8) per 1000 MVD ( Figure S2 Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale. S. aureus VAP incidence per 1000 mechanical ventilation days and 95% CI for groups from Northern European countries. For comparison, the summary S. aureus VAP incidence (vertical line) derived from the studies in Figure 2 is shown for reference. Studies are listed in Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale. S. aureus VAP incidence per 1000 mechanical ventilation days and 95% CI for groups from the studies from Asia and the Middle East. For comparison, the summary S. aureus VAP incidence (vertical line) derived from the studies in Figure 2 is shown for reference. Studies are listed in Table S1 (see the additional file). Note that the x axis is a logarithmic scale.  Table S1 (see the additional file). Note that the x axis is a logarithmic scale.        The highest and lowest S. aureus-associated VAP incidences were amongst reports from Mediterranean versus Asian ICUs, respectively. By contrast, the highest and lowest incidences of VAP overall were amongst reports from Mediterranean versus Northern European ICUs, respectively. The incidence of MRSA VAP was lowest within reports from Northern European ICUs (Table 1).
Meta-regression models of VAP incidence, S. aureus-associated VAP and MRSA-VAP incidence are presented in Table 2. In the meta-regression model for VAP incidence overall and for S. aureus-associated VAP incidence, origins from a trauma ICU, but not mode of diagnosis using bronchoscopic sampling were significant correlates. In these models, origin from a Northern European study was significantly associated with lower incidences of VAP overall and MRSA-VAP, but not S. aureus VAP. In none of the three models was the year of publication significantly associated with incidence. However, a closer examination revealed an interaction in that a decline in S. aureus-associated VAP ( Figure 8) and MRSA-VAP (see the additional file: Figure S2) incidences over the years was apparent for reports from non-trauma ICUs, but not for trauma ICUs (p = 0.001). The highest and lowest S. aureus-associated VAP incidences were amongst reports from Mediterranean versus Asian ICUs, respectively. By contrast, the highest and lowest incidences of VAP overall were amongst reports from Mediterranean versus Northern European ICUs, respectively. The incidence of MRSA VAP was lowest within reports from Northern European ICUs (Table 1).
Meta-regression models of VAP incidence, S. aureus-associated VAP and MRSA-VAP incidence are presented in Table 2. In the meta-regression model for VAP incidence overall and for S. aureus-associated VAP incidence, origins from a trauma ICU, but not mode of diagnosis using bronchoscopic sampling were significant correlates. In these models, origin from a Northern European study was significantly associated with lower incidences of VAP overall and MRSA-VAP, but not S. aureus VAP. In none of the three models was the year of publication significantly associated with incidence. However, a closer examination revealed an interaction in that a decline in S. aureus-associated VAP ( Figure 8) and MRSA-VAP (see the additional file: Figure S2) incidences over the years was apparent for reports from non-trauma ICUs, but not for trauma ICUs (p = 0.001). Interpretation. The reference group is the composite group of multinational and ungrouped studies, and this coefficient equals the difference in log from 0 (a log equal to 0 equates to a rate of 1). The other coefficients represent the difference in log for groups positive for that factor vs. the reference group. c The coefficient for trauma represents the increment in log for an ICU having a majority of admissions for trauma. d The coefficient for year of publication represents the linear increment in log for each year after 1980. e For sampling using bronchoscopic versus tracheal sampling. f Studies undertaken during an infection control intervention.

Discussion
This is a survey of the incidences of VAP overall, S. aureus-associated VAP and MRSA-VAP among over 100 published studies using meta-analysis. The incidences of S. aureus-associated VAP and MRSA-VAP each vary by up to three-fold for the lowest versus highest incidence region among seven geographic regions worldwide, whereas the incidence of VAP varies by less than two-fold.
This variation in incidence among seven broad multinational regions is partly explainable by a limited number of group level factors. Of note, in the meta-regression models, trauma is more strongly correlated with S. aureus-associated VAP than was the case for the incidence of VAP overall or MRSA-VAP. A decline in the incidences of S. aureus-associated VAP and MRSA-VAP in association with year of publication is apparent only for reports originating from non-trauma ICUs.
The less than three-fold variation in S. aureus-associated VAP contrasts with the greater than five-fold variation in Acinetobacter-associated VAP incidence observed in an analysis of mostly the same studies as included here [139].
The findings here reinforce and further characterize previous observations in four multi-national surveys [1, 3,5,140]. Rello et al. surveyed ICUs in the following four regions: Paris, Barcelona, Montevideo and Seville, and likewise found less variation between the sites in S. aureus-associated VAP than was the case for Acinetobacter-associated VAP [141]. Another multi-national prospective 24-month survey [5] was conducted across 55 ICUs of 46 hospitals in Argentina, Brazil, Colombia, India, Mexico, Morocco, Peru and Turkey. This anonymized survey also found less variation in S. aureus-associated VAP than was the case for Acinetobacter-associated VAP. This survey found an overall rate of VAP of 24.1 per 1000 MV days with S. aureus accounting for between 13 and 53% of VAP isolates amongst the eight anonymized countries [5].
Kollef et al. prospectively surveyed VAP among 1873 mechanically-ventilated patients in 56 ICUs from 11 countries in the following four regions: Europe, United States, Latin America and the Asia-pacific region [3]. This survey [3] found that the incidence per 100 patients of VAP overall, S. aureus VAP and MRSA-VAP among all 56 ICUs was 293/1873 (15.6%), 65/1873 (3.5%) and 27/1873 (1.4%), respectively. This compares to incidence proportions for VAP overall and S. aureus VAP found here being 17.1 (14.0-20.6), 3.4 (2.6-4.5) and 1. 95 (1.14-3.3), respectively. Interestingly, these investigators found that the incidence of VAP overall, but not the incidence of P. aeruginosa VAP varied among the four broadly-defined worldwide regions. In this study [3], the incidences of S. aureus VAP and MRSA-VAP were each only reported in aggregate.
Finally, a survey of 27 European ICUs found a summary VAP incidence of 18.3 per 1000 MVD and that the dominant nosocomial pneumonia isolate varied among the nine European countries in the survey [1,138]. S. aureus was the dominant pneumonia isolate in Spain, France, Belgium and Ireland [1,138].
The extent to which any possible variation in VAP microbiology between regions is explainable by group level factors is difficult to establish in studies that are either short term or single centre. However, the findings here that admissions for trauma, but not bronchoscopic sampling undertaken as part of VAP diagnosis are significant factors towards S. aureus VAP are in line with findings reported from single centre studies [150,151].
With a literature survey, a considerable convenience is that the data are readily available, and the meta-regression methods for analysing these types of data are established. A random effects meta-regression methods is used as previously to enable the imprecision associated with each individual study estimate to be incorporated in the derivation of both the summary estimates and in the meta-regression models [139,152]. This is the major advantage of a meta-analysis in comparison to a simple narrative review. Estimates with random effects are more conservative (i.e., wider 95% confidence limits) than with fixed effects. This analytic approach enables an appreciation of the contextual influences of study-specific factors that would not be apparent within a single centre study examined in isolation.
There are several limitations to this analysis of the literature. Of particular note, this is not a systematic review, and the analysis is limited to English language articles. This is an analysis at the group level and therefore is unable to take account of patient-specific risk factors for S. aureus-associated VAP. For example, the duration of mechanical ventilation and usage of empiric antibiotic therapy are important VAP determinants at the level of the individual patient that cannot be explored at the group level of analysis. To include patient level information within a meta-analysis would require an individual patient data meta-analysis.
A further limitation relating to the survey of MRSA-VAP is that in some reports, MRSA-VAP may not have been reported as a consequence of being rare in that specific location. Hence, this reporting bias may lead to under-representation of zero counts. Moreover, the analysis is unable to account for strain type differences underlying the variation in both S. aureus VAP incidence and also in MRSA-VAP incidence. The 'intervention periods' studies include a broad range of interventions. Their inclusion is to indicate the influence relative to other group level associates on the incidences. However, the methodology used here is not appropriate for the evaluation of the effectiveness of infection control interventions. Randomised controlled trials were not included here for two reasons. Firstly, the populations of these studies usually are a small sub-group that meets specific inclusion criteria. Moreover, there is the potential for a contextual effect in the presence of any such intervention within an ICU on the incidence of VAP overall [152], and more so in relation to the potential contextual effect of prevention interventions on S. aureus-associated VAP [151][152][153].
As with any multi-national survey, a range of definitions was applied in the diagnosis of VAP among the studies here, and standardisation was not possible. The classification of studies into those that did versus did not use bronchoscopic sampling towards VAP diagnosis was a simplifying compromise.
The grouping of countries into near neighbour groupings is somewhat arbitrary. This grouping may conceal important variations within each of the regions. Country and even regional groupings could be confounded by other variables such as infection control practices, prevalence of antibiotic use and standards of care for patients receiving mechanical ventilation, which are not able to be considered in the analysis here. Indeed, even the imperative to publish may differ in different countries. The influence of publication bias is difficult to estimate.

Conclusions
There is a variation of up to three-fold in S. aureus-associated VAP and MRSA-VAP among published reports from seven broad geographic regions worldwide. This variation exceeds the variation in incidence of VAP overall. For MRSA-VAP incidence, there is a complex interaction between the year of publication and admission for trauma.
Funding: This research was supported by the Australian Government Department of Health and Ageing through the Rural Clinical Training and Support (RCTS) program. The funding agency had no role in the preparation of the manuscript, nor its approval for submission.

Conflicts of Interest:
The authors declare no conflict of interest.