Introduction

Methicillin resistant Staphylococcus aureus (MRSA) is a major cause for healthcare associated infections (HAI), and considered a relevant patient safety issue. Infection control programs worldwide have proposed and implemented various strategies against the spread of this pathogen. Risk factors have been associated with MRSA carriage and invasive disease [1], [2], yet, it has become clear that risk factors for MRSA acquisition in the hospital have to be separated from those associated with patients already MRSA-positive upon hospital entry. Subsequently, MRSA screening and risk factor analyses on admission have been performed, risk factor scores for selective screening have been developed [3], and in some countries directly neighbouring Germany (e.g., the Netherlands and Denmark) such early-implemented targeted screening and isolation strategies (‘search and destroy’) have been successfully applied for controlling the MRSA epidemics [4], [5]. Moreover, in many countries with MRSA endemicity, national guidelines now recommend application of such risk factor-based screening as part of an ‘active detection and isolation’ (ADI) strategy [6]. Moreover, in a multifactorial approach analyzing the association of MRSA prevalence and key infection control parameters in 146 European hospitals, the implementation of an MRSA screening policy was found to be associated with decrease in MRSA prevalence [7]. As a result, in a consensus statement the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) concluded that policies should be guided by local MRSA infection and colonization rates [8]. This consensus statement has recently been updated [9]. Despite intense research, guideline implementation, and policy making, the issue of an optimal cost-effective and patients safety-focussed approach towards admission screening remains debated as carefully performed studies have come to contrasting results [10]–[12]). While universal admission screening may most effectively prevent MRSA infections due to unrecognized transmission [13], [14], the presumed high costs associated with testing and contact precautions have also prevented its wide adoption [15].

This debate may be a consequence of the local/regional coverage of the screening policy in the various studies: An analysis applying extensive mathematical modeling confirmed previous observations, i.e. that admission screening will be less effective and more costly if neighboring hospitals do not screen [16]. Thus, it becomes clear that effective detection for MRSA has to be implemented with a strategy well beyond single departments or hospitals. The German Antibiotic Resistance Strategy (DART) [17] has addressed this issue by fostering regional German model projects for establishing regional networks on prevention and combat of antibiotic resistances and their spread. As part of this initiative, we have established the first State-wide German network for the control of MRSA, MRSAarNet (www.mrsaar.net).

The aim of this prospective cohort study was to define the MRSA and MSSA admission prevalence covering all hospitals in one German state. This evaluation was accompanied by an ascertainment of risk factors for MRSA acquisition. The overall goal was to provide a reliable database for establishing specific, region-tailored recommendations for an effective screening policy.

Methods

As a study prerequisite, we intended to exclude any bias due to inter-laboratory or seasonal variabilities by processing all samples in one diagnostic laboratory (Institute of Medical Microbiology and Hygiene, University of Saarland Medical Center) during the period of October and November, 2010. Accordingly, during the study period an anticipated processing work load of 20.000–30.000 specimens with peak numbers of up to 1.500–2.000 specimens per workday were expected. Despite this elevated workload, a minimal turn-around time needed to be assured because participating hospitals expected the report of confirmed results within same time intervals as those offered by their routinely commissioned laboratories. This high number of specimen required the implementation of novel, automated methods for processing.

Accordingly, we opted for an automated sample processing system, the Walk Away Specimen Processor, WASP® (Copan, Brescia, Italy) [18], [19]. The use of WASP relies on a liquid based specimen microbiology technique, and we employed the flocked swab system provided by the WASP manufacturer allowing for immediate release of swabbed microorganisms into the Amies medium (ESwab™, Copan) without the need of a separate enrichment step. This technique has been evaluated for MRSA screening purposes [20], [21]. Specimens were directly plated on CHROMagar MRSA/CHROMagar S. aureus detection biplates (Mast, Germany). Prior to initiation of the main study, the plating protocol for WASP (using 30 µl of the Amies medium eluate) was optimized by careful comparison with the results of parallel manual streaking either of the ESwab or of 30 µl of Amies medium. The streaking pattern by WASP was found to be of particular importance, and a distribution of the 30 µl calibrated loop content through the entire length of the 80 mm CHROMagar biplate was found to be essential to achieve optimal results.

A prestudy was performed to assure that this automated liquid microbiology technique results in similar MRSA detection rates compared to conventional detection rates using cotton swabs and broth enrichment. The ESwab system was compared with conventional cotton swab specimen followed by enrichment (tryptic soy broth, 18 h, 35°C) at 195 admission patients of the Department of Urology, University of Saarland Medical Center, prospectively examined with parallel nasal swabs. The overall number of S. aureus (MRSA and MSSA) detected by either method was 66/195 (33.8%) specimen. Both methods identified the identical 7 MRSA among these 66 S. aureus isolates (10.6%)(overall MRSA prevalence 7/195 [3.6%]) while of the remaining 59 MSSA isolates, 21 were detected only by either method (with the cotton swab/enrichment method missing 8 and the ESwab/Amies method missing 13 isolates). Hence, the advantages of parallel testing in all Saarland hospitals, analysis in one centralized laboratory, and rapid communication of any positive result (i.e. features ascertained by the novel ESwab/WASP method) to our opinion clearly outweighed a potential slightly reduced MSSA detection sensitivity, and this method was selected as the standard procedure for the main study.

Results

MRSA/MSSA Prevalence according to the Department of Admission

For 384/436 MRSA positive patients, a questionnaire with completed risk factor evaluation form was available. 22/384 (5.7%) patients were transferred from another hospital. A valid indication of the clinical department upon admission was available for 361/384 forms, and this cohort was analyzed with respect to the attribution of MRSA admission rates to the respective medical specialty.

Admission to a geriatric department was associated with the highest MRSA admission prevalence, followed by intensive care units. Admission to a general medical or surgical department was associated with MRSA prevalence close to the mean prevalence found in our study, while admission to a psychiatric and pediatric department was associated with the lowest prevalence (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. MRSA admission prevalence according to medical specialty.

https://doi.org/10.1371/journal.pone.0073876.t001

Age and Gender

For 17,244/17,975 patients of the study population, the date of birth was available. MRSA carrier status was demonstrated in 373 of these 17,244 patients. The mean age of the study population was 56.5 years (SD: 12.3 years), the median age was 63 years (25^th and 75^th percentile, 45 and 76 years, respectively). MRSA colonized study patients were significantly older than the overall study population (mean age, 67.5 years [SD: 36.0], median 76 years [25^th and 75^th percentile, 64 and 84 years, respectively]) (p<0.001)(for details in age related prevalence please refer to Figure 1). For 17,415/17,975 study patients, valid gender identification was available; 8333 of these patients (47.9%) were males. MRSA prevalence was significantly higher in males than in females (207/8333 [2.48%] vs. 171/9082 [1.88%]; p = 0.007).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. MRSA prevalence as a function of age.

https://doi.org/10.1371/journal.pone.0073876.g001

Risk Factor Analysis

Only 40/384 (10.4%) of all MRSA colonized study patients did not report any risk factor. The distribution of risk factors in MRSA positive study patients is shown in Table 2. Univariate analysis revealed that 10 out of 14 risk factors evaluated were found to be associated with a significantly increased MRSA positivity; only ’farm’, ’meat’, ‘burn’ and ‘transfer’ were not significantly associated with MRSA carriage upon admission (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariate analysis of risk factors associated with MRSA carriership upon admission and conditional logistic regression (only significant risk factors depicted in the table).

https://doi.org/10.1371/journal.pone.0073876.t002

In contrast, in a multivariate analysis using conditional logistic regression, only 7 factors were significantly associated with an increased risk: ’history’, ‘skin’, ‘catheter’, ‘diabetes’, ’chronic care’, ‘contact’, and ‘hospital’ (Table 2).

With respect to the relative risk for MRSA carriage of the various factors evaluated as a function of age, we selected the age of 40 years and above as cutoff (the MRSA prevalence in the patient group younger than 40 years was <1/100), and analyzed risk factors in this subpopulation. It appears that in this elder population, the multivariate ORs are very similar to the ORs in the entire study group (data not shown).

Risk Factor-based Number Needed to Screen and Resulting MRSA Detection Fractions

Risk factors were sorted according to the rank of significant OR results in the multivariate analysis, beginning with the risk factor with highest OR (Table 3). First, the absolute number of risk factor entries in the study population (n = 17,975) was ascertained. Next, upon hierarchical database retrieval beginning with the risk factor associated with highest OR (‘history’), the number of cumulative risk factor entries was assessed, and expressed in percent of the study population. This percentage indicates the proportion of admission patients to be screened. Correspondingly, the number of MRSA detected (as a function of risk factor entry) was determined, both in absolute as well as in cumulative numbers. The percentage indicates the proportion of MRSA detected upon risk factor based screening. For instance, if only the top 3 risk factors (‘history’, ‘skin’, 'catheter') are ascertained, 11.9% of admission patients need to be screened resulting in the detection of 41% of all MRSA positive patients.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Cumulative proportions of patients reporting one or several risk factors and resulting number-to-screen (applying risk factors significant in conditional logistic regression analysis, see Table 2).

https://doi.org/10.1371/journal.pone.0073876.t003

If patients are screened with respect to any of the first 6 risk factors in multivariate analysis significantly associated with MRSA carriership, 30.9% of patients would be tested to identify 66.7% of MRSA positive patients. In particular, the inclusion of ‘hospital’ as the least-ranked yet most frequently entered significant risk factor requires the screening of an additional 34.5% of patients (in our study, 4,574 patients) but reveals only an additional 16.6% of MRSA positive patients (64 patients).

In addition, it was of interest in our study to compare the results when employing the list of risk factors established by RKI with the risk factors determined in our study. In line with the RKI recommendations the following risk factors were analyzed: five risk factors (‘history’, ‘facility’, ‘hospital’, ‘farms’, ‘contact’) independently prompting an indication to screening plus six factors considering a risk only if at least two of these factors are present (‘chronic care’, ‘antibiotics’, ‘catheter’, ‘dialysis’, ‘skin’, ‘burn’). Overall, employing the RKI risk criteria, 9,099/17,975 (50.6%) patients would be needed to be screened conferring a detection rate of 298/384 (77.6%) of MRSA.

Discussion

The here presented large study analyzing the MRSA admission prevalence in an entire German Federal State has been carried out according to a unique study concept of identical preanalytical and analytical detection methods performed in a single laboratory during a short parallel study period of four weeks, in order to minimize bias by external factors. Such a parallel, prospective universal admission screening approach has only rarely been undertaken elsewhere; rather, in the literature, most admission screening studies were restricted to single or few hospitals often including only a few medical departments or smaller areas, and employing various culture based and molecular detection techniques.

Data from published studies allow to compare the admission prevalence of 2.2/100 patients determined in our study with the situation elsewhere. A non-exhaustive list from these reports from European countries [11], [31]–[38] and from North America [10], [25] reveals admission prevalence values ranging between 3/100 and 9/100. One of the largest MRSA intervention trials, the Veterans Affairs Initiative to prevent MRSA infections [39] reported admission prevalence values as high as 13.6/100. In German hospitals, an MRSA point prevalence or on-admission prevalence between 0.7/100 and 5.3/100 [40], [41] was reported (of note, the hospital admission/discharge rates reported by the European Union for Germany are among the highest in the Union with 240 discharges/1000 population, with a mean among 24 EU countries of 176/1000 [42]; this fact underlines the need to collect local data as a basis for targeted screening particularly in high hospital admission-rate regions such as in Germany). Two larger studies, the Euregio study in Münsterland, and the study in Siegen-Wittgenstein should be mentioned separately: The Euregio study revealed an admission positivity rate of 1.6/100 admissions (total patient number, n = 25,540) [43]; yet, analyses were carried out in different testing laboratories. The Siegen-Wittgenstein study reported 1.4/100 admission MRSA prevalence (total patient number, n = 6,985) (S. aureus 22,3%) with the latter being performed both in acute care and rehabilitation facilities [44]. One recent study performed in one hospital in the State of Brandenburg, Germany, reported 0.8/100 MRSA prevalence (total patient number, n = 13.855) at admission [45]. These literature data demonstrate that it is almost impossible to extrapolate secular national supra-regional MRSA prevalence data on the MRSA burden for individual hospitals or regions. However, it is clear that solely based on clinical-microbiological cultures the prevalence of MRSA is grossly underestimated with 85% of MRSA-colonized patients being missed [46]. Hence, routine performance of at least targeted MRSA hospital admission screening is now recommended [8], and has also been included in a number of health system and legislator policies. The scope of screening (universal versus targeted) as well as the method (molecular versus culture-based screening), however, depends on the baseline on-admission prevalence as summarized in a recent review [9].

Yet, how should a screening program be introduced for a given hospital or region, if such robust baseline data are not available? To respond to this question, we initiated this study in order to provide hospitals and State policy makers with the necessary data on the MRSA admission burden for the State of Saarland hospitals. As a prerequisite to convince the directors of all hospitals to collaborate in this universal screening approach, the time frame of the study needed to be limited to a maximally 4-week-screening period. Moreover, in a system with hospitals typically employing diagnostic services from different laboratories (including different microbiologic-analytical protocols), subtle but potentially important differences in the MRSA prevalence were expected. Thus, to overcome this problem, a single laboratory was selected to process all samples. In addition, due to the short and parallel time frame, the large number of daily analyses required automation in specimen handling, a prerequisite being met with the automated specimen processor employed.

Our study clarified a number of other important issues. While we observed a variable MRSA admission prevalence between the institutions, the overall differences were minor with exception of four hospitals with MRSA rates above 4.0/100 admissions (these latter receiving a larger number of elderly, gerontopsychiatric patients)(data not shown). In accordance with findings in the Euregio [43], the MRSA admission prevalence in 16/24 Saarland hospitals was found to be between 1/100 and 3/100 patients. Between various departments, important differences were noted, and geriatric departments as well as ICU’s in Saarland reported the highest MRSA admission prevalence putatively due to differences in hygiene, antibiotic prescription and disease severity issues [11], [25]. Elevated admission prevalence (2.6/100) in both geriatric/rehabilitative and ICU wards (when compared to the mean prevalence of 1.4/100) was also reported in the Euregio study [43], yet this was clearly inferior to the prevalence observed in the respective departments in Saarland hospitals.

The Saarland study population also revealed relevant details with respect to the risk factors significant for MRSA carriage. Interestingly, risk factors with highest OR’s in our analysis as well as in the Euregio study were MRSA ‘history’ and a ‘skin’ condition, yet, in our study the associated relative risks were lower when compared to Euregio (with odd’s ratios of these two factors reported of 67 and 55, respectively). In the Euregio study, residency in a care institution (‘facility’) proved also to be highly associated MRSA carrier status (significant in our study only in univariate analysis), while ‘catheter’ as a medical condition was again relevant both in the Euregio as well as in the Saarland study. The other medical conditions significant in our study, ‘diabetes’, was not ascertained in the Euregio study, yet, ‘contact’ with an MRSA carrier was a significant factor in both studies. Moreover, in both studies, previous hospitalization (‘hospital’) was also a significant albeit weak risk factor, yet, it is reported by many admission patients, and its inclusion in a targeted screening algorithm greatly augments the number-to-screen. Interestingly, the use of antibiotics in the past 6 months (‘antibiotics’) – a significant risk factor in other studies [11] – has not been found to be a significant risk factor in multivariate analysis in either the Saarland or the Euregio study. Other factors such as occupational contact with meat products or contact with farm animals were not found to be relevant in the Saarland region; this is also in line with the finding of only one isolate attributable to the t011/ST398 genotype and typically associated with livestock-associated MRSA (this might be mostly due to the fact that economically this region is dominated by metallurgy and other industry, and pig farming associated with livestock MRSA is of negligible importance in this area).

This study has potential limitations. The relatively low prevalence of S. aureus in our study population raises the question on the sensitivity of our screening and test system. The major methodological difference to all other previous studies is the consequent use the flocked ESwabs [20], [21] in our study, precluding the enrichment step used in other studies [43], [47]. The pre-study we performed yielded S. aureus admission screening rates corresponding to previously reported rates [25], [43], [44], and being comparable between the direct ESwab chromogenic plating method and the enrichment of conventional swabs (of note, some MSSA positive samples were missed by the conventional method with cotton swabs either). Ultimately, the reason for the lower prevalence of S. aureus in the main study remains unclear, and can not easily be attributed to the quality of sampling (given written and web-based instructions on the correct swabbing technique and a great deal of enthusiasm and support for this study in the participating hospitals). Another limitation of our study may concern the quality of risk factor evaluation which was not controlled for, and might have been underreported particularly for risk factors whose ascertainment depends on proper evaluation of patient history (e.g. ’antibiotics’). Finally, it should be mentioned that the laboratory diagnostic concept of our study, i.e. performing all tests in a single centralized laboratory staffed 7/7 days, and responding on a rapid schedule, may not necessarily correspond to a ‘true life’ experience in other regions or institutions served by centralized microbiologic laboratory facilities involving extended transport and storage times.

On the other hand, our study is a ‘true life’ experience in that risk factor evaluation was obtained by the same personnel which will subsequently identify admission patients for targeted screening. Therefore, a clear limitation of the number of risk factors prior to decision for or against routine screening is urgently needed. The German recommendations for screening of MRSA admission patients [22] with altogether 11 risk factors, several of them difficult to evaluate, and 6 of them only to be applied if reported in combination may overstrain the hospital personnel’s capacities especially during the admission process. As a consequence of our study, we recommended the participating institutions to develop their own risk attributed screening policy based on the results of our analysis; yet, as a ‘default strategy’ we suggest to employ the six factors with highest and significant risk whose ascertainment and subsequent screening of less than one third of Saarland admission patients would allow for identification of more than two third of MRSA cases. This set of risk factors also appears appealing as all these six factors – with exception of ‘contact’ – can be ascertained in the admission ward based on readily available medical and social information (‘chronic care’, ‘diabetes’), hospital MRSA alert systems (‘history’), and simple inspection of the patient (‘skin’ and ‘catheter’).

In conclusion, in this prospective admission prevalence study, the prevalence of MRSA, and thus the epidemiologic burden by patients colonized upon entry into acute care hospitals could be determined for an entire German federal state of Germany. In conjunction with an analysis of the absolute and relative importance of ascertained risk factors, such information provided a rational basis for participating hospitals to develop an institution-individualized policy for targeted or universal screening. Moreover – and probably as importantly albeit difficult to evaluate – this prevalence screening has raised the overall awareness among administrators and hospital employees of participating institutions as well as of local policy makers and the general public to control and prevent colonization and infection with this notorious nosocomial pathogen. This intervention is part of an ongoing network activity in Saarland involving hospitals and community-based institutions to reduce the burden of MRSA and other multiresistant pathogens.

Acknowledgments

We are indebted to C. Bartsch, K. Bleses, H. Czelniak, W. Forsch, S. Freis, U. Geipel, S. Germer, J. Hans, T. Kassel, S. Loibl, H.J. Stuppy, S. Rauch, D. Velten, and all other members of the medical, technical and hospital hygiene team of the Institute of Medical Microbiology and Hygiene; to M. Damm, M. Stöckle and the team of the Department of Urology for excellent support during the pre-study; to all participating Departments with their nurses and physicians; and to the Management Board (Chair, W.I. Steudel) and the Administration of the University of Saarland Medical Center. Special thanks to CelTech (C. Igel, R. Sturm and coworkers) for IT and website support, and for preparation of electronic instruction material; to the Saarland chamber of physicians, the Saarland association of health insurance physicians and the Saarland hospital association for general study support; and to the Saarland Ethics Committee and the Saarland Data Protection Commissioner for valuable support and recommendations. In addition, our thanks extend to the participating Saarland hospitals (admission prevalence screening study correspondents in parenthesis): Lebach Caritas Hospital (Finkler), Dillingen Caritas Hospital (Finkler), Saarbrücken Caritas St. Theresia Hospital (Bienmüller), Mettlach Red Cross Hospital (Müller), Merzig Medical Center (Rahns), Saarbrücken Medical Center (Quintes), Püttlingen Knappschaft Hospital (Ruffing), Sulzbach Knappschaft Hospital (Schützmann-Kirsch), Saarlouis Red Cross Hospital (Schäfer), Saarbrücken Caritas St. Josef Hospital (Schwarz), St. Ingbert Hospital (Reichert), Saarlouis St. Elisabeth Medical Center (Schäfer), Kohlhof St. Josef Hospital (Schüler), Ottweiler Hospital (Meiser), Wadern St. Elisabeth Hospital (Steuer), St. Wendel Hospital (Schröder), Saarbrücken-Sonnenberg Hospital (Grimm), Völklingen Hospital (Momper), Saarbrücken Diakonie City Hospital (Bindert), Neunkirchen Diakonie Hospital (Bindert), Neunkirchen City Medical Center (Werner), Neunkirchen St. Josef Hospital (Schüler), Wallerfangen Hospital (Hovenbitzer), and to the health authorities of the State of Saarland.