Nosocomial Transmission: Methicillin-Resistant Staphylococcus aureus (MRSA)

Abstract

Nosocomial, or hospital-acquired , infections are an important cause of morbidity and mortality in health-care settings. Within hospitals, we find a gathering of patients with a weakened immune system, who receive all kinds of treatment that may even further weaken host defense mechanisms and that may break natural barriers against pathogens by surgery or by inserting intravascular lines. In such circumstances, even microorganisms that are generally considered harmless may cause fulminate infections.

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Nosocomial, or hospital-acquired , infections are an important cause of morbidity and mortality in health-care settings. Within hospitals, we find a gathering of patients with a weakened immune system, who receive all kinds of treatment that may even further weaken host defense mechanisms and that may break natural barriers against pathogens by surgery or by inserting intravascular lines. In such circumstances, even microorganisms that are generally considered harmless may cause fulminate infections. Such potentially pathogenic microorganisms are ubiquitous in health-care settings. They are continuously introduced by patients being admitted and can also be transmitted between patients. The latter is facilitated by the frequent contacts between health-care workers and patients, creating many opportunities for cross-transmission. Finally, because of the frequent use of antimicrobial agents, there is a selective advantage for pathogens resistant to these agents, and, hence, reported prevalence rates of antimicrobial resistant pathogens, which are more difficult to treat, are usually high in these settings.

Apart from these factors, there are several other factors, which clearly distinguish nosocomial infections from other infections.

• The concept of colonization . Typically, infections with nosocomial pathogens are preceded by colonization (i.e., the presence of the microorganism at a body site without signs of infection) with the same pathogen. Since only a fraction of colonized patients will develop overt infection, dynamics of nosocomial pathogens are mainly determined by patients being colonized instead of infected. It is estimated that the ratio between colonized and infected patients is at least 10:1.

• Population size. The typical size of hospital units is small (5–20 beds) and, therefore, natural fluctuations in the prevalence of colonization are high. These fluctuations, together with dependency between colonized patients due to transmission, complicate the analysis of preventive interventions.

• Patient dynamics. The length of stay of patients in hospitals (or in certain units) is short (days), which leads to a high turnover of patients.

• Readmission of patients. Many patient categories, e.g., haemodialysis patients, are frequently readmitted. If colonization persists after discharge (which holds for most nosocomial pathogens), there will be frequent re-introductions of pathogens.

• External interactions. Units (and hospitals) are not isolated and will be in contact, through transfer of patients, with other units (or hospitals).

2 The Burden of Disease

An abundance of data demonstrates the high prevalence of hospital-acquired infections. In the United States these infections now rank among the leading causes of death, with conservative estimates of 260,000 years of life lost each year from premature deaths directly related to nosocomial bloodstream infections alone. Such infections lead to billions of dollars of expenditures and to an unquantified diminution of the quality of life for thousands of patients each year (Wenzel 2003). Actual estimated figures are 1.7 million individuals in the United States acquiring a nosocomial infection, yielding 100,000 deaths (Klevens et al. 2007a) and an additional $6.5 billion in health-care expenditures (Stone et al. 2005). Within hospitals, Staphylococcus aureus is the most frequent cause of opportunistic infections and the proportion of these infections caused by MRSA has increased rapidly in most developed countries in the last decade. More than 18,000 deaths were estimated to have occurred among patients with invasive MRSA infections in the United States during 2005, and most of these infections had been acquired in health-care settings (Klevens et al. 2007b).

2.1 Staphylococcus aureus and MRSA

S. aureus is a bacterium colonizing many healthy persons. The predilection site for colonization is the anterior nares. For unknown reasons, about 25% of healthy people are always found to be colonized and another 25% appear to be resistant to nasal colonization with S. aureus. The remaining proportion is intermittently colonized. In the hospital, S. aureus is the most important bacterial cause of hospital-acquired infections, such as postoperative wound infections and pneumonia. In most cases, these infections are caused by the bacteria that could initially be detected in the nose of the patient.

At the time that penicillin was introduced for clinical use, in the 1940 s, all S. aureus isolates were susceptible to penicillin. However, within 5 years almost all isolates causing infections in hospitalized patients had become resistant to penicillin, due to the bacterial production of enzymes that destroy penicillin (beta-lactamases). In the 1950 s a new class of penicillin antibiotics was developed that was resistant to these beta-lactamases, the first being methicillin. Within several years, though, methicillin-resistant S. aureus (MRSA) isolates were described as well. Now resistance did not result from production of beta-lactamases, but from a new protein in the bacterial cell wall that prevents the interaction of the antibiotic with the bacterium. As a result, MRSA are resistant to all beta-lactam antibiotics.

During the 1980s, MRSA became an important cause of nosocomial infections in virtually all developed countries. In this chapter, we focus on the transmission dynamics and modelling of MRSA in hospital settings. For simplicity we call these strains hospital-acquired MRSA (HA-MRSA). Recently, we have seen a rapid emergence of infections caused by MRSA strains in the community, especially in the United States, as well as among animals in Europe with secondary spread to professional caretakers. From an epidemiological point of view, these so-called community-acquired MRSA (CA-MRSA) have different characteristics. The modelling as described in this chapter, therefore, does not automatically apply to CA-MRSA.

2.2 Transmission Routes

Within a health-care setting, several different bacterial transmission routes can be distinguished:

• Airborne transmission . This occurs through aerosols after coughing or sneezing. This route is relevant for respiratory viral diseases such as influenza and severe acute respiratory syndrome (SARS) but it seems to be of little relevance for antibiotic-resistant bacteria such as HA-MRSA.

• Environmental contamination. Although this may occur for MRSA, its role in the epidemiology of nosocomial MRSA is unclear. There are reported outbreaks due to contaminated surfaces or equipment, but in most instances, the contribution of contamination to the transmission dynamics is unknown. Moreover, there are no studies demonstrating that a single intervention reducing environmental contamination as a sole measure reduces acquisition with MRSA by patients (Dancer 2008).

• Direct transmission . This can occur when a health-care worker is persistently colonized with MRSA and acts as a constant source for non-colonized patients. Such occasions have been documented for MRSA, but these are considered to occur only sporadically.

• Indirect transmission , usually through temporarily contaminated hands of health-care workers. This transmission route, usually considered most important for HA-MRSA, will be discussed in detail below.

2.3 Indirect Transmission of MRSA

Health-care workers (more specifically their hands) are considered important in the nosocomial transmission of HA-MRSA. In this respect, health-care workers act as vectors and they spread pathogens after direct contact with a colonized patient followed by a contact with an uncolonized patient (Grundmann et al. 2002). The dynamics of this process mimics the dynamics of malaria, with health-care workers acting as mosquitoes, and is described in the Ross–MacDonald model (Ross 1911; MacDonald 1957) (Figure from Austin et al. 1999). This model was, more or less simultaneously, applied to describe the epidemiology of nosocomial pathogens by Cooper et al.(1999) and Bonten et al. (2001), and later by Grundmann et al.(2002).

In this theoretical framework, patients can be admitted and discharged and on admission patients are either uncolonized or colonized. Uncolonized patients can only acquire colonization after a contact with a contaminated health-care worker. Health-care workers can become contaminated after contact with a colonized patient. With adequate hand disinfection, a contaminated health-care worker may clear contamination, and become uncontamined again. From this model, it becomes immediately clear that improved hand hygiene (through either more frequent or more effective cleaning) will reduce the number of contacts between contaminated health-care workers and uncolonized patients, and, hence, reduce the spread of HA-MRSA. This underscores the prominent role that improved hand hygiene campaigns nowadays have in many hospitals worldwide.

Another possibility to reduce the number of contacts between colonized patients and uncontaminated health-care workers as well as between uncolonized patients and contaminated health-care workers is cohorting . The level of cohorting can be quantified as the likelihood that a next health-care worker’s contact will be with the same patient. In the extreme case (cohorting level is 1) each health-care worker contacts only a single patient and colonization cannot be transmitted from one patient to another. The level cohorting has been determined in intensive care units in the United States, the United Kingdom and the Netherlands, and was approximately 70% in all (Grundmann et al. 2002; Nijssen et al. 2005; Bonten unpublished data). The importance of a high level of cohorting underscores the need of adequate staffing levels. In general, the number of contacts needed by patients depends on the severity of their disease, and should be seen as a given parameter that is difficult to change (without compromising the quality of patient care). A reduction in the number of staff, with a similar need of patient contacts, may well reduce the level of cohorting as nurses need to assist more frequently in the care of different patients. On top of that, the increased working load for nurses may also reduce the frequency (and appropriateness) of hand hygiene, which will even more facilitate transmission.

The likelihood that a health-care worker acquires contamination with HA-MRSA on his/her hands depends on the fraction of patients in the unit being colonized. This parameter has been termed colonization pressure (Bonten et al. 1998). Even when compliance of hand disinfection is far from optimal (in both frequency and efficiency), the rate at which the hands of health-care workers become decolonized is much higher than the rate at which the colonization pressure is changing, i.e. health-care workers will successfully clear contamination several times per shift. This is confirmed by studies where, even for high colonization pressure, contamination on the hands of health-care workers was only seen infrequently (Grundmann et al. 2002). The risk of an uncolonized patient acquiring colonization depends on the prevalence of hand contamination of health-care workers. However, due to the high decolonization rate of health-care workers, the probability that the hands of a given health-care worker are contaminated at a certain time is proportional to the colonization pressure, i.e., a doubling in the colonization pressures doubles the probability that the hands of a given health-care worker are contaminated. Therefore, from a mathematical point of view, the risk for an uncolonized patient of acquiring colonization can be described by the colonization pressure and health-care workers need not be modelled explicitly. In other words, the dynamics with health-care workers will resemble a situation without health-care workers where patients have direct contacts with each other. The frequency of such ‘direct contacts’ between patients depends of course on the frequency of contact between health-care workers and patients and on the efficacy of hand disinfection.

Environmental contamination is not part of the theoretical framework depicted in Fig. 22.1. If a colonized patient contaminates his or her direct inanimate environment (bed, washbasin, or air due to sneezing) and this contamination is only short-lived (for instance, because the environment is successfully cleaned after patient discharge) we can consider the inanimate environment as part of the patient. Contamination of a health-care worker then still fulfills the assumptions of indirect “patient-to-patient” transmission. However, if environmental contamination persists longer, the risk of an uncolonized patient to acquire colonization no longer depends on the actual number of colonized patients in the unit. In such a scenario, indirect “patient-to-patient” is no longer an accurate description of the transmission dynamics.

Fig. 22.1

Model of the transmission dynamics of HA-MRSA

This distinction between transmission routes that do (patient dependent), and that do not depend (patient independent), on the colonization pressure is extremely important. Examples of patient-independent transmission routes are transmission due to persistently colonized health-care workers, colonized (or contaminated) visitors, and persistent contamination of the inanimate environment. Naturally, efficacy of intervention measures will differ for both scenarios, but even more importantly, differences in dynamics become obvious in statistical analysis. If patient dependency is relevant, dynamics become nonlinear and standard statistical tests that are usually applied (e.g., chi square tests or student t-tests) for intervention studies are no longer appropriate (Nijssen et al. 2006). Therefore, distinction between both transmission routes, and determination of their relative contribution, is necessary when interpreting epidemiologic data and evaluations of interventions. Genotyping to determine events of cross-transmission and mathematical methods (Bootsma et al. 2007; Cooper 2007; Cooper and Lipsitch 2004; Drovandi and Pettitt 2008; Forrester and Pettitt 2005; McBryde et al. 2004; McBryde et al. 2007; Mikolajczyk et al. 2007; Pelupessy et al. 2002) are available to distinguish between the two.

For clarity, the models described in this chapter focus on nosocomial pathogens for which colonization pressure is relevant, as is the case for HA-MRSA and vancomycin-resistant enterococci (VRE).

2.4 Intervention Strategies

The aims of intervention strategies for HA-MRSA are to reduce the number of infections with MRSA, by reducing the prevalence of HA-MRSA carriage. Intervention strategies can be divided into those which work for many pathogens simultaneously and pathogen-specific strategies. General strategies generally include improvement of hygiene measures, such as hand hygiene (hand washing with an alcohol-based rub instead of water and soap), usage of gloves and gowns, better environmental cleaning, and cohorting of health-care workers.

Pathogen-specific intervention measures typically aim to physically separate colonized and uncolonized patients, by treating colonized patients in isolation rooms, single bed rooms, or by cohorting these patients (Cooper et al. 2003). Although such a separation in theory should be effective in reducing spread of pathogens, several clinical studies failed to confirm this (Cepeda et al. 2005; Cooper et al. 2003).

Detection of carriage is an important aspect for the success of such an intervention strategy. The simplest way is to use bacteriological cultures that have been obtained for clinical reasons. However, as could be assumed from the high colonization–infection ratio, most colonized patients will not develop infections and may, therefore, not be subjected to clinical testing. This problem can be overcome by active screening, either of all patients or of a high-risk population. Active screening means obtaining cultures to detect certain pathogens even in the absence of clinical suspicion of infection. Such strategies could include weekly surveillance of all patients, screening on admission of all patients, of those with a history of colonization or other risk factors for colonization. The faster colonized patients are isolated, the more effective this measure will be. Therefore, the best time to screen suspected carriers is directly after admission (Robotham et al. 2006). One should also realize that there is a diagnostic delay between obtaining the culture sample and the microbiological result, during which colonized patients would not be treated in isolation. For HA-MRSA, conventional microbiological techniques have a diagnostic delay of up to 5 days for patients not carrying HA-MRSA. Recent advances in diagnostic techniques have reduced this diagnostic delay to several hours, although these tests still have delay of about 1 day in routine daily practice (Harbarth et al. 2006). Preventive isolation of patients at high risk for colonization until culture results have excluded such carriage (or documented it) could be used to minimize the likelihood of transmission during the period of unrecognized carriage. However, if there is a trade-off between accuracy and rapidness of a test, the effect of rapid tests depends on the screening strategy (Bootsma et al. 2006; Raboud et al. 2005).

3 Mathematical Models for the Dynamics of Hospital-Associated MRSA

In the last decade, a considerable number of mathematical models to describe the spread of nosocomial pathogens have been published (Austin et al. 1999; Boldin 2007; Bonten et al. 2001; Bootsma et al. 2006; Cooper et al. 1999; Cooper et al. 2004; D’Agata et al. 2005; Grundmann et al. 2002; Lipsitch et al. 2000; McBryde et al. 2007; Robotham et al. 2006; Robothamet al. 2007; Sébille and Valleron 1997; Sébille et al. 1997; Smith et al. 2004; Smith et al. 2005). With these models investigators aimed to highlight key aspects of pathogen dynamics, as well as to predict the effect of intervention measures.

For models that want to go beyond qualitative relationships between parameters, two model ingredients are essential. The first one is that the model should be stochastic because hospital units are small with large fluctuations in prevalence (Bootsma et al. 2006; Cooper et al. 1999; Cooper et al. 2004; Raboud et al. 2005; Sébille and Valleron 1997; Sébille et al. 1997). The second is that readmission of discharged patients should be taken into account. The average number of secondary cases per primary case per admission if all other patients are uncolonized (the R _A -value ) (Cooper et al. 2004) is typically below 1. This implies that after the first introduction of HA-MRSA in a hospital, there may be a small outbreak, but it is unlikely that there is a major epidemic shortly afterwards in which a substantial fraction of the hospitalized patients is involved. However, a patient who was colonized at the time of hospital discharge can still be colonized when readmitted. So, although the spreading capacity of HA-MRSA per admission is insufficient to generate an epidemic, the readmission loop (or feedback loop) ensures that the average total number of secondary cases per primary case per infectious period, i.e. the R ₀-value , is above 1 (Bootsma et al. 2006; Cooper et al. 2004; Robotham et al. 2007).

Almost all models assume that the probability of an uncolonized patient acquiring colonization is proportional to the colonization pressure. As mentioned earlier, this mass action assumption is correct when health-care workers are vectors of transmission, and as long as contamination of their hands is short-lived. However, although mass action may be a reasonable assumption for the dynamics within a single ward, it is not a good description of the pathogen dynamics in a hospital with multiple wards. In general, most health-care workers are confined to a single ward, and some, e.g., physicians, move between wards. The compartmental structure of a hospital has several consequences for the dynamics. First, depletion of susceptible patients occurs much more rapidly in a small unit, which may reduce the growth of an epidemic. Second, when outbreaks occur within a single unit, intervention measures can be restricted to the affected unit, although such a local outbreak may exceed the local capacity to deal, e.g. because no isolation beds are available, with possible subsequent spread to other wards.

Another feature of the models is that all uncolonized patients are assumed to be equally susceptible and all colonized individuals to be equally infectious. The main reason for this assumption is simplicity and incorporation of heterogeneity in susceptibility and infectivity is challenging. Factors that influence susceptibility are known from clinical trials, the factors that increase infectivity, however, are less known. More importantly, correlations between infectivity, susceptibility, readmission rates, and decolonization rates are hardly known. However, variation in susceptibility and infectivity typically leads to smaller outbreaks (Kuulasma 1982) and a model with equal susceptibility and infectivity can therefore be seen as a worst-case scenario.

3.1 Insights Derived From Mathematical Models of Hospital-Associated MRSA

Here, we briefly describe the most salient observations in some models addressing the nosocomial epidemiology of HA-MRSA. Of note is that all models stress the importance of hand hygiene and cohorting, but these were not specifically investigated in any of the models (only by considering changes in the transmission parameter).

Grundmann and coworkers fitted observational data of MRSA colonization in an intensive care unit during a 1-year period to the previously described model of vector-borne indirect transmission (Grundmann et al. 2002). During the study period there were two outbreaks, and these were linked to temporal shortage of nursing staff. For most periods of the year the calculated R _A -value of MRSA was <1, but this increased to values just above 1 at the time of outbreaks. Overall, it was estimated that the R _A -value of MRSA (in the absence of any control measures) was close to 12 and that infection control measures as used had reduced this value to <1 for most of the time.

Several researchers have explicitly addressed the relation between antibiotic consumption and selection of resistant strains (Bergstrom et al. 2004; Boldin 2007; Bonten et al. 2001; D’Agata et al. 2005; Lipsitch et al. 2000; Stewart et al. 1998), and all demonstrate a direct relation between the amount of antibiotic consumption and the prevalence of antibiotic-resistant strains. However, all models assume that there is a fitness cost of resistance . This implies that, in the absence of the selective pressure of antibiotics, susceptible bacteria will have a growth advantage over the resistant ones and that in time carriage with resistant strains will disappear. Although this concept has been assumed for decades, there is increasing evidence that this paradigm of a fitness cost due to resistance is not always correct. In the absence of fitness costs of resistance, prevalence levels of resistant strains will not decline or decline much slower after withdrawing the selective pressure of antibiotics.

Cooper et al. (2004) used a Markovian stochastic simulation model to mimic the dynamics of MRSA both in the hospital as well as in the catchment population outside the hospital. They were the first to stress the importance of readmission of patients still colonized with MRSA for the nosocomial dynamics of MRSA. Especially, patients with frequent hospital admissions are of pivotal importance. Cooper et al. also demonstrated that when infection prevention resources (here the number of isolation beds ) are insufficient to control an outbreak, the efficacy of such measures decreases rapidly. Finally, they nicely demonstrated the importance of stochastic effects . With the same parameters, MRSA could be controlled (even without extensive control measures) for years, but catastrophic failure resulting in high endemic prevalence levels might occur as well. Their findings advocate the use of infection preventions strategies before a pathogen becomes a major problem.

In another study, a more “game-theory like” approach was used (Smith et al. 2004; Smith et al. 2005). In order to preserve or reach a low nosocomial prevalence rate of MRSA carriage or infection, it would be optimal if all hospitals implemented intervention measures. However, for an individual hospital, it may be cost-effective not to implement these costly measures, as long as the other hospitals do (the so-called tragedy of the commons). In return, the efficacy (and thus cost-efficacy) will be lower for the hospitals with implemented measures. This exemplifies the need and the economic benefits of coordinated actions for controlling the spread of nosocomial pathogens, such as MRSA. In the extreme, the outcomes of this model would even justify legal enforcement to comply with such measures.

We have also used a Markovian stochastic simulation model to investigate the dynamics of MRSA in hospitals and their catchment population, as well as the effects of different intervention measures hereon (Bootsma et al. 2006). The main difference with the Cooper model (Cooper et al. 2004) is the much more detailed modelling of the transmission dynamics within the hospital. This includes a compartmentalized structure due to wards, two different types of wards (with regard to transmission and patient flow), the presence of health-care workers persistently carrying MRSA (and being a continuous source for patients) as well as health-care workers temporarily contaminated with MRSA and acting as vectors for MRSA transmission. This more detailed modelling structure mimics reality more closely and allows for the evaluation of the contribution of the individual components of the so-called Search and Destroy policy. This policy consists of a set of infection prevention measures and is thought to be responsible for the low prevalence of MRSA among nosocomial S. aureus infections in Scandinavian countries and the Netherlands. The “Search and Destroy” policy aims at treating MRSA carriers in single rooms with barrier precautions. On admission, high-risk patients are screened for MRSA carriage and preventively isolated until documented absence of such carriage. When MRSA is unexpectedly found in a patient who is not treated in isolation, all patients and health-care workers in the same ward or that otherwise might have had contact with the index case are screened for carriage. Finally, eradication of MRSA (in those identified as infected or colonized) through nasal application of mupirocin (in combination with chlorhexidine body washing) is attempted after hospitalization.

First we analyzed how the cumulative use of these measures will benefit control of MRSA in a low-endemic setting . Isolation of MRSA carriers that are detected by clinical cultures alone will not be sufficient to control MRSA in the long term, even if isolation is 100% effective. However, it might take many years before the nosocomial prevalence of carriage will start to increase. With isolation we mean all measures that limit the transmission of MRSA to another patient. The highest likelihood to achieve this will be reached when such a patient is placed in a single room, with limited exposure to health-care workers that adhere maximally to barrier precautions. This may not be feasible in many circumstances, though, and therefore, we have used different efficacy levels for isolation in our model. In a low-endemic setting , addition of either admission screening and preventive isolation of high-risk patients or active search for carriage among contact patients in case of an unexpected case to isolation of carriers seems sufficient to control MRSA. Yet, in the analytical model perfect performance of these combined measures would only be sufficient to control a pathogen with an R value of 1.4–1.5. Considering that such measures will never be executed perfectly, application of both screening interventions (on admission and after detection of an index patient) simultaneously is much safer. In the combined approach, the R value that could be controlled successfully increased to 2.2. In such a scenario, active decolonization after hospital discharge would only add little to its effectiveness.

Second, we analyzed how a modified Search and Destroy approach would perform in a setting with a high prevalence of MRSA carriage in the hospital. Considering the problems of realizing perfect isolation in such setting, for instance, because of shortage of isolation facilities, two strategies were evaluated. In the first scenario, three measures were implemented (isolation of all documented carriers; screening of high-risk patients on admission with preemptive isolation; screening of contact patients when an index patient was identified). Because of anticipated problems in achieving appropriate isolation in the beginning (when prevalence of carriage is high) we assumed that efficacy of colonization would be 50% in the first 5 years and would then increase to 80% thereafter. From Fig. 22.2, it is obvious that the combined approach of all three measures is more effective than the other combinations. In the second scenario, all efforts were placed on achieving good isolation (80% efficacy) of documented carriers in the first 5 years, with the addition of screening high-risk patients and contacting patients after this period (while maintaining 80% efficacy of isolation). Again, the combined approach of all three measures was most efficient. This scenario analysis can also be used to determine the burden of infection prevention for the hospital, expressed as the number of isolation beds needed. This is represented by the area under the curves. Accepting the assumptions made, there were marked differences in the number of isolation beds needed, which might be helpful in choosing the most cost-effective strategy.

Fig. 22.2

Mean nosocomial prevalence in 1000 simulations. Left: The isolation efficacy is 50% during the first 5 years and 80% thereafter for three intervention scenarios. Admission screening of high-risk patients (HR), Screening of contact patients (contact) and a combination of both. Right: The first 5 years only documented carriers are isolated with an efficacy of 80%. After 5 years, additional interventions are implemented.

Finally, the effects of rapid screening methods for MRSA were compared to conventional microbiological culture methods. Conventional microbiological methods have a diagnostic delay of up to 5 days for excluding MRSA carriage and of several days (up to 3) for detecting MRSA. This diagnostic delay can be reduced, in theory, to several hours, although a turn-around-time of about 1 day has been realized in clinical trials (Harbarth et al. 2006). Rapid diagnostic tests can prevent unnecessary isolation days when using preventive isolation. When not using preventive isolation, rapid diagnostic tests can reduce the duration of non-isolation of colonized patients. Therefore, in theory, rapid diagnostic tests must be beneficial for controlling MRSA if these tests have equally good sensitivity and specificity as the conventional microbiological culture methods. Yet, rapid tests will be more expensive and these additional costs must be balanced against these benefits.

Several interesting scenarios arise if the test characteristics of the rapid test are different from those of conventional microbiological tests. Occurrence of false negative results ( sensitivity <100%) will prevent colonized patients from being isolated, which may increase transmission (and nosocomial prevalence) and will, thus, determine whether the use of rapid test reduces the nosocomial prevalence. As prevalence may grow only slowly, rapid tests may initially seem beneficial, but may ultimately increase the number of patient days in isolation. However, our model predicts that rapid tests will reduce the number of isolation days required in high-endemic settings, even if test characteristics are not optimal.

Going back to the two proposed scenarios for gradually increasing MRSA control in high-endemicity settings, we have evaluated to what extent the use of rapid diagnostic tests can reduce the number of isolation beds needed. Depending on the turn-around-time of test results and their specificity this number could be reduced by as much as 47%, which might make such control strategies feasible, even in high-endemicity settings.