Aspects of hospital infection control have achieved a great deal of interest from the media, the public and politicians in many countries and many hospitals worldwide are very active in improving the situation. Meanwhile, a considerable body of knowledge about the effectiveness and suitability of specific infection control measures has been established. Comprehensive guidelines have been developed in many countries to support the infection control recommendations in individual hospitals. However, knowledge about the best infection control measures is often not the most important point in improving the situation. In many hospitals compliance with these recommendations is far from ideal and barriers for low compliance must be identified to overcome them.

One of the most interesting infection control problems is the MRSA (Methicillin resistant Staphylococcus aureus) problem. Many hospitals have introduced surveillance systems to monitor MRSA import and the development of nosocomial MRSA cases. National surveillance systems support them by providing definitions, protocols and reference data.

The German national nosocomial infection surveillance system (KISS) has established a hospital-wide surveillance system in 2003 (pilot year) which is called MRSA-KISS to track MRSA cases on the hospital level [1]. The hospitals provide data about imported and hospital-acquired MRSA cases once a year, and the denominator data are the number of patient-days of the whole hospital. Reference data are stratified by hospital size and screening frequency and published once a year.

Further extensive surveillance of MRSA cases is performed on the unit level within the ICU surveillance component. The ICUs have been required since 2003 to provide information about all ICU-acquired and imported cases (infections and colonizations) of multidrug resistant organisms (multidrug resistant organisms (MDRO), such as MRSA, VRE, ESBL) in addition to and independently from surveillance of lower respiratory tract infections, primary bloodstream infections, urinary tract infections and meningitis/ventriculits [2, 3]. During ICU-MDRO surveillance the units enter this information monthly on a unit-by-unit basis in a web based surveillance system and are able to generate analyses of their data at any time.

Meanwhile we have an overview about more than 6 years in both MRSA surveillance systems. The objective of this study therefore was to analyze which surveillance system (hospital-based or unit- based) leads to a greater decrease in incidence density of newly acquired MRSA cases.

Two cohort studies of surveillance data were conducted. Data from a total of 224 hospitals providing hospital based nosocomial MRSA incidence rates and 359 ICUs providing unit based nosocomial MRSA incidence density rates analyzed with data from the years 2004 to 2009. Data from 2003 were excluded because this year was a pilot phase for both surveillance systems.

Annual incidence density was calculated (i.e., the number of MRSA cases standardized per 1000 patient days) to describe the change of MRSA cases over the years. MRSA cases in both surveillance systems were considered imported when MRSA reports from prior admissions or discharge reports or microbiological results from surveillance cultures or clinical specimens taken within the first 48 hours following admission were available. The screening frequency was defined as the number of nasal swabs taken per 1000 patient days, whereby only the first swab per patient was counted. The term "nosocomial was applied to hospital acquired cases in MRSA-KISS (case is acquired for the hospital following admission to that hospital) and to ICU-acquired cases (a case is acquired for an ICU following admission to that ICU) from data of the MDRO component of ICU KISS.

The development of incidence density over time (calendar years) was described first for both surveillance components. All participating hospitals or ICUs were included. Because a surveillance effect can be expected after at least four years of surveillance, the data from those hospitals/ICUs with continuous participation for at least four years were analyzed in a second step. Only the data from the first four years of participation were included in the analysis.

Incidence rate ratios (IRR) with 95% confidence intervals were calculated to compare incidence densities for hospital- and ICU-acquired cases between the different time intervals. SAS 9.2 (SAS Institute Inc., Cary, NC, USA) PASW Statistics 18 and EpiInfo version 6.04 were used for analysis.

National guidelines for MRSA prevention were published in Germany in 1999 [4] and each hospital must record the occurrence of MDRO such as MRSA since 2001, but without mandatory reporting [5]. German microbiology laboratories have been participating in the European Antimicrobial Resistance Surveillance System (EARSS) and have provided the MRSA percentage of S.aureus bacteremia isolates for many years. Since 2008 a more comprehensive surveillance system for MDRO from laboratory data has been established (ARS = Antibiotika Resistenz-Surveillance, https://ars.rki.de). In 2009 a mandatory reporting of MRSA bacteremia cases was introduced [6]. MRSA-KISS and the MDRO component of ICU-KISS were established in 2003 as voluntary surveillance systems. In 2009 about 10% of all hospitals were participating and about 15% of all ICUs.

Compared to the data from other countries the incidence density of nosocomial MRSA in German hospitals and ICUs is lower (especially considering that most publications report only nosocomial MRSA infections, not nosocomial MRSA cases (i.e. including colonized as well as infected patients) [7–9]. But of course there are also countries with lower MRSA rates. One example is the report from 38 French hospitals describing a decrease of the MRSA incidence density from 1.16 to 0.57 per 1000 hospital days between 1993 and 2007 [10].

Intensive care units participating in the MDRO component of ICU-KISS were able to achieve a significant decrease of ICU-acquired MRSA cases of 29% during a four-year surveillance period. This is in accordance with similar observations concerning the influence of surveillance activities on the development of nosocomial pneumonia and primary bloodstream infections in ICU-KISS. Significant reductions of infection rates of between 14 and 29% were demonstrated during various periods of analysis [11–14]. On the unit level, the surveillance staff together with the ICU staff can analyze the data at any time, present the information to the ICU staff and stimulate discussions to analyze reasons for infection control problems and to introduce the most appropriate interventions. Already in the second year of surveillance a significant effect was observed with further improvement in the following years.

However, no decrease in hospital-acquired MRSA rates on the hospital level was found. Normally, MRSA rates on the hospital level are presented during meetings of the hospital infection control committees. But, even if the hospitals have hospital-acquired MRSA rates above the median, the hospital committees are often not able to identify reasons for this situation and draw the most appropriate conclusions. A similar development was demonstrated within the other hospital-based surveillance system in KISS, the surveillance system for Clostridium difficile associated diseases (CDAD-KISS), where we also did not find a reduction of nosocomial cases within the first 3 years of participation. Perhaps the infection control staff has focused its activities on the ICU within their hospitals because they know that the highest MRSA rates can be observed on the ICUs and they are used to work in the field of quality management with the ICUs due to the experience from ICU KISS.

Moreover, many hospitals increased their admission screening frequency remarkably during the observation period. Without admission screening, cases with an MRSA positive microbiology report from day 3 on will be automatically considered hospital-acquired or ICU-acquired according to the MRSA-KISS and ICU-MDRO protocols. With admission screening, on the other hand, many cases will be classified as imported which would not be the case without admission screening. Due to this reclassification the constant hospital-acquired MRSA incidence in MRSA-KISS hospitals with continuous participation will be in reality associated with an increase of hospitals-acquired cases. For the ICUs we do not have the data about the development of admission screening. But we also expect an increase over the years along with the increase in the entire hospitals. Therefore the observed 29% decrease of the ICU-acquired MRSA incidence density has also to be regarded-at least in part-in the light of this re-classification bias.

Unfortunately we do not have comparable data from hospitals without continuous MRSA surveillance in Germany. The only data source available is the above mentioned ARS database, where the percentage of MRSA among S.aureus strains is provided for a large number of hospitals over time. According to this database a further increase of the percentage of MRSA was observed between 2008 and 2010 https://ars.rki.de.

Our data lead to the conclusion that a unit based surveillance approach (at least in the ICUs) is more useful for reducing nosocomial (i.e.acquired) MRSA rates compared to a hospital based approach. From the hospital level, the infection control staff has to identify the most problematic units and to provide a feedback on the unit level in order to stimulate appropriate interventions on the unit level.