The clinical and cost effectiveness of screening for meticillin-resistant Staphylococcus aureus (MRSA)

Transcription

1 ~ October 2007 This document can be viewed on the NHS QIS website. It is also available, on request, from NHS QIS in the following formats: electronic audio cassette large print. NHS Quality Improvement Scotland Glasgow Office - Delta House 50 West Nile Street Glasgow G1 2NP Tel Edinburgh Office - Elliott House 8-10 Hillside Crescent Edinburgh EH7 5EA Tel comments@nhshealthquality.org This document is produced from elemental chlorine-free material and is sourced from sustainable forests The clinical and cost effectiveness of screening for meticillin-resistant Staphylococcus aureus (MRSA) Health Technology Assessment Report 9 ~ October 2007 The clinical and cost effectiveness of screening for meticillin-resistant Staphylococcus aureus (MRSA)

2 Health Technology Assessment Report 9 The clinical and cost effectiveness of screening for meticillin-resistant Staphylococcus aureus (MRSA) Authors: Ritchie K, Bradbury I, Craig J, Eastgate J, Foster L, Kohli H, Iqbal K, MacPherson K, McCarthy T, McIntosh H, Nic Lochlainn E, Reid M, Taylor J

3 NHS Quality Improvement Scotland (NHS QIS) is committed to equality and diversity. We have assessed this report for likely impact on the six equality groups defined by age, disability, gender, race, religion/belief and sexual orientation. For a summary of the equality and diversity impact assessment, please see our website at: The full report in electronic or paper form is available on request from the NHS QIS Equality and Diversity Officer. NHS Quality Improvement Scotland 2007 ISBN First published October 2007 NHS Quality Improvement Scotland (NHS QIS) consents to the photocopying, electronic reproduction by uploading or downloading from the website, retransmission, or other copying of this Health Technology Assessment for the purpose of implementation in NHSScotland and educational and not-for-profit purposes. No reproduction by or for commercial organisations is permitted without the express written permission of NHS QIS.

4 CONTENTS 1 EXECUTIVE SUMMARY 1 2 INTRODUCTION 5 3 BACKGROUND ON NHS QUALITY IMPROVEMENT SCOTLAND 6 4 SETTING THE SCENE Description of MRSA in Scotland Recent policy initiatives Description of technology MRSA screening services in Scotland Objective and focus of HTA 10 5 CLINICAL EFFECTIVENESS Introduction Methods Literature search Results Prevalence rates Study selection Assessment of the evidence Model estimates for MRSA prevalence MRSA detection using clinical risk assessment Study selection Assessment of the evidence Model estimates for MRSA detection using clinical risk assessment Laboratory tests Study selection Assessment of evidence Model estimates for laboratory tests for MRSA Decolonisation Study selection Assessment of evidence Model estimates for mupirocin decolonisation clearance rates Association between colonisation and infection Study selection Assessment of the evidence Model estimates for MRSA colonisation and infection Transmission rates Isolation of patients and MRSA control Study selection Assessment of the evidence Staff screening Discussion 51 6 ECONOMIC EVALUATION AND MODELLING Introduction Evidence sources Literature search Study exclusion criteria Overview of studies Economic model Methodology MRSA screening Hospital setting MRSA screening strategies Patient management Model structure The patient care pathway Hospital setting 60 iii

5 6.3.4 Model inputs Model costs Other model parameters Model limitations Results Summary of strategies Results for a tertiary referral hospital Results for a large general hospital Sensitivity analyses Financial implications Discussion Conclusions 89 7 PATIENT ISSUES Introduction Methods Literature search Focus groups Results Patient literature review Patient views of MRSA and of isolation procedures Psychological implications of MRSA colonisation Experience of isolation nursing Public understanding of MRSA Focus groups Patient focus groups Staff literature review Staff compliance with MRSA control procedures Staff attitudes and views Staff focus group Patient consent Preadmission screening Patient isolation Discussion and conclusions 96 8 ORGANISATIONAL ISSUES Introduction Current policies and practice of MRSA screening in Scotland Screening issues Legal and ethical issues Building and site issues Staff resources and training Information management of the technology Quality assurance and audit Organisation of the service Information needs Guidelines on MRSA Discussion and conclusions DISCUSSION AND CONCLUSIONS Background to HTA Main findings Limitations Challenges for implementation Patient and public considerations Need for further research Conclusions RECOMMENDATIONS ACKNOWLEDGEMENTS REFERENCES 109 iv

6 LIST OF APPENDICES Appendix 1 Topic group members, peer reviewers and other contributors 121 Appendix 2 Strategy for literature searches 122 Appendix 3 Included studies 129 Appendix 4 Excluded studies 135 Appendix 5 Staff screening: excluded studies 140 Appendix 6 Patient pathways for all screening strategies 141 Appendix 7 Economic input data 147 Appendix 8 Tabulated results 152 Appendix 9 Financial implications for a large general hospital 160 Appendix 10 MRSA questionnaire 164 LIST OF TABLES Table 5-1 Summary of literature searches 11 Table 5-2 Prevalence of MRSA colonisation on hospital admission 13 Table 5-3 Characteristics of clinical risk assessments for MRSA colonisation on admission 15 Table 5-4 Relationship between the number of clinical risk factors and the diagnostic accuracy of clinical risk assessment 17 Table 5-5 Characteristics of culture method studies 19 Table 5-6 Characteristics of PCR method studies 21 Table 5-7 Laboratory test data 23 Table 5-8 Characteristics of decolonisation studies 25 Table 5-9 Colonisation and infection studies 34 Table 5-10 Studies of isolation or nurse cohorting to control MRSA 43 Table 5-11 Reviews of staff screening for MRSA 48 Table 5-12 Primary studies of staff screening for MRSA 50 Table 6-1 Laboratory test turnaround times 55 Table 6-2 Categorisation of hospital specialities 56 Table 6-3 MRSA screening strategies 57 Table 6-4 Scottish hospital categories 60 Table 6-5 Percentage of bed in high, low-risk and other specialties 60 Table 6-6 Tertiary referral hospital configuration 61 Table 6-7 Large general hospital configuration 61 Table 6-8 Costs associated with clinical risk assessment 62 Table 6-9 Costs associated with taking patient swab samples 62 Table 6-10 Proportions assumed for negative sample confirmation 63 Table 6-11 Cost per sample 63 Table 6-12 Costs associated with contact precautions 64 Table 6-13 Additional daily costs associated with single rooms 64 Table 6-14 Estimates of length of patient stay 65 Table 6-15 Decolonisation resource requirement 65 Table 6-16 Parameter values common to both tertiary referral and large general hospital settings 66 Table 6-17 Parameter values specific to tertiary referral hospitals 67 Table 6-18 Parameter values specific to large general hospitals 68 Table 6-19 MRSA screening stategies 69 Table 6-20 Laboratory test combinations 69 Table 6-21 Summary of prevalence and costs of combination of screening strategies and laboratory tests 70 Table 6-22 Cumulative MRSA infections 74 Table 6-23 Summary of results for a tertiary referral hospital using chromogenic agar for all laboratory tests 76 Table 6-24 Summary of results for a large general hospital using chromogenic agar for all laboratory tests 76 Table 6-25 Summary of results of sensitivity analysis 77 Table 6-26 Summary of results using chromogenic agar for all tests and reducing isolation room availability to one per ward 79 v

7 Table 6-27 Summary of results using chromogenic agar for all tests and reducing initial prevalence of MRSA colonisation to 3.6% 79 Table 6-28 Summary of results using chromogenic agar for all tests and reducing efficiency of isolation to 50% 80 Table 6-29 Summary of results if single rooms are occupied prior to introduction of screening strategy 80 Table 6-30 Annual and cumulative increase in patient throughput associated with chromogenic agar Strategy 2 compared with no screening 81 Table 6-31 Cost of screening using chromogenic agar and Strategy 2 81 Table 6-32 Annual number of laboratory specimens 82 Table 6-33 Cost breakdown of laboratory services 82 Table 6-34 Costs associated with decolonisation treatment 82 Table 6-35 Costs associated with isolation of MRSA-colonised patients in single-bed rooms 83 Table 6-36 Cost of contact precautions 83 Table 6-37 Total cost of Strategy 2 using chromogenic agar for a tertiary referral hospital 84 Table 6-38 Nursing resources 85 Table 6-39 Total cost of enrichment chromogenic agar Strategy 2 for a large general hospital 85 Table 6-40 Total cost for tertiary referral hospitals and large general hospitals in Scotland 86 Table 6-41 Cost components of chromogenic agar Strategy 2 for six tertiary referral hospitals and 19 large general hospitals 86 Table 8-1 Key aspects of a screening programme 99 LIST OF FIGURES Figure 6-1 Hospital and community population structure 58 Figure 6-2 Mean prevalence (%) using chromogenic agar for all laboratory tests 71 Figure 6-3 Change in mean prevalence (%) over time 71 Figure 6-4 Mean prevalence (%) for Strategy 2 at Year 1 72 Figure 6-5 Mean costs of screening using Strategy 2 at one year 73 Figure 6-6 Cumulative mean costs using chromogenic agar for all laboratory tests 73 Figure 6-7 Cost breakdown of screening strategies (mean costs) 74 Figure 6-8 Cumulative infections over 5 years using chromogenic agar for all laboratory tests 75 Figure 6-9 Cost components of chromogenic agar Strategy 2 for a tertiary referral hospital 84 Figure 6-10 Cost components of chromogenic agar Strategy 2 for a large general hospital 86 Figure 13-1 Strategy 1: no screening 141 Figure 13-2 Strategy 2: swab all patients 142 Figure 13-3 Strategy 3: swab high-risk specialty patients only 143 Figure 13-4 Strategy 4: risk assess all patients 144 Figure 13-5 Strategy 5: swab high-risk specialty patients, risk assess low-risk specialty patients 145 Figure 13-6 Strategy 6: risk assess all patients and isolate likely carriers 146 Figure 13-7 Screening and chromogenic agar 149 Figure 13-8 Cost components of chromogenic agar Strategy 2 for a large general hospital 163 vi

8 1 EXECUTIVE SUMMARY Background Staphylococcus aureus (S. aureus) is a gram-positive bacterium commonly carried by about three in 10 of the healthy population. If S. aureus enters the body it can cause a range of illnesses, from minor localised skin and wound infections to systemic life-threatening infections such as endocarditis, pneumonia and bacteraemia. The introduction of penicillin in the 1940s to treat S. aureus infections led to the rapid development of resistance to the antibiotic. The introduction of meticillin in 1959 which was stable to the enzyme penicillinase, the cause of earlier resistance, initially appeared to have solved the problem. However, in 1961 meticillin-resistant S. aureus (MRSA) was identified and subsequently epidemic strains have developed. MRSA strains caused major infection outbreaks in various countries during the 1980s and have been endemic in healthcare facilities since the 1990s. In comparison with infections due to meticillin-susceptible S. aureus (MSSA), infections due to MRSA are associated with greater risk of treatment failure, increased patient mortality and higher costs. A major aspect of controlling the spread of MRSA is screening to identify colonised or infected patients, and then managing them to reduce the risk of MRSA transmission to other individuals. Screening involves taking swabs, and then carrying out laboratory analysis of these samples. The approach of screening all patients for MRSA colonisation was advocated by early guidelines. However, as the problem has grown, such an approach has been suggested to be impractical, and increasingly screening is targeted on the basis of a risk assessment of the likelihood of MRSA carriage. Methods This Health Technology Assessment (HTA) considers evidence indicating the possible clinical and cost effectiveness benefits associated with MRSA screening programmes, and assesses the potential impact of the findings in terms of patient management procedures, the patients themselves and the National Health Service in Scotland (NHSScotland). Evidence identified by systematic literature searching and provided by experts and patient interest groups was critically appraised, and expert staff undertook robust analyses. Peer review and wide public consultation were undertaken to ensure that all views were considered. Patient issues were considered by critical review of the research relating to public understanding of MRSA and consideration of staff and public experiences of the management of MRSA in hospitals. In addition, focus groups were commissioned to directly ascertain the opinions of staff and members of the public. A survey questionnaire was used to assess the current level and type of MRSA screening undertaken in Scottish hospitals. Data from the survey questionnaire and other information sources were used to consider the organisational issues associated with delivering an effective MRSA screening service within NHSScotland. Results and conclusions Clinical effectiveness Clinical effectiveness assessment identified a large body of literature on MRSA. However most publications reported on observational studies, infection outbreaks or routine information collection and, as a result, the literature base was considered methodologically weak. The findings which were used to support cost-effectiveness analyses are summarised below. The incidence of MRSA colonisation among patients admitted to acute inpatient care in the UK is not known. Given the restricted data, the results of a recently published study of screening patients admitted to a UK district general hospital and citing an MRSA prevalence of around 7% were selected for the economic model (Rao, 2007). For modelling purposes, screening agar culture of bacteria from swabs was considered to identify MRSA with a sensitivity of 68%, a 6% incidence of falsepositive tests and a turnaround time of 48 hours. Chromogenic culture was considered to identify MRSA with a sensitivity of 98%, a 0.2% incidence of falsepositive tests and a 24-hour turnaround time. Realtime polymerase chain reaction (PCR) was considered to identify MRSA with a sensitivity of 96%, a 5% incidence of false-positive tests and a 24-hour turnaround time. The results of Lucet et al. (2003) were considered most appropriate in terms of assigning an effectiveness measure to clinical risk assessment of carriage. These data indicate that limited swab screening of patients with at least one risk factor for MRSA colonisation would target 65% of patients for screening and identify 88% of MRSA carriers. MRSA transmission is likely to be dependant on a number of factors. For the purposes of this HTA the method of Cooper et al. (2003) was used, where transmission rate is assigned based on the equilibrium colonisation rate, applying admission and discharge rates seen in the absence of infection control measures. Decolonisation of MRSA-positive patients has been recommended by recent guidelines. A 53% successful decolonisation estimate was assumed for the economic model, based on the study of Rohr et al. (2003) where outcome was assessed 1-2 days following 5 to 7-day mupirocin treatment. Examination of the evidence considering screening of staff for MRSA colonisation was requested by patient representatives. Guidelines recommend that staff with skin lesions are referred for screening and that all staff are screened in the event of unexplained, persistent outbreaks. Until the effects of colonisation pressure on staff caring for patients with MRSA and the nature of transient colonisation are better understood, the current recommendations are considered appropriate. Current hospital staffing levels have implications for the feasibility 1

9 of regular screening of staff and sending them home whilst awaiting results, a procedure comparable to isolating a patient potentially colonised with MRSA. Overall, there was a lack of high-quality studies from which values to populate the economic model could be obtained. The lack of robust estimates has an impact on the strength of the recommendations that can be derived from the model. Cost effectiveness The objective of the economic model was to inform recommendations on the cost-effective use of MRSA screening options and to identify the level of isolation facilities required to reduce MRSA prevalence. A stochastic compartmental model based on that of Cooper et al. (2003) was developed and populated with, as far as was possible, Scottish clinical and cost data. The outputs obtained when modelling the tertiary referral hospital setting are detailed below. Of the screening strategies modelled, screening by laboratory test of all patients admitted to hospital with isolation of those identified as potential carriers proved most effective at reducing prevalence rates. Screening all patients by laboratory test without pre-emptive isolation was only marginally less effective. Screening by laboratory test of only those patients admitted to high-risk specialty units was least effective. Using chromogenic agar for all tests was the most clinically effective method of laboratory testing, given the high sensitivity and specificity of this method and the turnaround time. Using chromogenic agar for all tests was also the most cost-effective method of screening. The impact of MRSA on hospital resources was measured as the number of days needed to treat patients whose hospital stay was prolonged as a result of MRSA infection. At a colonisation prevalence of 7.1%, not implementing a screening and isolating policy would result in 1,671 MRSA infections over 5 years when compared with treating the population in the absence of MRSA. Screening all patients using chromogenic agar would, for a tertiary referral hospital, reduce the total number of MRSA infections to around 80 and would allow around 4,000 additional patients to be treated in hospital at a total cost of 2.9 million over 5 years. Strategies using chromogenic agar as the laboratory test resulted in fewest infections and the lowest costs. Sensitivity analyses showed that prevalence rate and fiveyear cost findings were affected by: changing the number of single-bed isolation rooms per ward variation from the assumed prevalence of MRSA of 7.1% reduction in the effectiveness of isolation to reduce transmission of MRSA. These economic data consider the costings appropriate to a tertiary referral hospital, however, similar findings were seen when modelling was based on input values for a large general hospital. Financial implications The annual estimated costs, using the parameters in the model presented in this report, for a nominal 840-bed tertiary referral hospital range are as follows: 158,800 in Year 1 rising to 160,600 in Year 5 for screening 289,500 in Year 1 increasing to 291,500 in Year 5 for laboratory test analysis 209,800 in Year 1 decreasing to 32,000 in Year 5 for the increased cost associated with isolation in single rooms 41,200 in Year 1 decreasing to 9,300 in Year 5 for administering decolonisation treatment 60,300 in Year 1 reducing to 9,200 in Year 5 for administering contact precautions. The total costs to NHSScotland of screening all patients for MRSA on admission to hospital using a chromogenic agar test are estimated to be approximately 14.3 million in the first year reducing to 9.7 million in the fifth year, giving a total of 55 million over a 5-year period. Patient issues The research evidence base concerning patient issues was limited and of poor quality. However, a number of points raised in the literature were reiterated by the focus groups. There was evidence that patients do not always understand the nature and implications of MRSA colonisation or infection, or the requirement for contact precautions. Both the literature and focus groups highlighted the fear that individuals have of MRSA infection, which is associated with the stigma of infected individuals being perceived as unclean. The focus group emphasised the absence of clear information about MRSA and their reliance on sources such as the media and the internet. Staff were generally positive about the concept of preadmission MRSA screening but did have concerns over the logistics of providing the service. The literature showed that there are varying degrees of compliance with isolation practices by professional healthcare staff. This in itself can exacerbate problems, with non-compliance sending mixed messages to patients. Some research has indicated that being nursed in a single-bed room can be a a negative experience. However, other literature and focus groups considered single-bed rooms to have a number of advantages. 2

10 Current MRSA screening practices A survey was conducted to detail current MRSA screening practices in hospitals across Scotland. Almost 60% of hospital infection control units reported that they carry out some assessment of MRSA colonisation risk at the time of patient admission and 41% assess patients susceptibility to MRSA infection. Risk assessment is carried out prior to elective admission in a number of services. Three services reported MRSA screening of all patients undergoing elective procedures at the time of admission. Selective preadmission screening did take place in some services including orthopaedic and cardiac surgery units. A variety of laboratory methods are currently in use to identify MRSA and carry out susceptibility testing, with the most common method in Scotland being agar screening (67%). PCR analysis of MRSA screening swabs is not in use in Scotland. Some units reported a wait of up to 5 days for test results. The preferred management strategy for MRSA colonised or infected patients is to isolate or cohort them whenever possible (63%). However, isolation or cohorting is not implemented if other patients on the ward are not physically ill, as in a psychiatric unit. Most services indicated that colonised patients would undergo decolonisation. Organisational issues The organisational issues associated with an MRSA screening policy are linked to the wider issues of infection control in hospitals. Ethical considerations, such as the right of the individual to make an informed choice and the balance of benefit over harm, are associated with MRSA screening and subsequent patient management. A balance must be struck between fully informing patients of these issues and overloading them with unnecessary information. The benefits of screening are primarily seen in terms of vulnerable others within the hospital experiencing reduced exposure to MRSA. Patient concordance with a screening programme could be maximised through open and informed discussion regarding the reasons for screening and by emphasising that the care provided to those requiring isolation is acceptable and appropriate. Expansion of MRSA screening will require coordination and communication between preadmission and admission units and laboratory services. This will increase demands, in terms of time and knowledge of MRSA, on nurses carrying out the screening. However, these demands should be considered relative to the resources required to fire-fight outbreaks and to manage patients with serious MRSA infections. The quality of screening should be maximised through training and protocol development. Laboratory facilities will require sufficient capacity to process additional samples, and the adoption of new techniques will require additional resources and staff training. There is currently no standard process for MRSA identification and susceptibility testing across Scotland. The current provision of isolation facilities is inadequate in the short to medium term, to meet the demand for single-bed isolation rooms for patients colonised with MRSA. An increase in the identification of those colonised with MRSA would result in a greater demand for alternatives to open-ward accommodation. Scottish hospitals will require a higher proportion of single-bed rooms to increase capacity for isolation. Discussion This HTA has raised a number of issues that require consideration in the light of implementing an optimal MRSA screening and patient management policy. The assessment applies an MRSA screening model to the hospital setting. The simplified nature of a model means that many of the interactions observed at ward level are not taken into account. The only driver of MRSA prevalence explored was colonisation pressure, which was assumed to be fixed. All other factors associated with transmission of infection, known and unknown, were assumed to be comparable across the hospital setting and to be independent of colonisation pressure. There is a lack of robust data to populate the economic model. The prevalence of MRSA colonisation in Scotland is unknown and literature on transmission rates, the effectiveness of decolonisation, assessment of MRSA colonisation risk and swab screening is limited. To provide more robust models of the economic impact of infection control strategies, there is a requirement for data from high-quality research studies. The resource costs required for PCR screening of swabs are high, meaning that the technology is unlikely to represent a realistic replacement for culture methods. In routine clinical practice, where samples require to be batched before testing, PCR does not have a turnaround time advantage over chromogenic agar testing. Given sufficient isolation facilities, swab screening of just those patients admitted to high-risk specialty units would reduce MRSA prevalence, but to a lesser extent than the other strategies modelled. It has been assumed for the purposes of the modelling described here, that one strategy would be adopted and maintained for a five-year period. Given that some strategies predict a rapid reduction in MRSA prevalence in the first year following implementation, a change to a less resource intensive approach might be warranted but the implications have not been explored for this report. Being identified as MRSA positive and being nursed in isolation has implications for this group of the patient 3

11 population. Policies will need to be developed to ensure that patients in isolation do not experience an unnecessarily reduced quality of care compared with those on the open ward. This model is not predictive, and although it can estimate the resources required to implement a range of screening strategies, the impact of the strategies on MRSA prevalence cannot be categorically determined. However, the model does provide an indication of the direction and magnitude of the effect of screening on MRSA prevalence. The results of this modelling exercise would indicate that the resource requirement associated with implementing a programme of screening for MRSA are substantial, however, the opportunity costs of not screening are even greater. These requirements will have to be considered in the light of competing pressures for NHS resources. Recommendation 5 Care of patients isolated as a result of MRSA colonisation or infection should not result in their being or feeling unnecessarily disadvantaged. Much distress will be avoided with high-quality patient information and effective communication between healthcare staff and patients about their condition and its management. Recommendations Recommendation 1 A primary study should be set up in acute inpatient care within a whole NHS Board area (which should include a tertiary referral hospital and one or more large general hospitals) to assess whether screening all patients for MRSA is effective in preventing MRSA infection as predicted by the economic model. Data from this study should be collected for at least one year to decide whether MRSA screening results in a reduction in prevalence of MRSA. The Scottish Government* should fund and manage this study. Recommendation 2 There is currently insufficient evidence on staff MRSA transmission to determine an appropriate schedule of screening and subsequent management. Therefore, current guidelines indicating screening on occasion of unexplained outbreaks should be followed. Further research of the extent and implications of staff colonisation is urgently required. Recommendation 3 Systems should be developed to collect patient-based data on the prevalence of MRSA colonisation and infection to determine the effectiveness of infection control strategies. The resource implications of establishing such systems are recognised, however, these are necessary to plan and evaluate future strategies to control MRSA. Recommendation 4 High-quality patient information on MRSA, the purposes of screening and methods to achieve infection control should be distributed to all patients and relatives on admission to hospital. * On 3 September 2007, Scottish Ministers formally adopted the title Scottish Government to replace the term Scottish Executive as an expression of corporate identity. 4

12 2 INTRODUCTION This report presents the recommendations arising from the NHS Quality Improvement Scotland (NHS QIS) HTA on the clinical and cost effectiveness of screening for MRSA in patients admitted to acute hospital care. Section 3 of the document provides information on NHS QIS and on the HTA process. A description of the nature of MRSA, the methods for screening, policy initiatives developed in Scotland to address MRSA and other healthcare associated infections (HAI), and the rationale for undertaking this work are presented in Section 4. The studies used to define clinical estimates for the economic model are summarised in Section 5 and the economic analyses are detailed in Section 6. Patient issues are considered in Section 7 and organisational issues associated with MRSA screening are explored in Section 8. Section 9 discusses the HTA findings including its limitations and uncertainties, and Section 10 presents the recommendations. NHS QIS is committed to equality and diversity. This document, and the research on which it is based, have been assessed for any likely impact on the six equality groups defined by age, gender, race/ethnicity, religion/faith, disability and sexual orientation. For a summary of the equality and diversity impact assessment, please see 5

13 3 BACKGROUND ON NHS QUALITY IMPROVEMENT SCOTLAND NHS QIS was set up by the Scottish Parliament in 2003 to take the lead in improving the quality of care and treatment delivered by NHSScotland. NHS QIS sets standards, monitors performance and provides NHSScotland with advice, guidance and support on effective clinical practice and service improvements. Health Technology Assessment HTA is an internationally recognised process used by NHS QIS to advise NHSScotland about a specific health intervention, eg medicine, equipment or diagnostic test. HTA evaluates the clinical and cost effectiveness of the various ways in which a particular intervention can be used, comparing alternatives where appropriate. Patient and organisational issues are also considered. Evidence is identified by systematic literature searching and assimilating expert evidence, the views of patient interest groups and manufacturers. The evidence is then critically appraised and robust analyses are undertaken by expert staff. Surveys may also be undertaken to ascertain current clinical practice and patient preferences. NHS QIS staff from a variety of disciplines conduct the assessment, with considerable input from healthcare professionals expert in the particular area of interest (see Appendix 1). Peer review and wide public consultation ensures that all views are considered. 6

14 4 SETTING THE SCENE 4.1 Description of MRSA in Scotland Each year around one in 10 NHS patients contracts an HAI in the UK (Department of Health, 2003). A recent study of the prevalence of HAI in Scottish hospitals found that 9.5% of inpatients in acute hospitals had an HAI and of these, around 17% were MRSA infections (Health Protection Scotland, 2007a). In Scotland, HAIs are a major factor in an estimated 457 deaths each year and a contributory factor in a further 1,372 deaths (3% of all deaths). The economic burden of HAI to the hospital sector is substantial, with the annual cost to the NHS estimated to be 186 million in Scotland (Walker, 2001) and in excess of 1 billion in England and Wales (Plowman et al., 1999). MRSA is a major cause of HAI (Boyce et al., 2005). S. aureus is a gram-positive bacterium, commonly carried by about three in 10 of the healthy population without causing any infection. It is most often found in the nose or on the skin (Wertheim et al., 2005). S. aureus can be introduced into the body through broken skin, insertion of a device or an invasive procedure. Once within the body, it can cause a range of illnesses from relatively minor localised skin and wound infections such as boils, to systemic life threatening infections such as endocarditis, pneumonia and bacteraemia. Following its introduction for therapeutic use in the 1940s, penicillin was used to treat Staphylococcal infections. By the end of that decade, however, approximately 60% of all strains of S. aureus found in hospitals had become resistant to penicillin and resistance rapidly developed to other antibiotics used (Boyce et al., 2005). The introduction of the antibiotic meticillin in 1959 which was stable to the enzyme penicillinase, the cause of earlier resistance, initially appeared to have solved the problem. In 1961 however, meticillin-resistant S. aureus was identified. Subsequently, epidemic strains of MRSA have developed and caused major infection outbreaks in hospitals in various countries during the 1980s. These strains have been endemic in healthcare facilities since the 1990s (Enright et al., 2002). The meca gene, which is found on a mobile genetic element called the staphylococcal cassette chromosome (SCCmec) is responsible for conferring meticillin resistance. This gene changes proteins in the cell wall and allows the bacterium to continue to proliferate in the presence of meticillin and other beta-lactam antibiotics. The introduction of the mobile SCCmec element into different MSSA clones resulted in the development of a number of MRSA strains, some of which had characteristics that would enable them to be transmitted widely (Boyce, 2002). Eleven major MRSA clones have been identified to date (Enright et al., 2002) and currently most MRSA infections in UK healthcare facilities are due to two different clones, Epidemic MRSA (EMRSA) 15 and EMRSA 16 (Boyce et al., 2005). Recent years have seen the emergence of new virulent strains of MRSA that have spread within the community. These strains are genetically distinct from healthcare associated MRSA isolates and their true prevalence is unclear (Wertheim et al., 2005). The prevalence of MRSA has increased considerably since the 1990s (Reacher et al., 2000) with rates varying greatly between countries. Giannoudis et al. (2005) noted that the level of meticillin resistance in S. aureus isolates is reported to be less than 3% in Scandinavia, Switzerland and the Netherlands and greater than 30% in Spain, France, Italy and Portugal. The proportion of UK S. aureus isolates that are meticillin resistant are among the highest in Europe at 43% ( MRSA is not confined to Europe, with the USA reporting 64% of S. aureus isolates to be meticillin resistant and Japan citing 74% (Dr D Morrison, Principal Clinical Scientist, Scottish MRSA Reference Laboratory. Personal communication). The low prevalence in certain countries is generally attributed to restricted antibiotic usage and more importantly, to strict infection control policies (Boyce, 2002). In Scotland, national data on the incidence of MRSA bacteraemia are collected through two surveillance systems: the Health Protection Scotland (HPS) laboratory reporting system and from referrals of isolates to the Scottish MRSA Reference Laboratory, as part of the European Antimicrobial Resistance Surveillance System (EARSS). Data from both systems are now combined and are reported quarterly as part of the Scottish Surveillance of Healthcare Associated Infection Programme (SSHAIP). The latest SSHAIP report (Health Protection Scotland, 2007b) concluded that quarterly MRSA bacteraemia rates for Scotland have been relatively stable between 2003 and 2007, at between 0.15 and 0.2 cases of MRSA per 1,000 acute occupied bed days. The data from EARSS showed that in 2006, 42% of S. aureus bacteraemias were due to MRSA. Infections due to MRSA are associated with a higher risk of treatment failure, increased mortality and increased costs than those attributed to MSSA (Cosgrove et al., 2003; Melzer et al., 2003). This higher mortality is thought to be due to delays in starting treatment with appropriate therapies, the inferior quality of some of the antistaphylococcal agents used to treat MRSA infection (Lodise et al., 2003) and the severity of the underlying condition in the affected patient (Cooper et al., 2003) rather than due to increased virulence of MRSA strains. MRSA infections are usually treated with the glycopeptides vancomycin and teicoplanin, and new products such as linezolid and synercid. However, MRSA strains resistant to these agents have been documented (Boyce et al., 2005; Cooper et al., 2003). As the range of therapeutic options becomes more limited, there is an urgent need to control the spread of colonisation and infection by MRSA. Within healthcare facilities, there are three major reservoirs of MRSA: patients, healthcare workers and the environment. The importance of the latter has not been 7

15 established. MRSA is most commonly spread between patients on the hands of healthcare workers, who have become transiently contaminated (Rubinovitch & Pittet, 2001). Healthcare workers with nasal colonisation can also spread the bacteria through droplet transmission. Airborne transmission can occur when the dressings and bedding of MRSA-infected or colonised patients with burns, wounds or dermatitis are changed (Boyce, 2005). Up to 1% of patients admitted to acute care hospitals endemic for MRSA may become colonised and of those individuals, 20-60% will go on to develop an infection (Rubinovitch & Pittet, 2001). The major route for becoming colonised or infected by MRSA is direct contact (Cooper et al., 2003; Muto et al., 2003). Therefore hospital infection control policies focus on preventing contact-mediated transmission, but are normally part of wider infection management strategies involving a multiplicity of measures. Guidelines prepared by the Society for Healthcare Epidemiology of America (SHEA) (LeDell et al., 2003) on preventing nosocomial transmission of multidrug-resistant strains of S. aureus and Enterococcus recommend the following procedures: active use of surveillance cultures to identify the reservoir for transmission cohorting colonised or infected-mrsa patients together (this can range from the use of side-room isolation, to cohorting affected patients in rooms or bays with or without the simultaneous cohorting of staff, to the use of isolation wards) improve hand hygiene use of contact precautions for patients known or suspected to be MRSA colonised (including the use of masks, gowns and gloves) judicious antibiotic usage decolonisation of colonised patients (attempting to clear patients of MRSA carriage using interventions such as mupirocin) educational programmes adequate disinfection methods use of computer systems to track infected or colonised patients where possible to dedicate the use of non-critical equipment to a single patient or cohort. There is considerable uncertainty over the effectiveness of individual control measures, as it is difficult to determine the effects of one measure in comparison with another when they do not act independently (Cooper et al., 2003). Some reports have questioned whether the merits of isolating patients in preventing MRSA spread are outweighed by the disruption and emotional upset caused. In a recent high-quality systematic review however, Cooper et al. (2003) found that there was evidence that intensive concerted interventions that include isolation can substantially reduce MRSA prevalence. Isolation policy is liable to failure, however, if inadequately resourced and managed. The SHEA guidelines (Muto et al., 2003) note that multiple studies implementing surveillance cultures in conjunction with contact precautions have resulted in a significant reduction in the rates of both MRSA colonisation and infection. 4.2 Recent policy initiatives In response to the development of new MRSA strains and the limited availability of novel antimicrobials, the Department of Health s Specialist Advisory Committee on Antimicrobial Resistance (SACAR) asked the British Society for Antimicrobial Chemotherapy (BSAC), the Hospital Infection Society (HIS) and the Infection Control Nurses Association (ICNA) to update previously published guidelines (Duckworth et al., 1998) on the control of MRSA in UK hospitals (Coia et al., 2006). The group was asked to extend their remit to cover laboratory diagnosis and susceptibility testing of MRSA, and the prophylaxis and treatment of MRSA infections. Guidelines on the former were published in 2005 (Brown et al.) and the group has recently finished consulting on draft guidelines relating to prophylaxis and treatment (Gemmell et al., 2006). Based upon the 1998 guidelines, the Royal College of Nursing (RCN) also prepared specific guidance for nurses, a new edition of which was also published in 2005 (Royal College of Nursing, 2005). Within Scotland, the Scottish Infection Standards and Strategy (SISS) Group, a bicollegiate grouping bringing together a number of societies and organisations representing clinical and laboratory professions, have prepared guidance on good practice in MRSA management (Scottish Infection Standards and Strategy Group, 2006). While there are currently no specific national guidelines relating to MRSA in Scotland, extensive work has been undertaken covering the wider area of HAI. This includes the establishment of an HAI Task Force by the Scottish Executive in Key policy documents published to date include: Preventing Infections Acquired While Receiving Healthcare: the Scottish Executive s Action Plan to Reduce the Risk to Patients, Staff and Visitors (Scottish Executive Health Department, 2002b) The Antimicrobial Resistance Strategy and Scottish Action Plan (Scottish Executive Health Department, 2002a) The Risk Management of HAI: A Proposed Methodology for NHSScotland (Scottish Executive Health Department, 2004b) The NHSScotland Code of Practice for the Local Management of Hygiene and HAI (Scottish Executive Health Department, 2004a). NHS QIS is committed to providing guidance and support in this area. NHS QIS has published national standards for HAI control (Clinical Standards Board for Scotland, 2001) and a review of progress based on these standards (NHS Quality Improvement Scotland, 2004). The standards have recently been revised and are currently in draft for consultation (NHS Quality Improvement Scotland, 2007). In addition, NHS QIS has launched a poster campaign within the NHS to alert visitors to the spread of HAI, and 8

16 published an assessment of the effectiveness of alcoholbased hand hygiene products (Ritchie et al., 2005) 4.3 Description of technology A major aspect of controlling the spread of MRSA is the identification of colonised or infected patients by screening, and then managing them to reduce the risk of transmission (Boyce, 2002; LeDell et al., 2003). Screening involves taking swabs from one or more body site and carrying out laboratory tests on these samples. Swabs are taken from the site of primary infection if present or sites where carriage is most likely, such as the nose, throat, groin, perineum, wounds or broken skin if present, urine if catheterised or sputum if productive cough is present (Coia et al., 2006). The conventional method for detecting MRSA is culture followed by antibiotic susceptibility testing, which usually take several days. The most widely used methods for MRSA detection in the UK involves culture on oxacillin/salt screening plates, followed by oxacillin disc diffusion susceptibility testing. Other options include broth and agar microdilution tests, disc diffusion, automated systems, penicillin binding protein (PBP2a) latex agglutination and PCR. The PBP2a latex agglutination methods can provide results within 15 minutes (Farr, 2004) and PCR can be completed within 2 hours of taking the swab (Brown et al., 2005). Such methods, therefore, offer the potential for more rapid introduction of control measures or therapy. Screening can be carried out preadmission for non-emergency admissions, on admission, regularly for long-stay patients, following hospital transfer, after a case of MRSA is detected and on discharge from the hospital. The approach of screening all patients for MRSA was advocated by the early guidelines, from the late 1980s and early 1990s. As time has gone on and the extent of the MRSA problem has grown, such an approach has been considered impractical and there has been a move towards targeted screening based on risk assessment (Cooper et al., 2003). This risk assessment approach involves two separate assessments. Patients are assessed on their likelihood of being colonised or infected with MRSA, with there being a greater requirement to screen high-risk patients. Patient groups considered most likely to be colonised or infected with MRSA (Coia et al., 2006) are: those previously infected or colonised with MRSA patients with frequent admissions to healthcare facilities patients transferred from another hospital those who have been recent inpatients of hospitals in the UK or abroad with known or likely high MRSA prevalence residents of care facilities with a known or likely high MRSA prevalence patients that fall into other high-risk groups such as injecting drug users, HIV-positive individuals and professional contact-sport players. Patients are also assessed for the risk that MRSA is likely to pose to them, ie how susceptible or vulnerable they are to MRSA infection. Patients considered to be most vulnerable to MRSA (Coia et al., 2006) include those admitted to the following units: intensive care burns transplantation cardiothoracic orthopaedic trauma vascular surgery renal. There is greater need to screen patients considered vulnerable, both for their own protection and for the safety of other patients in the units to which they are admitted. Prior to the results of screening being obtained, risk assessment can be used to guide interim patient management. When screening results become available, these, in combination with the results of the risk assessment, inform the management strategy to be adopted for each patient. The approach can be influenced by ward design, the availability of single-bed rooms, the facilities for patient isolation, whether the patient is likely to be a heavy MRSA shedder, and the antibiotic resistance patterns, virulence and potential transmissibility of the identified organism (Coia et al., 2006). It is believed that most staff colonisation with MRSA is transient and localised to the hands. Hand hygiene practice of healthcare workers is therefore considered of primary importance in MRSA infection control, although this is unlikely to be successful in the absence of other measures (Muto et al., 2003). Occasionally, staff, particularly those suffering from dermatitis, remain colonised for longer periods and can present a significant and ongoing clinical threat. If a persistent problem is identified, it could be appropriate to screen staff (Boyce, 2005; Coia et al., 2006; Scottish Infection Standards and Strategy Group, 2006). 4.4 MRSA screening services in Scotland A survey to establish current implementation of MRSA screening policies and practice across Scotland was conducted during August and September The survey considered the following areas: assessment of the risk of MRSA carriage patient screening programmes laboratory screening methods the management of infected patients the epidemiology of infection staff screening programmes the economic implications of the presence of MRSA. The results of the survey are presented in Section 8.2 of this report. 9

17 4.5 Objective and focus of HTA Recent UK guidelines provide contradictory recommendations on the MRSA screening process and those groups of patients to be targeted for screening. As a result, this HTA was commissioned to consider the different MRSA screening strategies and appropriate management of patients found to be colonised. This work only considered the screening of patients admitted to acute hospital inpatient care, although it is recognised that MRSA is a problem within long stay and other community healthcare settings. 10

18 5 CLINICAL EFFECTIVENESS 5.1 Introduction This section appraises the literature considering various aspects of MRSA colonisation and infection, with the principal intent of populating the economic model described in Section 6. Studies used included those determining the prevalence of MRSA colonisation, the effectiveness of clinical risk assessments and laboratory screening tests, the transmission of MRSA in the clinical setting, the effectiveness of decolonisation procedures and the effectiveness of isolation policies. In addition, the issue of screening staff for MRSA colonisation was identified by a patient issues subgroup and merited examination of the clinical effectiveness literature. Accordingly, separate literature searches were conducted to identify and review staff screening studies. 5.2 Methods Literature search To assess the size and quality of the relevant primary literature, initial MEDLINE scoping searches were undertaken in April These searches were designed to retrieve studies relating to screening patients and staff for MRSA, laboratory tests for MRSA and their diagnostic accuracy, and studies relating to managing MRSA including those outlining patient pathways and protocols. Searches to identify high-level evidence such as HTAs, systematic reviews and ongoing relevant research were undertaken in January Sources included the HTA database, the Cochrane Database of Systematic Reviews and the Database of Abstracts of Reviews of Effects (DARE). To identify the primary literature, the following bibliographic databases were searched using the OVID platform: MEDLINE EMBASE CINAHL MEDLINE in PROCESS. Searches were undertaken in July 2005, September 2005, December 2005 and June 2006, as shown in Table 5-1. Table 5-1 Summary of literature searches Search Date Limits/Filters 1. Prevalence of MRSA July Risk management July 2005 None 3. Tests for MRSA July 2005 Diagnostic accuracy methodological filter 4. Management of MRSA July 2005 None 5. Decolonisation July Staff screening Sept 2005 None 7. Association of colonisation to infection 8. Decolonisation treatmentsnamed in Scottish Infection Standards and Strategy Group (2006) 9. Decolonisation - update of Search 5 June June June 2006 July 2005-June 2006 Randomised controlled trials (RCTs) methodological filter 11

19 All searches combined the concepts Staphylococcal infection with meticillin resistance and aimed to be as comprehensive as possible, identifying all relevant subject terms and keywords. The MRSA prevalence search was restricted by year to retrieve only studies indexed between 2000 and 2005, the association of colonisation to infection to retrieve those indexed , and the decolonisation treatments search was restricted to the years In-house methodological filters, as presented in Appendix 2, were used to identify studies on MRSA tests and decolonisation allowing both search sets to be filtered. For MRSA tests, this involved retrieving studies focusing on the sensitivity and specificity of each test or tests. For decolonisation, this involved retrieving only RCTs. The decolonisation search was updated during the consultation period, and a related search specifically looking at the decolonisation treatments named in Scottish Infection Standards and Strategy Group (2006) was undertaken to identify any additional studies. A general search for RCTs relevant to MRSA was undertaken in Cochrane Central Register of Controlled Trials (CENTRAL), Current Controlled Trials (mrct), Clinical Trials.gov, Trials Central, CORDIS and CRISP. 5.3 Results Prevalence rates Study selection The literature search for papers with data on the prevalence of MRSA in patients admitted to acute hospital care identified 1,941 titles, of which 77 abstracts were selected. These included primary studies investigating the prevalence of MRSA colonisation amongst hospital patients published since Two reviewers assessed the full papers and 21 reports were selected as giving details of prevalence rates derived from UK and Irish studies. For the purposes of the economic model, data from studies considering the prevalence of patients admitted to acute inpatient care were required. Studies of point or period prevalence and those conducted in nursing homes, hospices or comprising specific patient groups (eg children with cystic fibrosis) were excluded. Four papers met these criteria and are summarised in Table 5-2. A summary of the 17 excluded studies is presented in Appendix 3, Table Extended searching was undertaken in December 2005 for studies relevant to screening, in particular considering staff screening. The following additional sources were searched: Web of Science Health Management Information Consortium British Nursing Index PsycINFO Applied Social Sciences Index and Abstracts HAI Portal - NHSScotland elibrary. Literature was also identified using Zetoc (the British Library s table of contents alerts service) by citation searching on key papers and from scanning the bibliographies of retrieved items. Language restrictions were not imposed on any search. A list of the sources searched and a copy of the strategy used to search the MEDLINE database is presented in Appendix 2. This strategy was adapted to search all other databases. A complete listing of all search strategies can be obtained by contacting NHS QIS. 12

20 Table 5-2 Prevalence of MRSA colonisation on hospital admission Study Aim Study design Population Screening and infection control Porter et al. (2003), Manchester Prevalence of nasal carriage on admission to a medical surgical intensive care unit (ICU) Retrospective review of all patient (pt) records from Apr 1998 to Mar 2000 (24 months) 565 pts with records of nasal swabs Nose swabs; enrichment broth. Samad et al. (2002), Wales Theaker et al. (2001), England Prevalence of colonisation in surgical pts on admission; risk factors Rates of colonisation and infection in a general ICU Prospective; questionnaire; Nov 2000 to Jan 2001 Prospective observational survey of pts admitted over 18 months; dates not reported; Pts in ICU <48 h and -ve (n=334) were excluded. Dates of study not reported. 430 pts, including 261 emergency admissions; mean age 60 years (range ) 308 pts +ve within 48 h of admission or in ICU >48 h Nose and pre-existing wound site swabs on admission; no enrichment broth. Barrier nursing, treatment. Culture unclear. Infected pts treated with vancomycin. Thompson (2004), England Prevalence on admission and acquisition rate in a general ICU Screening of all admissions between Oct 2001 and Mar 2004 (30 months) 1,361 admissions of 1,252 pts Nose, groin swabs on admission then weekly, other cultures as clinically indicated; enrichment broth. Isolation, cohort nursing. Results Notes MRSA isolated from 16/565 (2.8%) pts. Too few MRSA cases for further statistical analysis. 23/430 +ve initial swab (5.3%); 15 were nasal carriers. Reports only P values for analysis of risk factors. 19/642 +ve at admission (3.0%). 308 pts admitted +ve or stayed >48 h, 78 pts acquired MRSA as inpatient 54/97 colonised, 43/97 infected pt days, 713 pt days of colonisation, 294 pt days of infection. Median time to colonisation 7 days (IQR 4 12), to infection: 10 days (IQR 6 16). Reports statistical analysis of length of stay ve admissions (8.7%; 95% CI 7.2 to 10.2) of 111 pts; 21/119 known before admission; 68/1,242 pts -ve on admission acquired MRSA. Reports +ve admissions by source and acquisition rate per length of stay. 13

21 Assessment of the evidence No studies reported colonisation rates on admission in the acute hospital population. Three studies (Porter et al., 2003; Theaker et al., 2001; Thompson, 2004) were conducted in ICU patients, with the fourth being conducted in surgical patients (Samand et al., 2002). The prevalence of MRSA colonisation on admission ranged between 2.8% and 8.7%. A more recent study of approximately 10,000 patients admitted at a UK Accident and Emergency (A&E) department over a one-year period found an overall prevalence rate of 6.7% (Rao et al., 2007) Model estimates for MRSA prevalence It is not possible to determine with any certainty the prevalence of MRSA colonisation in patients admitted to Scottish hospitals as, in the absence of a comprehensive screening programme, this data is unavailable. It is generally believed that colonisation rates are likely to be higher in patients admitted to high-risk units, as these patients are more likely to have a history of hospitalisation and repeated antibiotic exposure. Therefore MRSA prevalence rates collected from ICU patient populations may not reflect the general hospital patient population. As such, the recent data collected from screening hospital admissions referred at a UK A&E unit were considered most applicable to the current Scottish patient MRSA prevalence rate and were used in the economic model MRSA detection using clinical risk assessment Study selection Literature search identified 283 titles and abstracts, from which 33 full papers reporting on the effectiveness of clinical risk assessment at predicting MRSA colonisation were obtained. Studies with a focus on clinical risk assessment on hospital admission, based on identifiable clinical risk factors, were included. Studies published in Japanese and Chinese were excluded. This process identified six papers that assessed the effectiveness of clinical risk assessment to predict MRSApositive status on admission, as summarised in Table Assessment of the evidence Literature analysis showed that the risk factors considered during clinical risk assessment were highly variable across studies. The numerous factors considered included: the type of admission, direct admission or transfer from another healthcare facility, history of hospitalisation over 1-5 years prior to admission, the presence of skin lesions and antimicrobial therapy during the last year. No standard clinical risk assessment tool has been established and as different studies used different sets of risk factors, this prevented data from different studies being combined. In addition, no study assessed the risk factors recommended by the BSAC/HIS/ICNA guidelines and considered pertinent to the clinical risk assessment criteria used for this HTA (Coia et al., 2006). 14

22 Table 5-3 Characteristics of clinical risk assessments for MRSA colonisation on admission Study Population Factors associated with MRSA-positive screen (as reported in publication) Hidron et al. (2005), USA Prospective study; 1,000 bed public inner city hospital;all admissions on Sun, Tue or Thur; 726 pts screened. Hospitalisation in past 12 months: OR 4.0 (95% CI 1.97 to 8.15), P= Skin or soft tissue infection at admission: OR 3.4 (95% CI 1.46 to 7.90), P= HIV +ve and no antimicrobial use in past 3 months: OR 13.8 (95% CI 4.34 to 43.04), P< HIV-ve and antimicrobial use in past 3 months: OR 2.5 (95% CI 1.20 to 5.03), P<0.01(multivariate analysis). Lucet et al. (2003), France Prospective study; 14 ICUs Jul 1997 to Dec 1997; 2,347 pts screened within 24 h of admission. Transferred patients: >60 yrs: OR 2.1 (95% CI 1.07 to 3.90), P=0.02 hospital stay 21 days before ICU admission: OR 3.1 (1.75 to 5.49), P= Directly admitted patients: >60 yrs: OR 3.3 (95% CI 1.69 to 6.43), P< History of ICU stay in last 5 years: OR 3.5 (95% CI 1.75 to 6.94). History of surgery in last 5 years: OR 2.6 (95%CI 1.38 to 5.02), P= Presence of open skin lesions: OR 2.1 (95% CI 1.02 to 4.14), P=0.04. Manian et al. (2002), USA Prospective study; Teaching hospital, all newly admitted pts to acute rehabilitation unit from Jul 1999 to May 2000; 411 of 427 consecutive pts screened for MRSA. History of colonisation: OR 32.0 (95% CI 11 to 95), P< Referral source: OR 1.2 (95% CI 1.03 to 1.40), P= Notes Prevalence 53 of 726 Prevalence 162 of 2,347 (6.9%; 95% CI 5.9% to 8.0%). Cost-benefit analysis of different screening strategies on basis of risk assessment presented. Prevalence 51 of

23 Study Population O Sullivan & Keane (2000a), Republic of Ireland Prospective survey; residents of six longterm residential care facilities for the elderly; 910 screened for MRSA, 786 questionnaires completed. Troillet et al. (1998), USA Prospective study; teaching hospital, admissions to two wards (vascular surgery and podiatry, and general surgery and general medicine) Mon - Thur (no dates); 387 of 746 eligible admissions screened for MRSA. Vovko et al. (2005), Slovenia Prospective survey; residents of long-term care facility; 102 participants screened. Factors associated with MRSA-positive screen (as reported in publication) Male: OR 2.2, P=0.004 (multivariate analysis). Known MRSA carrier: OR 47.8, P= Open persistent wound: OR 8.1, P= Diabetes: OR >6.0, P= Nursing home stay within a year: OR 10.2, P<0.001 (multivariate analysis). Treatment with antibiotics: OR 5.1 (95% CI 1.02 to 25.48), P= No of hospital admissions in 3 months before study 2: OR 6.3 (95% CI 1.31 to 30.05), P=0.22(multivariate analysis). Notes Prevalence 121 of 910 Prevalence 10 of 387 OR computed from cases vs controls with no S. aureus ie those with MSSA excluded. Prevalence 12 of

24 Model estimates for MRSA detection using clinical risk assessment Lucet et al. (2003) reported sensitivity and specificity data for detecting MRSA carriage on admission to an ICU, based on the frequency of a wide range of potential risk factors, rather than on the presence of a limited number. In the absence of studies to support estimates for an established standard clinical risk assessment protocol, the results from Lucet et al. (2003) were used to derive estimates for the economic model. The frequency approach adopted was considered the most encompassing of the wide ranging risk factors reported in other studies and offered the possibility of adapting these estimates to various settings. The data from transferred and directly admitted patients were pooled, as the source of admission would be considered as a risk factor in NHSScotland. The estimates used in the economic model are presented in Table 5-4. These results indicate that limiting MRSA screening on admission for patients with at least one risk factor would target 65.2% of patients and identify 87.5% of MRSA carriers Laboratory tests Study selection The literature search identified 1,822 titles and abstracts, published from 1990 onwards, from which 145 primary studies reporting the diagnostic accuracy of tests to screen for MRSA were selected. Studies focusing on postoperative patients, molecular typing, antibiotic sensitivity and meticillin-resistant coagulase-negative Staphylococcus were excluded. Studies published in Chinese, Japanese, Serbian and Turkish were also excluded. Full reports of the 145 selected studies were assessed by two reviewers independently and disagreements were resolved by discussion. Although there is a substantial literature on laboratory testing for MRSA, relatively few studies were able to provide the information required for this project. Studies of efficacy of tests on pure cultures rather than clinical specimens could not be included and as has been highlighted in recent guidelines (Brown et al., 2005), the performance of any method is dependent on the comparator tests and in the absence of a gold standard reference test, this results in variation in results between studies. The use of robust inclusion and exclusion criteria is critical in ensuring that bias is not introduced when reviewing such studies. Studies were retained if patients were screened on admission, swab specimens were tested (rather than bacterial isolates), the reference standard was defined and data were reported to construct a 2x2 sensitivity and specificity table. Studies of mixed populations were included if data from patients screened on admission were reported separately. Twenty studies met these criteria. Study selection criteria required that all swab specimens were subject to both the index and reference standard tests, and that the index test results were not used as part of the reference standard. On this basis, three culture method studies (Apfalter et al., 2002; Nsira et al., 2006; Stoakes et al., 2006) and three real-time PCR studies (Desjardins et al., 2006; Huletsky et al., 2005; Warren et al., 2004) met the inclusion criteria. Estimates of sensitivity and specificity from these studies were used in the economic model (Section 6.3). No studies were identified that reported the diagnostic accuracy of MRSA detection directly from swab specimens using Vitek systems (biomérieux) compared with a standard reference method. An updated search was conducted post consultation to source the most recently published literature. Following review of this updated search, a further 10 full reports were assessed against inclusion and exclusion criteria as previously described. Five studies met the initial criteria, however one study (Nilsson et al., 2005) used the index test as part of the reference standard and was subsequently excluded. Two further studies of newly developed chromogenic agar were subsequently included (Stoakes et al., 2006; Nsira et al., 2006) and two real-time PCR studies (Desjardins et al., 2006; Wren et al., 2006). Table 5-4 Relationship between the number of clinical risk factors and the diagnostic accuracy of clinical risk assessment Number of False-positive Positive predictive Number of Sensitivity risk factors rate value patients screened % 100.0% 4.4% 100.0% % 64.1% 5.9% 65.2% % 16.6% 11.9% 18.0% % 3.1% 15.4% 3.6% 17

25 Assessment of evidence The QUADAS tool was used to assess the quality of the studies included (Whiting et al., 2006) and the results are summarised in Table 5-5 and Table 5-6. Excluded studies are presented in Appendix 4. Study quality was fair in general, although two studies did not clearly define the criteria used to select patients for testing (Huletsky et al., 2005; Wren et al., 2006). The validity of all reports was compromised by an imperfect reference, a feature known to bias estimates of the performance of index tests. In addition, interpretation of the index or reference tests could have been influenced by knowledge of the results of the other. The studies did not report on whether clinical data were available when the test results were evaluated. The latter two points could be deficiencies in reporting rather than methodology. 18

26 Table 5-5 Characteristics of culture method studies Study Population Test(s) Reference standard Apfalter et al. (2002), Austria 579 screening swabs from 83 hospital inpatients; no pt details; results not stratified by site swabbed (nose=131, axilla=119, groin=70, throat=60, wound=61, perineum=34, device insertion site=24, other=80); year 2001 Screening agars ORSA (Oxoid) and MS-Oxa (mannitol salt oxacillin) MRSA identified by conventional method: BA 24 and 48 h in combination with BHI enrichment broth 48 h, from which 236 swabs from 43 pts were MRSA +ve Method Results Each swab plated on ORSA and MS- Oxa in random order, then on BA and into BHI enrichment broth, all incubated 24 and 48 h; if growth only in BHI 24 or 48 h the broth was plated as above. Typical S. aureus/mrsa colonies on BA, ORSA, MS-Oxa Pastorex Staph-Plus latex agglutination. Colonies on BA, ORSA oxacillin susceptibility test using disk or screening agar. All MRSA confirmed by PCR SA442, meca & femb. Sensitivity & specificity calculated from number of +/- swabs, sample site not specified. 24 h without BHI enrichment ORSA: Sensitivity=50.8% Specificity=95.6% PPV=88.9% NPV=73.9% MS-Oxa: Sensitivity=53.8% Specificity=92.7% PPV=83.6% NPV=74.5% 48 h without BHI enrichment ORSA: Sensitivity=68.2% Specificity=94.5% PPV=89.4% NPV=81.2% MS-Oxa: Sensitivity=65.7% Specificity=91.8% PPV=84.7% NPV=79.5% After BHI enrichment 48 h: 70/78 MRSA isolated on ORSA and MS-Oxa only after 48 h enrichment, this increases the agar 48 h incubation sensitivity to: ORSA: Sensitivity=97.9% [Specificity=91.8%, PPV=88.9%, NPV=96.6%] MS-Oxa: Sensitivity=95.3% [Specificity=94.5%, PPV=92.4%, NPV=98.4%] Turnaround time: 24 h presumptive MRSA identification ca 50% for both ORSA and MS-Oxa without enrichment 19

27 Study Population Test(s) Nsira et al., (2006) 666 screening samples, 648 nasal admission swabs, 18 other swabs MRSA Select chromogenic medium Stoakes et al., (2006) 2,125 nasal and perineal swabs from consecutive pts admitted to hospital MRSASelect (MRSAS) chromogenic medium Reference standard Mannitol salt agar (MSA), oxacillin resistance screening agar base (ORSAB), inhouse tobramycincolistin-fungizone medium (TCF) CHROMagar MRSA (CMRSA), and mannitol salt medium supplemented with oxacillin (MSA- OXA) or cefoxitin (MSA-CFOX) Method Results Swabs were inoculated onto all media and incubated for h and for further 24 h. TCF plates were incubated for 24 h. Meticillin resistance was checked or confirmed by Oxa-screen test. In cases of discrepancy, PCR was used to test for meca gene. Sensitivity and specificity based on number of +/- specimens. MRSA Select: Sensitivity=99.0% Specificity=99.8% MSA + oxacilllin screen: Sensitivity=97.0% Specificity=100% ORSAB: Sensitivity=91.9% Specificity=92.6% TCF: Sensitivity=94.1% Specificity=99.6% Turnaround time for all media not given, but incubations all 24 h. Swabs plated onto each of four media with reading at 18 and 48 h (MRSAS read at 18 h only). Colonies suspected to be MRSA were tested by multiplex PCR amplifying nuc, femb and meca genes, detected by ethidium bromide gels. Presence of meca with nuc and/or femb identified MRSA. Sensitivity and specificity based on number of +/- specimens. MRSAS: Sensitivity=97.3% Specificity=99.8% Turnaround time= available within 24 h. CMRSA: Sensitivity=82.9% Specificity=99.1% Turnaround time not given. MSA-OXA: Sensitivity=80.2% Specificity=79.0% Turnaround time not given. MSA-CFOX: Sensitivity=99.1% Specificity=84.8% Turnaround time almost 25% of MRSA strains required 48 h incubation. 20

28 Table 5-6 Characteristics of PCR method studies Study Population Test(s) Reference standard Desjardins et al. (2006) 287 specimens of nasal, rectal or wound, pooled or individual swabs from patients at high risk of colonisation Overnight broth incubation followed by IDI- MRSA assay using SmartCycler II device Broth subcultured to 5% sheep blood agar and screen for resistance on oxacillin salt agar screen plates (MH- Ox, Becton Dickinson) Huletsky et al. (2005), Canada 331 nose swabs from 162 adults screened during a hospital surveillance programme on the basis of risk factors for colonisation; year 2002 Real-time PCR using primers that formed the basis of IDI- MRSA, and Smart-Cycler (Cepheid) Standard MSA culture Method Results Swabs incubated in brain heart infusion broth overnight and subcultured. Aliquot of overnight broth also used for PCR analysis. MSSA isolates recovered from PCRpositive broths but that failed to grow on MH-Ox further characterised by cefoxitin disk diffusion, repeat testing on MH-Ox screen plated and determination of PBP2 expression by latex agglutination. Sensitivity and specificity based on number of +/- specimens. Sensitivity=96% Specificity=96% PPV=90% NPV=98% Post-implementation evaluation of IDI-MRSA assay of 4,201 broths gave PPV=65% (insufficient information to compute sensitivity, specificity and NPV). Samples incubated in broth overnight. PCR assay required less than 4 h additional time for a negative result (40 to 72 h for culture confirmation of a positive result). Routine cultures required 48 h for negative result (and 40 to 72 h as above to confirm MRSA positive). Swabs plated on MSA 24 and 48 h; DNA then extracted from the same swab for PCR. Typical S. aureus colonies on MSA BA catalase, tube coagulase, slide agglutination (Staphaurex); oxacillin susceptibility test using screening agar; MRSA confirmed by MicroScan Panel. Sensitivity and specificity based on number of +/- specimens. Sensitivity=100% (95% CI: 94, 100) Specificity=96.4% (95% CI: 93.2, 98.3) PPV= 89.4% (95% CI: 80.4, 94.7) NPV=100% (95% CI: 98.1, 100) Turnaround time: PCR min; at least 48 h for the culture method. results of PCR performed directly on nasal specimens can be obtained in <1 h. 21

29 Study Population Test(s) Warren et al. (2004), USA 288 nose swabs from hospital inpatients at high risk of colonisation; year 2003 Real-time PCR using of IDI- MRSA (Infectio Diagnostic), and Smart- Cycler II (Cepheid) Reference standard Culture by direct plating on MSA, enrichment (or either method) Method Results Each swab plated on MSA 24 and 48 h then into solution for PCR then into enrichment broth. Enrichment broth 24 h BA 24 h Typical S. aureus colonies on MSA and BA Gram stain, catalase, coagulase, Staph Latex agglutination (LifeSign) S. aureus from BA oxacillin screen agar to confirm MRSA. Sensitivity and specificity based on number of +/- specimens. PCR versus direct plating (MSA) Sensitivity=98.5% (95% CI: 95.5, 100) Specificity=92.8% (95% CI: 89.4, 96.2) PPV=80% (no CI) NPV=99.5% PCR versus enrichment before plating (BA) Sensitivity=92.4% (95% CI: 86.0, 98.8) Specificity=91.4% (95% CI: 87.8, 95.1) PPV=76.2% (no CI) NPV=97.6% Turnaround time: time from centrifugation* to PCR results for a batch of 8-10 specimens was 1.5 h; entire specimen handling time not reported (*does not include the whole specimen processing procedure). [PCR] allows same day results. 22

30 Model estimates for laboratory tests for MRSA The estimates used in the model are given in Table Decolonisation Study selection The literature search identified 1,203 titles and abstracts, from which full reports on the effectiveness of MRSA decolonisation procedures were selected. Studies focusing on the nursing home setting, MRSA outbreaks, healthcare workers, herbal remedies and studies published in Chinese and Japanese were excluded. Full reports of the studies selected from the initial search were assessed by two reviewers independently and disagreements were resolved by discussion. Ten studies were identified for inclusion at this stage. Following the updated search, one reviewer selected reports from identified titles and abstracts and a following review of these publications, a further five studies were identified for inclusion at this stage. One further study was identified from reference lists included in these publications. Studies were included if: patients were tested for MRSA colonisation in a hospital setting, decolonisation with mupirocin was attempted for positively-identified patients and clearance rates were reported. Studies assessing preoperative, prophylactic mupirocin use were excluded. The 16 studies meeting the selection criteria are summarised in Table Assessment of evidence Secondary evidence The Cochrane review selected and appraised RCTs assessing the effectiveness of topical and systemic antimicrobials among patients colonised with MRSA. Two studies considering the use of topical mupirocin were included (Harbarth et al., 1999; Parras et al., 1995). The study of Parras et al. (1995) involved decolonisation of both patients and staff members and did not present the results separately and so this publication was not considered further. The conclusion of the Cochrane review was that there is insufficient evidence to support the use of antimicrobial agents for MRSA decolonisation. Primary evidence The 16 selected primary studies included: three RCTs (one of which was placebo controlled), 10 observational studies (each comprising one cohort and no control group), and three retrospective chart review studies. The studies identified by this HTA assessed various intranasal mupirocin dosing regimens including a 2% preparation administered twice or three times daily over 5 days as recommended by Coia et al. (2006). Some studies did not clearly indicate the dosing regimen. Intranasal mupirocin was administered either as monotherapy or in combination with a body wash (chlorhexidine, octenidine, dihydrochloride) or with a body wash and a sulphonamide or oral antibiotics. The site of carriage was reported as nasal or multisite but for some studies, these were not specified. A number of studies were very small. Follow-up testing for MRSA colonisation after the start of mupirocin therapy occurred at timepoints ranging from 24 hours to 4 weeks. The definition of decolonisation failure in some studies was based on a single assessment, but in others was considered over a range of timepoints. Three RCTs were identified: the study of Harbarth et al. (1999) which compared the effectiveness of mupirocin with a placebo ointment the study of Dryden et al. (2004) which compared nasal mupirocin in conjunction with chlorhexidine gluconate soap and silver sulfadiazine cream for skin lesions with a regimen comprising nasal tea tree cream, tea tree body wash and tea tree cream for skin lesions the study of Horiuchi et al. (2006) which assessed the effectiveness of decolonisation regimens to reduce wound infections where a group of MRSA-positive patients received mupirocin treatment with some also treated with oral antibiotics was compared with a control group having no intervention and a group of MRSA-negative patients. Table 5-7 Laboratory test data Screening test Source of data Sensitivity False-positive rate Screening agar (ORSA) Screening agar (MRSA Select) Turnaround time (hours) Apfalter et al. (2002) 68.3% 5.55% 48 Nsira et al. (2006); Stoakes et al. (2006) 98.0% 0.2% 24 Real-time PCR Huletsky et al. (2005) Desjardins et al. (2006); Warren et al. (2004) 96.0% 5.0% 24 23

31 In the Harbarth et al. publication, of the 48 patients randomised to mupirocin treatment, 25% were found to be clear of MRSA carriage at any body site throughout the approximate one-month follow-up period, whereas 18% of the 50 patients given the placebo treatment were similarly clear of MRSA, which is a non-significant difference. Clearance of nasal carriage was 44% among the treatment group as compared with 25% in the placebo group, again a non-significant difference. Dryden et al. reported 48% of 114 patients cleared of MRSA at any site 14 days after receiving treatment with the mupirocin regimen as compared with 42% of 110 patients treated with tea tree; there was no significant difference between the effectiveness of the different treatments. Of patients with nasal carriage, the clearance rate was 78% for mupirocin treatment and 47% for the tea tree regimen, with mupirocin reported as being significantly more effective at clearing nasal carriage. Horiuchi et al. (2006) reported 100% effective decolonisation of the arm receiving the active intervention, although multiple courses were required for more than a third of the group. However, this trial was methodologically weak with unclear randomisation procedures and the assignment of patients to different active interventions was unexplained. The majority of the observational studies were very small, with only two comprising cohorts of more than 40 patients. These studies varied considerably in the groups of patients included (eg liver transplant patients, trauma patients, digestive diseases patients), the timing and sites of swab screening and in the reported effectiveness of the treatment for clearing treated patients of MRSA carriage. In general, better rates of decolonisation were reported in those studies with short follow-up periods, ie where patients were tested for colonisation within a week or so of treatment (eg Kampf & Kramer, 2004; Dupeyron et al., 2002) compared to those where testing was delayed or repeated at a later time point (Walker et al., 2003). Recolonisation of patients was observed in over 25% of patients initially cleared of MRSA carriage in the study of Dupeyron et al. (2002), although Rohr et al. (2003) indicated that the proportion of patients successfully decolonised increased between the initial swab following initiation of treatment and a second post-treatment swab; although it is not entirely clear if both swabs were taken following completion of the course of treatment. In a number of studies, mupirocin was administered in combination with oral antibiotics for at least some of the patients in the study cohort (Austin & Austin, 2001; Darouiche et al., 1991; Fung et al., 2002; Hill et al., 1988; Horiuchi et al., 2006; Muller et al., 2005). Although patients received mupirocin with or without systemic therapy in the studies of Austin & Austin (2001) and Hill et al. (1988), data for the mupirocin monotherapy group were not presented separately and therefore the effectiveness of a regimen which did not include systemic antibiotics could not be determined. The study of Dupeyron et al. (2002), which assessed the effectiveness of clearance of nasal carriage only, reported the greatest proportion of patients decolonised in the short term. The authors indicate that this is similar to other reports of clearance of S. aureus and the lower rates reported in the clinical of Harbarth et al. (1999) may be attributable to multisite carriage of MRSA. One study (Walker et al., 2003) required patients to have two positive MRSA cultures 7 days apart, prior to initiation of mupirocin decolonisation therapy. This study would therefore indicate the patient minimum length of stay to be at least 9 days prior to initiation of mupiricon therapy (assuming 2 days time delay before the second set of test results became available). This is beyond the mean length of stay considered applicable for NHSScotland. Due to the extent of variability across studies, it was difficult to meaningfully combine the estimates of successful mupirocin decolonisation. An estimate of effectiveness would ideally be based on the context in which mupirocin would be used in NHSScotland, ie as monotherapy or in combination with a body wash for patients testing MRSA positive on admission to a high-risk hospital specialty unit, and where the mean length of hospital stay is approximately 5 days. 24

32 Table 5-8 Characteristics of decolonisation studies Study Study design Interventions Study aim Population Results Austin & Austin (2001), Canada Retrospective chart review of adults identified as MRSA decolonised Systemic therapy, topical mupirocin with or without systemic therapy. To examine the success rate of mupirocin/ chlorhexidine therapy in a group of hospitalised pts colonised with MRSA 146 pts in an adult tertiary care hospital. 32 received systemic therapy only, 114 received mupirocin with or without systemic therapy. Clearance: 6.25% systemic therapy only,35.4% mupirocin and systemic therapy. Darouiche et al. (1991), USA Observational study with one group and no control Two-week course of oral antibiotics consisting of 100 mg of minocycline bid and 600 mg of rifampin each morning and intranasal mupirocin 2% bid for 5 days (only if nasal colonisation) To control the spread of MRSA 11 pts with MRSA colonisation colonised ( 1 site) in a spinal cord injury unit. 91% clearance [10/11] Clearance: Absence of MRSA at previously positive sites on follow-up cultures on Day 7 and Day 18 after enrolment. Dryden et al. (2004), UK RCT Intranasal mupirocin 2% tid and chlorhexidine gluconate 4% soap and silver sulfadiazine 1% cream (for skin lesions) (standard treatment 5 days),tea tree oil regimen (5 days). To compare the efficacy of topical tea tree preparations with the standard regimen for the eradication of MRSA colonisation in patients in an acute-bedded hospital and to assess their safety 236 pts colonised ( 1 site) with MRSA, within a hospital. 114 standard treatment, 110 tea tree oil regimen, 12 not evaluable. Eradication of MRSA carriage: 49% standard therapy, 41% tea tree oil regimen, Eradication defined as: the absence of MRSA in both sets of post-treatment swabs (after Day 2 and Day 14). Notes Retrospective. No dose levels given for mupirocin or systemic treatment. Details of systemic therapy not provided. Results not separately presented for mupirocin only group. Unclear if the same clearance criteria were used to assess clearance for each patient. No control group. Mupirocin administered in combination with other antibiotics, therefore effect of mupirocin alone cannot be assessed. Only pts with nasal colonisation received mupirocin (6 pts). No information about identification as an MRSA carrier on admission to hospital. 25

33 Study Study design Interventions Study aim Population Results Dupeyron et al. (2002), France Observational study with one group and no control Intranasal mupirocin 2% tid and alternating whole body wash, 4% chlorhexidine solution or soft liquid soap on alternate days (5 days). An aim of this study was to evaluate the efficacy of MRSA nasal carriage eradication with mupirocin. 86/117 pts with nasal carriage treated with mupirocin in a digestive disease unit. 57 carriers on admission and 60 acquired during hospitalisation. 98.8% clearance [ie 85/86]. 25.9% became nasal carriers again after several weeks. Clearance: the last nasal swab culture taken at least one week after the last day of therapy was negative. Fung et al. (2002), Canada Retrospective chart review of adult inpatients with MRSA decolonisation therapy Intranasal mupirocin 2% tid and alternating whole body wash, 4% chlorhexidine solution, rifampin (300 mg bid) and either trimethoprim/sulfamethoxazole (160 mg/800 mg bid) or doxycycline (100 mg bid). Treatment period 7 days. How to eradicate MRSA colonisation in hospitalised patients. 207 pts in tertiary care with MRSA colonisation hospitalised between Feb 1996 and Mar of 207 pts received decolonisation. Of 103 decolonised pts, only 9 received treatment with 2% mupirocin alone. Other patients received combination antimicrobial therapy. Notes Title of study indicates clinical trial, but no information is provided about the randomisation or control arms. This appears to be an observational study with no control. Clearance results not presented separately for patients with MRSA colonisation on admission and that acquired during hospitalisation. No explanation as to how the 86 pts treated were selected from the pts with colonisation became infected with MRSA during study. As this is a retrospective chart review, there is much variability with respect to choice of decolonisation therapy. Clearance rates not presented separately for the mupirocin only group as this is not the standard decolonisation therapy at the hospital. 26

34 Study Study design Interventions Study aim Population Results Harbarth et al. (1999), Switzerland Randomised, placebo controlled, double-blind trial Intranasal mupirocin 2% bid for 5 days; placebo. Both groups used chlorhexidine soap for body washing. The assessment of the clinical efficacy of mupirocin in eradicating overall MRSA carriage in patients who were MRSA carriers on admission to the hospital or became colonised during the hospital stay. 98 pts colonised ( 1 site) with MRSA: 48 mupirocin group and 50 placebo group. Tertiary acute care and geriatric long-term care facility. Surgical, geriatric, multiple comorbidities and multiple MRSA colonisation sites included. Concomitant systemic antistaphylococcal therapy excluded. Overall MRSA eradication rate 25% mupirocin/ 18% placebo 26 days following initiating therapy. Hill et al. (1988) Observational study, one group, no controls 2% nasal mupirocin. Chlorhexidine gluconate added to bathwater of patients with MRSA. Evaluation of mupirocin for the eradication of nasal carriage of MRSA by patients and staff. 40 pts and 32 staff with nasal carriage of MRSA from at least two consecutive swabs. Of 40 pts, 32 (80%) remained clear of nasal MRSA for duration of their follow-up (1 day to 9 weeks, mean 2 weeks). Of 32 staff, all were decolonised after one week, but 5 became recolonised at between 2 and 12 weeks post intervention. Notes Definition of eradication failure: If any of the screening cultures in the month following treatment were positive for MRSA and if the strain was identical with the original strain. Note, it is unclear if all patients had swabs taken at each specified timepoint ie 12, 19 and 26 days after initiating therapy. Authors note the variety of MRSA pts and the illness severity. 58% of subjects had documented nasal carriage at baseline. 16 pts had had previous therapy with systemic antibiotics against MRSA and 4 pts were receiving vancomycin at start of trial. 27

35 Study Study design Interventions Study aim Population Results Notes Horiuchi et al. (2006) Randomised trial Intranasal application of combinations of mupirocin, arbekacin inhalation and oral sulfamethoxaxole/ trimethoprim To assess efficacy of nasopharyngeal decolonisation of MRSA to reduce wound infection post percutaneous endoscopic gastrostomy (PEG) placement 72 pts hospitalised for PEG. 48 MRSApositive pts and 24 MRSA-negative pts. MRSA-positive pts in two groups, one group received decolonisation treatment. All 24 MRSApositive pts in decolonisation arm successfully decolonised using one of the following regimens: 5 days mupirocin three times daily; intranasal mupirocin plus inhalation of arbekacin sulphate twice daily; intranasal mupirocin plus inhalation of arbekacin sulphate plus oral sulfamethoxzole trimethoprim (800 mg/160 mg) twice daily. Not clear if randomisation of pts was carried out. Not clear which pts assigned to which treatment regimen or if treatment regimens tried consecutively to achieve decolonisation. Some pts had two rounds of treatment to achieve decolonisation. Not clear how long to achieve decolonisation in these pts. Only 12.5% (3 pts) were successfully decolonised by one course of intranasal mupirocin. 2% of pts in decolonisation arm developed MRSA infection compared with 100% in MRSApositive no decolonisation arm and 0% in MRSA-negative arm. 28

36 Study Study design Interventions Study aim Population Results Kampf & Kramer, (2004) Observational study of patients in six hospitals, no control Daily application of Stellisept scrub and twice daily mupirocin for 5 days. To assess the efficacy of Stellicept in combination with mupirocin for decolonisation of patients with MRSA. 35 pts with dermal colonisation without concomitant systemic therapy (21 of 35 with nasal colonistion). After one treatment cycle, 25/35 (71.4%) were found to be free of MRSA. A further treatment cycle resulted in 7/8 pts being decolonised (two were discharged in interim) and a third treatment cycle decolonised the remaining pt. Follow-up 5 days post treatment (2 days wash out, 3 days screening culture). Muller et al. (2005) Retrospective observational study Screening at hospital admission and once per week during hospital stay. Nasal mupirocin for 5 days. To determine the effectiveness of intranasal mupirocin prophylaxis in preventing ICUacquired infections with MRSA. Two study periods. Period 1: 47 pts colonised or infected with MRSA were treated with mupirocin. Period 2: 66 pts; no mupirocin treatment. 86.4% of MRSApositive pts in Period 1 successfully decolonised and 46.4% in Period 2 [spontaneous decolonisation]. Notes No control group. Treatment failures after first round were in different area of skin in four of 10 pts, in same area in five of 10 pts and presistant nasopharynx in one of 10 pts. Successful eradication of MRSA not confirmed by three consecutive swabs in a number 10 pt due to early discharge. MRSA considered to be eradicated when two consecutive screening tests were negative. Study included patients hospitalised for more than 14 days. No distinction between patients colonised and infected with respect to efficacy of mupirocin and three of 19 pts in Period 1 received concomitant treatment with vancomycin. 29

37 Study Study design Interventions Study aim Population Results Neumaier et al. (2006) Observational study Limited isolation and 5 days mupirocin treatment. To assess the incidence of nasal MRSA colonisation in trauma patients, the relationship between colonisation and infection and the efficacy of mupirocin therapy. 643 trauma pts screened, 13 pts found positive and MRSA eradication attempted in 8 pts. All eight treated patients were successfully decolonised although one patient became recolonised 41 days after the initial positive screening. Paterson et al. (2003) Prospective observational study Mupirocin calcium ointment applied twice daily for 5 days. To evaluate effectiveness of mupirocin to prevent S. aureus infections in liver transplant patients. 31 of 70 pts positive for S. aureus, 9 pts with MRSA, 22 pts with MSSA. Follow-up cultures 5 days post final application (or 2-14 days in discharge patients) of mupirocin indicated 8 of 9 MRSA pts to be decolonised, but subsequently three became recolonised (unclear at what time point post successful decolonisation). Notes Not clear exactly how successful eradication was determined as not all patients would appear to have had an MRSA test post treatment or at the follow up at 6 months. No control population. 23% of patients developed an S. aureus infection despite mupirocin treatment. Only nares decolonistion attempted no attempt to eradicate from other sites. 30

38 Study Study design Interventions Study aim Population Results Rohr et al. (2003), Germany Observational study, one group, no control Intranasal mupirocin and octenidine dihydrochloride whole body wash, 5-7 days. Evaluate the efficacy of the decolonisation regimen used in the hospital, to reduce MRSA multisite skin carriage. 32 newly colonised ( 1 site) MRSA extranasal carriers within a tertiary hospital. Pts showing only nasal carriage were excluded. 25 surgical ward, 7 medical ward. MRSA was eradicated in 53% and 68% of pts, hours and 7-9 days after treatment, respectively. No adverse events resulting from the decolonisation regimen. Semret & Miller, (2001) Prospective observational study Colonised patients cohorted or placed in private isolation; chlorhexidine gluconate twice weekly; decolonisation usually attempted with topical mupiroicin. To assess the effect of mupirocin resistnce on success of mupirocin to decolonise patients with MRSA. 71 pts enrolled. 38 had mupirocin resistant strains. At follow-up cultures at minimum of 72 h, (mean 8 days) 68% of pts colonised with mupirocinsensitive strain were successfully decolonised compared with clearance of 52% of pts with mupirocin-resistant strain. Notes Mupirocin dose level not specified. Patients with nasal colonisation alone, excluded and not clear which patients were selected for decolonisation treatment. Patients isolated on detection of MRSA colonisation. Study run over two separate time periods: Oct 1998 to Nov 1999 and Nov 2000 to Aug No control group. Text would indicate not all patients received mupirocin treatment. Not clear which patients or why. Discussion notes that practice in this hospital was application of mupirocin four times daily for 2 weeks. Not clear why timing of tests for successful decolonisation varied. 31

39 Study Study design Interventions Study aim Population Results Sloot et al. (1999), Germany Observational study, one group and no control Intranasal mupirocin tid and 1:1 dilution of octenidine dihydrochloride whole body wash, 5 days. To study the efficacy and safety in MRSA colonised patients, of whole body washing with an octenidine dihydrochloridebased antiseptic as a supplement to nasal mupirocin treatment, in order to shorten the costly period of isolation. 28 isolated pts in whom MRSA was detected during routine screening or in clinical examination material.13 surgery, 6 ICU, 5 internal medicine, 3 cardiac surgery, 1 dermatology department. Both infected and colonised pts were included. Clearance: 75% [21 out of 28] Swab time points: pre-washing, Day 4 of treatment Days 1, 4 and 7 post treatment. Side effects: 4/28 pts experienced redness of the skin during the washing programme. Walker et al. (2003), USA Observational study with one group and no control Intranasal mupirocin 2% bid for 5 days. To determine the efficacy of mupirocin ointment in reducing nasal colonisation with mupirocinsusceptible MRSA and mupirocinresistant MRSA. 40 hospitalised pts with two anterior nares culture positive for MRSA colonisation within a 7-day period. Clearance 3 days post treatment: 80% low-level MRSA, 28% highlevel MRSA. Clearance 2 to 4 weeks post treatment: 25% for low and high-level MRSA. Notes Dose level of mupirocin not stated. Five pts screened on admission. Results not presented separately for infected/colonised patients. Of the 21 pts, only 12 were monitored for the full 7 days post treatment. It is unclear if MRSA colonisation detection was at admission. Patients had two positive MRSA cultures 7 days apart to assess colonisation. Post-period assessments: Day 3, and 2-4 weeks post treatment. 32

40 Model estimates for mupirocin decolonisation clearance rates In the absence of more robust data, a 53% mupirocin decolonisation clearance estimate was assumed for the economic model based on the study of Rohr et al. (2003), where results were reported at 1 to 2 days after the end of 5-7 day mupirocin treatment Association between colonisation and infection Study selection A search of the literature identified 495 titles and abstracts from which 33 full reports were selected and obtained. These publications were reviewed by one reviewer and included if the report was of a primary study where the proportion of MRSA-colonised hospital inpatients who became infected with the organism could be determined. Seventeen studies matched these criteria. The majority of these studies were conducted within a single specialised hospital unit or group of patients. Seven studies were of ICU patients (Corbella et al., 1997; Garrouste-Org et al., 2001; Keene et al., 2005; Khurram et al., 2004; Mest et al., 1994; Pujol et al., 1996; Truffault et al., 2000), three studies were of orthopaedic patients (Khan et al., 2002; Levy et al., 2004; Sott et al., 2001) and four studies were of patients in specialised liver units (Bert et al., 2000; Desai et al., 2003; Dupeyron et al., 2001; Campillo et al., 2001). The proportion of patients colonised with MRSA who developed an MRSA infection in these studies of high-risk patients varied between 0 and 87%, with some studies being very small cohorts of patients, and the periods of follow-up identified also varied considerably. The remaining three studies were conducted among mixed groups of patients. The studies meeting the selection criteria are summarised in Table

41 Table 5-9 Colonisation and infection studies Study Aim Study design Population Bert et al. (2000) Nasal carriage and risk of infection in liver transplant patients Prospective laboratory testing of nasal swabs and follow up of 2 months Consecutive liver transplant pts who underwent screening for S. aureus between Sep 1996 and Oct 1998, France. Screening and infection control Nasal swabs day prior to surgery, prophylactic antibiotics. Microbiological and medical records reviewed for 2 months post-liver transplant. Results Notes 87 pts included in study. 24 had MSSA, and 8 had MRSA (9.2%). Within 2 months following transplant, five pts developed MSSA infection and 15 MRSA infection. Of the 8 MRSA carriers, 7 developed an MRSA infection (87.5%). Of these infections, isolates pre and post implantation had similar pattern in three pts, closely related in two and distinct patterns in two. Mean time to MRSA infection 14.1 days post surgery. 6 of 55 non-carriers (13%) developed MRSA infection. Comparison of MRSA and MSSA carriers showed significantly longer pre-operative stay amongst MRSA carriers and more likely stay in ICU of 3 days. Also trend to higher disease severity score and frequency of alcoholinduced cirrhosis in MRSA patients. 34

42 Study Aim Study design Population Coello et al. (1997) Risk factors for developing MRSA infection among colonised patients Prospective follow up of pts identified as colonised All pts identified as being colonised but not infected in 2 years from Nov 1990, teaching hospital, Spain Corbella et al. (1997) Association between S. aureus colonisation and infections Prospective laboratory testing Jan to Dec 1994 ICU pts, tertiary centre, Spain with ICU stay of 48 h or more Screening and infection control Colonised pts those with positive specimens from nares, throat, perineum, groin or axilla, without clinical symptoms of infection. Screening specimens taken from roommates of known carriers, and patients exposed to MRSA +ve staff. Prospective screening at admission and weekly thereafter. Identification of infection by clinical specimen testing. Results Notes 479 colonised pts identified, 53 (11.1%) developed 68 MRSA infections. Varied by department (6% medical, 11% surgical, 30% ICU). Risk of infection associated with use of invasive devices, previous use of antibiotics, presence of ulcers. Multivariate analysis indicated surgical wound, pressure ulcers and intravenous catheter independent risk factors for infection. Used Cox regression model to avoid bias due to colonised pts without clinical problems being discharged after short hospital stay. 752 pts included in study, 3 MRSA +ve at admission. 14 pts acquired MRSA during stay. 7 pts became MRSA infected, 2 post colonisation, 5 at time or prior to colonisation. In addition, 3 pts colonised at admission all became infected ie 10 of 17 total. No detail of time from colonisation to infection. PFGE analysis indicated all MRSAinfected pts harboured same clone. Authors indicated not possible to quantify risk of nasal colonised pts to develop infection because of nature of study. 35

43 Study Aim Study design Population Davis et al. (2004) Prevalence of MRSA infection at admission and impact on subsequent MRSA infection Prospective study with follow up for threemonth study period and one year thereafter All pts admitted to five representative hospital units between Jun 2002 and Aug 2002 eligible, USA. Desai et al. (2003) Comparison of liver transplantation patients with and without MRSA colonisation and subsequent outcomes Retrospective review of database of liver transplantation patients Consecutive liver transplant patients 1998 to 2001, tertiary centre, UK. Dupeyron et al. (2001) Relationship between S. aureus colonisation and subsequent infections during hospitalisation Prospective lab testing for S. aureus and gram-negative bacteria resistant to CephRs Pts admitted to rehabilitation unit with liver cirrhosis over 30-month period (Jan June 1998). Screening and infection control Nares culture within 48 h of admission, at transfer to other units, weekly during hospital stay and at discharge. Infection defined as recovery of MRSA from normally sterile sites (blood or urine) or sites concomitant with diagnosis of infection. Laboratory testing of patients pre and post transplantation Admission screening of nasal and stool specimens within 48 h of admission Results Notes 758 pts had nares swabs taken, 26 were colonised at admission with MRSA. Of these, 5 pts became infected with MRSA (19%). 1.5% of MSSA-colonised pts developed MRSA infection and 2% of uncolonised pts developed MRSA infection. Comparisons between MSSA and MRSA infection rates made. 157 of 374 patients were screened. 35 colonised pretransplantation. Carriers had higher scores in indices of severity of disease. 31% of carriers developed MRSA infection vs 9% of non-carriers (P=0.002). No adjustment for severity of disease in quantifying association between post-transplantation infection and carrier status. 551 pts in study, 96 were MRSA +ve on admission, (79% +ve nasal swabs). 34 cases of MRSA infection, 22 of these in MRSA carriers, ie 23% of carriers developed MRSA infection vs 2.6% of non-carriers. Severity of disease was measured, but not included in analysis of association between colonisation and infection. 36

44 Study Aim Study design Population Garrouste- Org et al. (2001) MRSA colonisation and occurence of S. aureus infections, use of glycopeptides and outcome in ICU patients Prospective study of admission screening and subsequent follow up Medical/surgical ICU pts from July 1995 to July 1998, France. Huang & Platt, (2003) Risk of MRSAassociated morbidity amongst MRSA colonised pts including during post-discharge period Retrospectively identified colonised/infected pts with subsequent follow up of 18 months Pts for whom MRSA infection control procedures had been newly initiated between Apr-Jun or Oct-Dec Screening and infection control Admission nasal samples tested and weekly thereafter (on Tuesdays). Pts isolated until known to be not colonised. Nares specimens taken for surveillance purposes and clinical specimens when clinically indicated. Results Notes Of 1,044 pts admitted, 106 (10.1%) found to be colonised with MRSA (54 on admission, 52 ICU acquired). 35 pts had 42 MRSA infections (MRSA and MSSA). Of these 35, 29 had been colonised with MRSA. (Assuming all MRSA infections, 27% of colonised pts became infected). No attempt to determine if isolates from clinical infection specimens were same as those colonising the pt. Authors report that univariate analysis showed association between infections and severity of illness, intubation and MRSA infection on admission. Of 435 pts included in study, 97 were identified as colonised (112 as infected) and of these 97, 30 developed subsequent infections (31%). Median time to developing subsequent infection was 29 days (but this includes pts with initial MRSA infection who developed subsequent infections). 37

45 Study Aim Study design Population Keene et al. (2005) Incidence of S. aureus infections following nasal or tracheal colonisation in ICU pts Prospective screening and follow up for ICU stay and 30 days post discharge from ICU, or death or hospital discharge. All pts in medical ICU from Feb to June 2002 on Monday and Thursday. Khan et al. (2002) MRSA colonisation and morbidity from subsequent infection Prospective laboratory testing for MRSA with post-operative follow up. All pts presenting with neck of femur fracture in Mar and Apr 2000, UK hospital. Khurram et al. (2004) MRSA colonisation and risk factors for infection Retropective casecontrol study by review of patient notes. ICU pts from Jan 1998 to Dec 1999, Karachi, Pakistan. Screening and infection control Nasal and tracheal samples taken and tested for S. aureus. Pre-operative screening for MRSA within 24 h of admission and before administration of prophylactic antibiotics. Pts who had clinically significant infection with MRSA at any time during stay were cases. Controls were those who remained colonised with MRSA but did not develop infection. No details of screening procedures used or frequency. Results Notes 208 of 325 admissions to ICU enrolled in study. 14 had nasal colonisation (11 at admission) with MRSA (33 with MSSA), five of 14 MRSA-colonised pts developed MRSA infection (36%). Mean time to infection of colonised patients: 11 days (6 days in those without colonisation). Comparisons made of risk factors associated with MRSA vs MSSA reported. 66 pts screened, four MRSA positive. No pts developed deep wound infection or other MRSA infection within first week post operatively. No wound infections developed post discharge. Small number of MRSA colonised pts. Of 1431 pts admitted, 57 were MRSA positive and did not develop an infection within first 48 h. 37 pts went on to develop an MRSA infection (65%). Risk factors which indicated a significant association with case status were diabetes mellitus, presence of a central line and age. Authors assumed risk factors were independent when conducting analysis. 38

46 Study Aim Study design Population Levy et al. (2004) Association between colonisation and infection with MRSA following femoral fracture of nursing home vs pts living independently Prospective laboratory testing and follow-up over 20-week period (dates not given). Pts with femoral fracture presenting at A&E (UK) Mest et al. (1994) Nasal carriage of MRSA perioperatively and risk of infection Prospective laboratory testing of nasal swabs and follow up. All pts of 15-bed SICU between Aug 1991-Jul1992, Veterans hospital, USA. Pasdeloup et al. (2000) MRSA and Acinetobacter baumanii colonisation and nosocomial infection in medical ICU pts Prospective study of admission screening MRSA colonisation, acquired colonisation and infection. Medical/surgical ICU pts from June 1993 to June day study period, France. Screening and infection control Screening for MRSA colonisation in A&E with follow up until wound healing or deep wound infection identified. All pts of SICU swabbed within 1 h of admission. Pts observed for duration of stay in SICU for evidence of infection by same ICN. Nasal, throat and digit skin swabs taken at admission and weekly intervals thereafter. Results Notes Of 105 admissions, 11 colonised (3 of 70 from own home, 8 of 35 from nursing or residential home or hospital). Two of 11 pts developed deep wound infection compared with 2 of 94 pts uncolonised at admission. 10.5% prevalence at admission. 18% of colonised pts became infected. Authors indicate prophylaxis should be directed at pts at high risk of carriage. 484 pts included in study. 19 (3.9%) colonised with MRSA. 11 post-operative MRSA infections, 5 among 19 (colonised pts (26%), 6 in other 465 pts (1.3%). Mean time to infection of colonised pts = 16.6 days. Multivariate analysis for MRSA infection showed perioperative nasal colonisation and prior spinal cord injury stay were significant risk factors. Of the four paired isolates available, infecting strain had same immunoblot pattern as colonising strain. 103 pts hospitalised for more than 7 days included. 10 pts were colonised at admission with MRSA. In French - from abstract only. 39

47 Study Aim Study design Population Pujol et al. (1996) Nasal carriage of S. aureus and risk of nosocomial S. aureus bacteremia Prospective cohort, nasal swabs and follow up to discharge or death. All pts admitted to ICU Mar 1991-Apr 1992 for more than 48 h. Tertiary hospital, Spain. Sott et al. (2001) To determine if screening for MRSA reduces risk of post-hip replacement sepsis with MRSA Intervention group (conducted 1996) compared with historical controls (1995) Pts admitted for total hip replacement Truffault et al. (2000) MRSA colonisation and subsequent infection Retrospective review of medical and laboratory reports Patients admitted for at least 48 h to surgical ICU from Apr 1996 to Dec 1997, France. Screening and infection control Weekly screening for S. aureus carriage. Blood cultures when bacteremia suspected by attending physician. Intervention group screened pre-admission and decolonisation using nasal mupirocin and chlorhexidine baths for until clear set of screening swabs obtained. Infection rate compared with historical controls where no pre-admission screening or intervention. Screening for nasal and perineal colonisation on admission and at weekly intervals. Results Notes 488 pts admitted for more than 48 h, 63 (12.9%) were MRSA carriers - 56 developed colonisation after first week in ICU. 24 (38%) of MRSA carriers developed S. aureus bacteremia vs 6 (1.7%) of non-carriers of S. aureus [8(9.5%) of MSSA carriers developed bacteremia]. After adjustment for severity of illness and presence of 3 intravascular catheters, RR for S. aureus bacteremia was 3.9 (95% CI 1.6, 9.8; P=0.002) for MRSA carriers compared with MSSA carriers. Molecular typing indicated same susceptibility pattern in nasal swabs as blood cultures. Of 113 pts in historical control group, five developed an MRSA infection (4.4%). Four of 123 pts in intervention group were decolonised and none developed MRSA, but one pt from this intervention group infected (0.8%). No statistically significant difference. Increased awareness of MRSA may have resulted in highlighting other aspects of infection control. 540 pts files reviewed, 15 were colonised at entry and 63 became colonised during their stay. 20 of the 63 colonised pts became infected (32%).Median delay between colonisation and infection 5 days. In French - from abstract only. 40

48 Assessment of the evidence Huang & Platt (2003) in a study of a cohort of patients (both inpatients and outpatients) at a general hospital in the USA found that of the 97 patients identified as colonised, 31% subsequently became infected with the organism over an 18-month follow-up period. The time to subsequent infection of this group cannot be assessed from the report as the study included patients identified as infected with MRSA who went on to develop another MRSA infection in addition to patients only colonised. Davis et al. (2004) in a prospective study assessed the risk of developing an MRSA infection following a positive admission swab indicating colonisation. The study involved patients of five units within a tertiary care hospital in the USA. Patients were followed for one year after the study period. Of the 26 patients colonised at admission with MRSA (3.4% prevalence), five developed an infection (19%). In addition, 12 patients acquired MRSA colonisation during their hospital stay and a quarter of these patients also developed a subsequent infection. Coello et al. (1997) prospectively followed all patients identified as being colonised with MRSA over a two-year period at a teaching hospital in Spain. Overall, 11% of 479 colonised patients developed a subsequent MRSA infection. A multivariate analysis showed that surgical wounds, pressure ulcers and intravenous catheterisation were independent risk factors for developing an infection. In addition, ICU patients had a significantly higher MRSA infection rate during the first 4 days after admission when compared with medical patients. This study indicated that 6% of medical ward patients, 11% of surgical patients and 30% of ICU patients colonised with MRSA subsequently developed an MRSA infection Model estimates for MRSA colonisation and infection For the purposes of the economic model, it has been assumed that 6% of colonised patients being admitted to low-risk specialties will become infected and 30% of those admitted to high-risk specialties will become infected Transmission rates For the purposes of the model, a measure of the intrinsic MRSA transmissibility, in terms of the basic reproduction number per colonised admission, was required. Previous authors have discussed why assuming independence in empirical studies is invalid, given that the majority of HAI control studies use these statistical methods (Cooper & Lipsitch, 2004; Cooper et al., 2003). The assumption of independence is inappropriate, given that the chance of a patient becoming infected is not independent of the MRSA status of the other patients in the hospital population. In addition, the majority of empirical transmission studies have been conducted in single wards or units, which may not be relevant to the whole hospital setting. In the absence of robust empirical data, various groups have used mathematical models to estimate transmission rates. Factors likely to account for differences in MRSA transmission rates between populations include: antibiotic use, infection control policies, the ratio of staff to patients and the size and type of the hospital or ward (Smith et al., 2004). As rates would be expected to vary according to local conditions and MRSA strains, Cooper et al. used transmission rates ranging from to transmissions per source per susceptible patient day in an HTA on isolation policies for the management of MRSA (Cooper et al., 2003). The size of the hospital ward in which patients are accommodated is a key element in determining the transmission rate, as colonisation pressure (the ratio of colonised to susceptible patients) is closely associated with MRSA acquisition (Merrer et al., 2000). The transmission rates used in the model of Cooper et al. (2003) were based on the assumption that all patients in the hospital were accommodated as if on one large ward. For the HTA presented here, the model assumes that patients are accommodated in 25-bed wards. Whilst closer to the situation in Scottish hospitals, it is unlikely that this precisely matches the situation for many patients. In order to replicate more closely the pattern of patient accommodation in Scottish hospitals, a more sophisticated mathematical model would be required. The transmission rates used by Cooper et al. were based on earlier modelling (Cooper et al., 1999). Section 6 details the methods used for estimating the transmission rate, given the prevalence of colonisation and hospital size and occupancy. Essentially transmission rate is based on the MRSA prevalence on admission and discharge from the hospital. As it was clear in the early stages of structuring the economic model that information on transmission rates was unlikely to be present in the literature, no specific database search was conducted (Cooper et al., 2003). However, in the process of assessing references from other searches, a number of publications were identified which included information relating to transmission. These included both modelling and empirical studies of MRSA transmission and the publications are summarised below. Grundmann et al. (2002) developed a stochastic model to predict the effectiveness of infection control interventions. The authors computed a theoretical basic reproduction number of as high as 10 in the absence of any infection control measures, using empirical data from a study conducted in a UK tertiary referral hospital. However, the combined effects of cohorting and handwashing compliance reduced the average number of secondary MRSA cases to Hotchkiss et al. (2005) modelled pathogen transmission in an ICU and used a per encounter acquisition probability of 5 or 10% (ie the contact between a colonised patient and carer, which transiently colonised the carer, and the subsequent contact with an uncolonised patient), which gave a per patient dissemination risk of around 0.1 to 0.4 per unisolated patient day. 41

49 5.3.7 Isolation of patients and MRSA control The systematic review of Cooper et al. (2004) detailed the methodological weaknesses of studies reporting infection control interventions for MRSA management. However, the authors concluded that there was evidence suggesting that isolation as a feature of MRSA infection control policies could reduce hospital infection prevalence. Although it was anticipated that little more robust evidence would have been published since the Cooper et al. review, a literature search was conducted (Appendix 2) to identify any recent studies assessing the relative effectiveness of isolation policies Study selection The literature search identified 2,778 titles and abstracts published since 2001 (the systematic review of Cooper et al. reported on literature from 1966 to 2000). Papers describing interventions using isolation facilities or dedicated nursing staff to reduce MRSA infection rates, where MRSA was endemic in an acute hospital setting, were selected. Papers reporting interventions to control MRSA outbreaks were excluded, as were those considering long-term care facilities. The criteria of Cooper et al. were applied, to include prospective studies involving an isolation ward or unit with nurse cohorting. Seven reports were identified that met the assessment criteria and these are summarised in Table

50 Table 5-10 Studies of isolation or nurse cohorting to control MRSA Study Setting Design Main interventions Biant et al. (2004), UK Elective orthopaedic ward (28 beds) Prospective before and after study in two phases over 2 years ( ) Ward ring fenced not to admit pts who had ever had MRSA (they were isolated on another orthopaedic ward), screening and decolonisation in the community, nurses disposed of apron and gloves after each patient contact and instructed to use alcohol hand rub. Cepeda et al. (2005), UK Three general medical/surgical ICUs in two hospitals (18, 4 and 10 beds) Prospective ITS in three phases over one year (2000 to 2001) Phase 1 and 3: MRSA colonised or infected pts moved to single room or cohorted bays. Phase 2: No movement of MRSA pts (unless carrying other multiresistant or notifiable pathogens, or in need of protective isolation). Standard infection control measures reinforced throughout including glove/apron removal on/before exiting patient s room and avoidance of sharing of equipment; compliance with hand hygiene was audited. Patient outcomes Before: 43 post-operative infections, 9 MRSA out of 417 pts. After: 15 postoperative infections, 0 MRSA, among 488 pts. No evidence of increased transmission during the non-move phase shown using unadjusted Cox proportional hazards model. Hand washing compliance was 21%. Assessment of evidence Observed change possibly influenced by regression to the mean and chance effects. Potential confounding factors not reported. Unable to differentiate effects of various simultaneous infection control measures. Included only pts in ICU 48 h, the minimum for screening to become available. 80 to 87% of pts screened at admission, movement of patients not initiated until results of screening were obtained, approximately 3 days post screening. Model adjusted for potential confounders and colonisation pressure, sensitivity analyses did not change the findings. 43

51 Study Setting Design Main interventions Patient outcomes Assessment of evidence Fitzpatrick et al. (2000), Ireland Cohort unit (11 beds) Observational study to evaluate the unit over the first 6 months (1999) Purpose built MRSA unit, screening, decolonisation, supervised cleaning protocol. Cohorting did not decrease successful decolonisation or increase the risk of staff colonisation. Assessment confined to the cohort unit, which the authors do not view as an isolation unit. No pt MRSA transmission data. Johnston et al. (2005), UK Two orthopaedic and one general surgery ward (case ward 30 beds, control wards bed numbers not reported) Prospective study over 83 months ( ) comparing a ward with a patient segregation policy to two control wards without segregation Preadmission screening and pt segregation in designated MRSA positive isolation bays on the segregation ward with separate nursing staff. Average number of new cases per month: segregation ward 1.17 (SD 1.41), control wards 3.57 (SD 0.71) and 3.18 (SD 2.12). Analysis did not include potential confounding factors. Comparability of intervention and control populations not reported. Unable to differentiate effects of segregation from other measures. Myatt & Langley (2003), UK Cardiothoracic unit (bed number not reported) 2 weeks closure for cleaning, preadmission screening, isolation in side room with decolonisation, nurse cohorting, hand gel, aprons and gloves, MRSA care pathway devised. Histogram of the 5 months before and after the changes was implemented showing number of pts with MRSA acquired on ward decreasing (maximum 9 per month, minimum 0, no denominators). Observational Poor reporting precluded assessment of potential for bias. Regression to the mean and chance effects likely. Unable to differentiate effects of isolation or nurse cohorting from several other changes in practice. 44

52 Study Setting Design Schelenz et al. (2005), UK Cardiac surgery unit (47 beds) Retrospective uncontrolled before and after study, two phases of 16 months ( ) Talon et al. (2003), France Dedicated orthopaedic cohort facility (20 beds) and three standard orthopaedic units (86 beds) Retrospective, observational, over one year ( ) Main interventions Enhanced targeted infection control programme including screening, isolation in single rooms (one patient could be nursed on the open ward during which time it would be closed to new admissions), decolonisation, nurse cohorting. Control programme implemented in the dedicated cohort facility including screening, isolation in separate rooms or in cohorts, decolonisation, gowns, gloves, hand washing. Patient outcomes Proportion of pts acquiring MRSA was 38/1,036 before and 14/921 after (P=0.03, Fisher s exact test). Statistical modelling predicted that without a dedicated cohort facility the hospital wide risk of acquiring MRSA would increase by 160% per year (from 14 to 23 cases). Assessment of evidence Intervention prompted by increased infection rates, regression to the mean likely. Study design vulnerable to information and reporting bias. Potential confounding factors not reported. Unable to differentiate effects of isolation or nurse cohorting from other infection control measures. Authors do not view the cohort facility as an isolation unit. Study design vulnerable to information bias in data collection and subsequent calculation of model parameters. 45

53 Assessment of the evidence The study of Fitzpatrick et al. (2000) did not report patient data on transmission of infection. One study conducted in the ICU setting showed that moving MRSA patients into single rooms or cohorted bays did not reduce transmission rates (Cepeda et al., 2005). Transmission rates were not reported but MRSA incidence varied between 15.5 and 28.2 acquisitions per 1,000 patient days at risk. MRSA prevalence rates on hospital admission were 7-9% for one hospital in the study and 30-34% for the second, with no explanation of the difference in prevalences being given. The reported handwashing compliance rate of 21% was very low. The rest of the studies reported reduced MRSA acquisition following introduction of the isolation facility. These reports were, in general, subject to similar limitations as those identified by Cooper et al. (2003) in that there was potential for other infection control measures to have accounted for changes in colonisation rates (Biant et al., 2004; Johnston et al., 2005; Myatt & Langley, 2003; Schelenz et al., 2005), there were other possible explanations for the findings such as regression to the mean (Biant et al., 2004; Schelenz et al., 2005) or the studies were vulnerable to sources of bias (Myatt & Langley, 2003; Schelenz et al., 2005; Talon et al., 2003). Three studies (Biant et al., 2004; Fitzpatrick et al., 2000; Talon et al., 2003) used isolation wards as defined by Cooper et al. (2003). Fitzpatrick et al. (2000) described three aims for their study: to determine whether patient cohorting reduced the likelihood of successful decolonisation, to evaluate the effect of staff decolonisation and to determine if successful environmental control of MRSA was possible. No attempt to measure change in MRSA incidence in the hospital as a whole was reported. The study of Talon et al. (2003) reported on the use of a dedicated orthopaedic cohort unit consisting of a 20-bed ward, a surgical block and a consultation facility to which all infected patients were admitted. Empirical data on the colonisation pressure (ratio of MRSA-positive patient days to total patient days), the numbers of MRSA-positive patients admitted and the number of patients acquiring MRSA were modelled to determine the effect of isolation. The authors estimated that the risk of acquiring MRSA in the absence of the dedicated cohort facility would increase by 160%. The premise of this paper was that by demonstrating that colonisation pressure affects MRSA prevalence, the policy of isolating MRSA-positive patients was an effective control method. In the paper of Biant et al. (2004), patients scheduled for elective joint replacement were screened for MRSA in the community prior to admission. MRSA-positive patients were excluded from a ring-fenced ward, and this reduced the MRSA infection rate from around 20% in the year before the intervention to no cases of infection in the year after. This study was subject to some limitations, in that the intervention was developed in response to a dramatic increase in the infection rate following restructuring of clinical services. A number of infection control policies were introduced and no indication of the effect of ring fencing on infection rate across the hospital was given. 46

54 5.4 Staff screening Current guidelines state that routine staff screening is not recommended but is indicated if transmission persists in a unit despite active control measures, or if epidemiological aspects of an outbreak are unusual or suggest persistent MRSA carriage by staff (Coia et al., 2006). The paucity of supporting evidence from research is reflected in the strength of this recommendation, which is categorised as suggested for implementation, supported by suggestive clinical or epidemiological studies or a theoretical rationale. A literature search identified three reviews that considered screening staff for MRSA, which are summarised in Table These reviews concluded that the evidence for routine staff screening in outbreak control was inconclusive (Jeanes & Rao, 1999; Wessex Institute for Health Research and Development, 1997) and that there was no evidence to support routine staff screening as an infection control measure in the endemic setting (Marshall et al., 2004). All three reviews failed to identify any controlled trials of the effects of staff screening on MRSA transmission. There appears to be little overlap in the studies included in the two reviews that addressed staff screening in outbreak control (Jeanes & Rao, 1999; Wessex Institute for Health Research and Development, 1997), although both undertook reasonably extensive searches for studies published up to The reason for this lack of overlap is not clear. One review in particular lacked transparency as to how studies were selected and gave no information concerning relevant material that was not included (Wessex Institute for Health Research and Development, 1997). The other review provided insufficient detail about studies used in relation to the stated inclusion criteria (Jeanes & Rao, 1999). Despite these concerns, both reviews reached similar conclusions. They found that most staff screening studies were case series and retrospective analyses, which were difficult to compare due to differences in carriage rates, extent of outbreak, location and patient type. Both highlighted the difficulty in distinguishing the effects of staff screening from other infection control measures implemented. The analysis of MRSA control by Marshall et al. (2004) was not a systematic review but addressed staff screening within the wider perspective of the control of endemic MRSA, based on a limited literature search and subjective study selection. A literature search was undertaken to identify primary studies considering the effect of staff screening to reduce MRSA transmission, either during outbreak control or in the endemic setting. Study selection focused on data published subsequent to the existing reviews. In order to reduce uncertainty as to whether colonised staff contributed to the initiation or spread of MRSA, studies were sought that used molecular typing to determine whether staff and patients carried the same MRSA strains. No controlled trials were identified and most studies did not assess the effect of screening on MRSA transmission (see table of excluded studies in Appendix 5, Table 13-5). Three studies not already included in the existing reviews met the inclusion criteria. 47

55 Table 5-11 Reviews of staff screening for MRSA Review Objective Search Inclusion criteria Included studies Authors conclusions Comments Jeanes & Rao (1999) Evaluate evidence for efficacy of routine screening and treating asymptomatic healthcare worker carriers in controlling MRSA outbreaks MEDLINE, CINAHL, Cochrane Library , English language, handsearch of two journals, Laboratory methods to detect asymptomatic colonisation among healthcare workers (all) where observed incidence greatly exceeds expected incidence (outbreak). No controlled trials found, 10 studies included, clinically heterogeneous, internal validity not assessed. There is little evidence to support routine screening of healthcare workers to identify carriers in an outbreak. Further research needed. Excluded studies and reasons given; sparse details of individual included studies; authors conclusion consistent with the evidence presented. Marshall et al. (2004) To determine whether evidence exists for the value of MRSA infection control measures that are widely recommended in the endemic setting including staff screening MEDLINE and references, no date. Inclusion criteria not reported. The authors stated that we have tried to include illustrative examples with evidence for and against certain measures ; data included from three studies: proportion of screened staff found to be carriers. Unless there is an epidemiological link between MRSA in a patient and staff member, there is no evidence to support routine screening of healthcare workers. Limited search and subjective inclusion. Wessex Institute for Health Research and Development (1997) Assess value of staff screening for MRSA carriage in outbreak control MEDLINE, HealthSTAR and Cochrane Library to 1997, Gears and NNR The authors applied modified Wilson Junger criteria for an effective screening programme because no controlled trials of screening were found. No study selection criteria reported. No controlled trials found; number of included studies not reported, included were examples of some key papers that are often quoted in support of screening ; internal validity not assessed. Evidence on the role of staff in transmission of MRSA is lacking, therefore, not possible to judge whether benefits of staff screening outweigh potential harms and costs. Further research needed. Decisions about study selection not transparent; insufficient details of included studies; no details of excluded studies. 48

56 Two epidemiological studies of staff screening in outbreak control were identified and are summarised in Table 5-12 (Berthelot et al., 2003; Wang et al., 2001). These studies screened all staff on the affected units at a single timepoint and used molecular typing of MRSA isolates to investigate the source of unusual outbreaks. In both studies, a healthcare worker with colonised dermatitis or eczema and nasal colonisation was identified as the likely source. Following treatment with mupirocin (Wang et al., 2001) or removal of the staff member from the affected unit (Berthelot et al., 2003), no new cases of colonisation or infection were observed among patients. These studies illustrate the potential value of staff screening to investigate small outbreaks confined to specialist units, where known risk factors help identify staff at higher risk of MRSA carriage and transmission. What remains unclear is the role of MRSA carriage among healthcare workers in general in initiating and maintaining outbreaks. As a result, extrapolating the data from one study to different situations is not straightforward. One study of staff screening in the endemic setting investigated the molecular similarity of MRSA isolates between hospital employees and patients (Eveillard et al., 2004). The study, conducted in France, found the highest prevalence of nasal carriage among ward staff (9.0%; 95% CI 6.7, 11.3), but the molecular identity between MRSA isolates from staff and patients on the same wards was only 25%. The highest prevalence of MRSA carriage among staff and patients was seen in the long-term care facility, where the identity of MRSA strains between staff and patients was the same in all cases. The same study assessed the duration of MRSA carriage by staff on two medical wards where 72 of 159 staff agreed to participate. Staff found to be MRSA positive had swabs taken every 3 weeks until MRSA negative. The prevalence of MRSA carriage was 14 of 72 staff (19.4%; 95% CI 10.3, 28.5) and length of service greater than 5 years was associated statistically with higher prevalence. The median time between the first positive swab and the first negative swab was 83 days (range days). Nasal mupirocin was prescribed for two staff members who were still carriers after 6 months and follow-up cultures after treatment were negative. The authors reported using molecular typing to compare MRSA strains between staff and patients in this part of the study, but did not report the data. The authors concluded that MRSA transmission between patients and staff is likely to depend on how often and for how long staff are exposed to MRSApositive patients and on the infection control measures used. The study investigated MRSA transmission from staff carriers to their households rather than transmission within the hospital setting. In summary, research evidence for the effectiveness of staff screening to control MRSA transmission is limited. Many argue that screening and decolonisation of healthcare workers is crucial to MRSA control, whilst acknowledging that this premise is not evidence based (Voss, 2004). Support for this argument comprises accumulated microbiological knowledge about S. aureus, empirical evidence from countries with low MRSA prevalence and intensive multicomponent infection control policies, and reports of outbreaks traced to colonised healthcare workers. The BSAC guidelines reflect current knowledge concerning routine staff screening, with regard to the recommendation that staff with skin lesions should be referred for screening and treatment if new MRSA colonised patients are found on a ward (categorised as strongly recommended for implementation) (Coia et al., 2006). 49

57 Table 5-12 Primary studies of staff screening for MRSA Study Aim Design Culture and typing Staff decolonisation Results Berthelot et al. (2003), France Investigate source of paediatric ICU outbreak Epidemiological study. Nose swabs from all staff in unit, weekly throat and stool screening of neonates, retrospective review of medical and microbiological records. 2 months later nose swabs taken again from staff identified as carriers plus skin swabs from one found to have eczema, Culture methods not reported, PFGE molecular typing. Nasal mupirocin and polyvidone iodine soap for 5 days. 2/42 staff identified as nasal carriers were decolonised and hygienic practice reinforced in the unit; follow-up cultures 2 months later when another patient became infected, both nasal cultures negative but one nurse had colonised eczema absent at first screening; no new cases following exclusion from the unit. Eveillard et al. (2004), France Assess prevalence and duration of nasal carriage among hospital staff in the absence of an outbreak, and MRSA transmission to households Part 1: Hospital-wide point prevalence surveys; ; Part 2: Survey and follow up of medical ward staff; Participation voluntary. Screening agar and disk diffusion pulsedfield gel electrophoresis (PFGE) molecular typing. Nasal mupirocin for 7 days when carriage exceeded 6 months. Part 1: Point prevalence 60/965=6.2% (95% CI 4.7, 7.7), highest among ward staff 52/577=9.0% (95% CI 6.7, 11.3). Part 2: 14/72 MRSA +ve=19.4% (95% CI 10.3, 28.5) Length of service 5 years 11/38 +ve=28.9% (no CI). MRSA identity between staff and patients varied from 25% to 100% depending on specialty. Wang et al. (2001), Taiwan Investigate source of postoperative surgical wound outbreak Nose swabs, and hands if dermatitis present, from ICU and surgical staff; epidemiological typing; Non-selective agar, Staphylase (Oxoid) test and disk diffusion, PFGE molecular typing. Nasal and dermatological mupirocin for one week. Screening of 33 staff identified one surgeon as a nasal carrier who also had colonised dermatitis, decolonisation follow-up cultures negative and no new cases observed. 50

58 5.5 Discussion Despite a very large literature concerning MRSA, few journal articles comprised planned experimental studies. The majority of studies were observational, following the introduction of new infection control policies and procedures or MRSA infection outbreaks, or reports of routine information collection. The weaknesses in the methodology and reporting of such studies have been addressed by the development of the ORION statement (guidelines for transparent reporting of Outbreak Reports and Intervention studies of Nosocomial infection) which was drafted to improve the quality of published literature to allow subsequent synthesis of results for systematic reviews and HTA (Stone et al., 2005). The literature searches had the primary purpose of identifying data to populate the economic model. As such, the context of MRSA screening in NHSScotland was important and specific issues, such as the relatively short average length of stay in Scottish hospitals, were taken into account. The prevalence of MRSA colonisation amongst the patient population admitted to acute inpatient care is not known in the absence of blanket screening of this group. No UK study published within the last 10 years cites such information, although four studies assessing colonisation in patients admitted to specialist units were identified. The results of a study recently published, which reported screening of patients admitted to a UK district general hospital with a prevalence of around 7% (Rao et al., 2007), were selected as appropriate for the economic model in the absence of more robust data. Clinical risk assessment of MRSA carriage has been investigated by a number of groups, however, no studies were identified which applied the criteria recommended in the BSAC/HIS/ICNA guidelines (Coia et al., 2006). The paper by Lucet et al. (2003) was, therefore, most useful in terms of assigning a measure of effectiveness of clinical clinical risk assessment. Further studies, considering the number of patients that would be correctly identified using a predefined set of clinical risk assessment criteria amongst the Scottish patient population, are required in order to verify the use of the Lucet et al. model. The transmission of MRSA is likely to be dependent on a number of factors including hospital and ward size, colonisation rate, the use of antibiotics and other infection control policies, and the ratio of staff to patients. It is difficult, although possible, to determine the transmission rate from empirical studies (Cooper & Lipsitch, 2004) and for the purposes of this HTA, it was decided that this method would be too complex and time consuming. The method described by Cooper et al. (2003) was, therefore, used where the transmission rate is assigned dependent on the equilibrium colonisation rate, using admission and discharge rates in the absence of any infection control measures. This approach is detailed in Section 6. Evidence considering the effectiveness of different management policies for patients colonised with MRSA was systematically reviewed by Cooper et al. (2004). The authors concluded that despite the methodological weaknesses of the studies reviewed, there was sufficient evidence to support the recommendations of current guidelines advocating the use of isolation facilities. In this report, we identified a number of studies published since Cooper et al., all with similar methodological limitations. None of the recent studies provided more robust evidence of the effectiveness of isolation. Many studies of the effectiveness of isolation have been conducted in ICU and it may be difficult to generalise from such units to other hospital wards and units. In particular, it has been observed that staffing ratios in ICU are associated with changes in rates of transmission of infection (Halwani et al., 2006; Dancer et al., 2006). Given the limitations of the evidence base, for the purposes of the economic model, it was assumed that following identification of colonised or infected patients, measures can be implemented to isolate them which would result in a reduction of the transmission rate of MRSA to zero. How this could be achieved practically in the hospital setting is unclear. A Cochrane review of antimicrobial treatments for MRSA decolonisation concluded that there was insufficient evidence to support use of topical or systemic antimicrobial therapy for eradicating nasal or extranasal MRSA (Loeb et al., 2003). In addition, this review found evidence of potentially serious side effects as a result of such treatments. Nevertheless, decolonisation of MRSApositive patients has been recommended in recent guidelines (Coia et al., 2006). Therefore, a search of decolonisation studies was conducted using the treatment regimen recommended by Coia et al. (2006). Although a large body of literature was identified, few studies provided data suitable for use in the economic model. In some studies, patients were treated with mupirocin in combination with oral antibiotics, in others both colonised and infected patients were included. For the purpose of finding estimates for the economic model, length of follow up was an important consideration as decolonisation during the hospital stay was required. Our assessment of the literature compares with a similar exercise (although with different objectives) carried out for the development of guidelines for the laboratory diagnosis and susceptibility testing of MRSA (Brown et al., 2005). The authors of these guidelines found only a small proportion of the published literature met the criteria of testing the efficacy of methods for screening of clinical specimens rather than pure cultures. Furthermore, many of the studies identified by Brown et al. did not provide sufficient data to compute both the sensitivity and specificity of the test method as required for our economic model. Subsequently, only a small number of the studies of the effectiveness of laboratory tests met the HTA inclusion criteria, as the majority of published studies compare the diagnostic accuracy of tests using bacterial isolates rather than patient swabs. There were shortcomings in the reports which did meet this criterion, as there is no agreed reference standard for MRSA testing. It is likely, with a greater than chance probability, that the culture method used as the reference standard and the method under test will misclassify patients as screening 51

59 positive or negative in the same way as has been described in screening tests for other infectious disease (Valenstein, 1990). This would distort the test performance and may confer an upward bias on sensitivity and specificity estimates. If the specificity of a test is lower than the estimate used in the economic model, beds in isolation facilities may be occupied unnecessarily, potentially excluding MRSA colonised patients. If the sensitivity of a test is lower than the estimate used in the economic model, the risk to noncolonised patients will be increased as false-negative MRSA patients may be managed in the open ward. Having high-quality data to populate any screening model is important, but currently there is a paucity of studies providing such data. Overall, there was considerable difficulty in identifying high-quality studies from which estimates for use in the economic model could be obtained. The lack of robust estimates will impact on the strength of any recommendations that emerge from the model outputs. The need for high-quality research studies and better reporting of research within the HAI field are well recognised. An examination of the evidence for MRSA screening of staff was initiated following a discussion with patient representatives (Section 7). Current guidelines recommend that patients with skin lesions are referred for screening and that all staff are screened in the event of unexplained persistent outbreaks (Coia et al., 2006; Scottish Infection Standards and Strategy Group, 2006). Literature on the effectiveness of staff screening was limited. Although intuitively it might be considered less effective to screen patients while not screening staff, there is no evidence to guide the frequency with which staff should be screened or the relative benefits of screening staff and allowing them to continue working until a positive test result is obtained. It is highly unlikely that staff resourcing levels would permit screening of staff and sending them home (the equivalent of patient isolation) whilst awaiting the test results. However, if prevalence rates of MRSA colonisation could be reduced to levels comparable to those seen in countries such as the Netherlands, regular screening of staff may be an important component in maintaining these low levels. Until the effects of colonisation pressure on staff caring for those with MRSA and the nature of transient colonisation are better understood, the current recommendations appear appropriate. 52

60 6 ECONOMIC EVALUATION AND MODELLING 6.1 Introduction The objectives of the evaluation were to: review existing literature on the cost effectiveness of screening patients for MRSA upon hospital entry extract relevant data and models from existing studies identify the care pathways followed by patients with MRSA colonisation and/or infection in Scotland develop an economic model applicable to the Scottish population and healthcare system, and to populate the model with cost and outcome data make due allowances for uncertainties in parameters and in the structure of the model, by conducting appropriate sensitivity analyses interpret the results, including the outcomes of sensitivity analyses, in the context of original HTA questions provide an analysis of the financial implications to NHSScotland if the recommended MRSA screening strategy is adopted. 6.2 Evidence sources Evidence was obtained from a variety of sources including published and grey literature, and information from manufacturers and clinical experts. Other sources of evidence used for the economic model included the results of the MRSA screening questionnaire survey (Section 8.2) and data supplied by the HTA Topic Group members, NHS personnel and other experts and National Services Scotland (Information and Statistics Division) Literature search Initial scoping searches were undertaken in January 2005 to identify economic evaluations relevant to the HTA. The NHS Economic Evaluation Database (NHS EED) and the Health Economics Evaluation Database (HEED) were searched. In addition, websites of the world s major health economics research centres were searched for relevant economic evaluations. Full searches to identify primary literature were undertaken in June 2005 and updated in January The search combined the concept MRSA with an inhouse economics search filter. Search results were restricted to papers published in English from 2000 onwards. The following bibliographic databases were searched using the OVID platform: MEDLINE EMBASE CINAHL MEDLINE in PROCESS. Literature was also identified using Zetoc (the British Library s table of contents alerts service) by citation searching on key papers and from scanning the bibliographies of retrieved items. A list of the sources searched and a copy of the strategy used to search the MEDLINE database is presented in Appendix 2. This strategy was adapted to search all other databases. A complete listing of all search strategies can be obtained by contacting NHS QIS Study exclusion criteria Criteria were applied to the economic literature to exclude: review articles not containing costs, outcomes or models antimicrobial resistance studies studies conducted in long-term care facilities mortality studies community-acquired MRSA studies studies on the cost effectiveness of different laboratory screening media studies showing the cost effectiveness of drug therapy studies published before 2000 studies not published in English. The literature search yielded 980 references. Full text citations were obtained for 97 and the remainder were excluded as irrelevant, on the basis of title and or abstract alone. A total of 29 articles were selected using the exclusion criteria. Seven studies were included in the final analysis (Bissett, 2005; Chaix et al., 1999; Cooper et al., 2004; Farr, 2004; Herr et al., 2003; Rubinovitch & Pittet, 2001; Wernitz et al., 2005), with the remainder being excluded on the basis of full article content Overview of studies The selected literature included three reviews and one systematic review considering the role of MRSA screening and the subsequent isolation of colonised patients in the prevention and control of infection in the hospital setting. All reviews noted methodological shortcomings in the evidence base, but concluded that active MRSA detection programmes and isolation of colonised or infected individuals should be implemented. Farr (2004) cited two studies indicating that patient isolation on admission and the removal from isolation following confirmation of negative-mrsa test results was cost effective. Bissett (2005) recommended universal MRSA screening on admission to hospital and subsequent isolation if indicated, particularly for patients considered at high risk of colonisation. This strategy was considered important to minimise MRSA transmission both within and between healthcare institutions. Rubinovitch & Pittet (2001) specifically considered the role of screening in institutions where MRSA is endemic. It was noted that elimination of the MRSA reservoir in this setting may not be feasible. However, early identification of colonised or infected individuals and prompt implementation of appropriate control measures could reduce transmission. The authors advocated the screening of high-risk patients on admission to hospital and noted cost effectiveness in the acute care MRSA endemic setting. 53

61 The evidence base did not generalise to Scotland in terms of costs and clinical data so a model was built. 6.3 Economic model Methodology The economic modelling methodology adopted is presented in Guidance for Manufacturers (Health Technology Board for Scotland, 2002). Best practice economic evaluation should comprehensively assess the changes in health states and associated cost changes that arise from the adoption of a technology, ideally from a full societal perspective (Drummond & McGuire, 2001). Adopting a societal perspective means that all changes in resource costs and outcomes should be included in the economic evaluation. However, there is considerable uncertainty surrounding the values of key parameters in the NHS model, particularly with respect to MRSA transmission and prevalence rates. Introducing societal costs increases uncertainty in the model and, therefore, in the value of its findings. As a result, this economic evaluation adopted an NHSScotland perspective to capture costs and benefits. The objective of the economic model was to inform recommendations to NHSScotland on the efficient use of MRSA screening options by comparing strategies, including the different combinations of clinical risk assessment and swab screening tests currently available in Scotland, and to identify the level of isolation facilities required to reduce the prevalence of MRSA. To this end, a form of cost-consequence analysis was carried out. Cooper et al. (2003) modelled a single hospital and community population pool to assess the impact of isolation policies in the management of MRSA transmission in the hospital setting. The basic structure of the Cooper et al. model was used and then significantly extended for the economic model used in this HTA. The model presented here is a stochastic compartment mass-action model which considered the impact of admission screening and isolation policies in the management of MRSA transmission in the hospital setting. To this extent, the model represented one tertiary referral hospital and a community population pool. Patients were admitted to hospital from the community pool and returned to the community pool when discharged from hospital. Further details on the structure of the model are listed in Section The patient population used in the model comprised adult inpatient acute admissions only. Day case and outpatients were excluded. The patient population was disaggregated in the model into two groups: patients under 65 years of age patients aged 65 years and over. Each age group had a different MRSA prevalence rate and the impact of this was assessed in the model. The time horizon used in the model was 5 years. It should be noted that this may not be sufficient to capture all the benefits resulting from MRSA screening, but the time frame was selected in recognition that modelling over longer periods increases the degree of uncertainty in the findings MRSA screening MRSA screening methods Screening to identify patients with MRSA can take the form of swab testing or clinical risk assessment. During swab testing, a swab sample is taken from the patient for laboratory analysis. Swab tests can indicate MRSA colonisation or infection but do not distinguish between the two. Patients with a positive swab test result may: be colonised and subsequently acquire a secondary infection during their hospital stay be colonised and remain without infection have an infection at the time of hospital entry be a false positive, where the laboratory test wrongly classifies the patient as having a positive MRSA status when their true status is negative. Clinical risk assessment estimates the patient s risk of being an MRSA carrier, by reviewing their medical history and assessing risk factors that would indicate that they were more likely to be a carrier. Such risk factors include: known previous MRSA carrier status hospital admission within last 12 months transfer from another healthcare facility. If a patient is identified as a likely carrier, a swab test can be performed to confirm their MRSA status. Laboratory tests The implication of using different types of laboratory test, in terms of costs and turnaround times, was assessed by the model. There are three stages in the MRSA laboratory testing process: 1. Culture of bacteria from swabs using different media, such as screening and chromogenic agar. 2. Identifying S. aureus using latex slide tests or tube coagulase tests. 3. Confirming that the S. aureus isolate is meticillin resistant using the disc diffusion method (observing reaction to the antibiotic) or protein binding tests. A separate disc diffusion test can be performed to assess the responsiveness of the isolate to different antibiotics. Traditional PCR can be used to analyse a pure bacterial isolate from culture. New real-time PCR systems can test for the presence of MRSA directly from swab samples. 54

62 Laboratory test usage The results of the NHS QIS questionnaire survey showed that several different combinations of laboratory tests were being used across Scotland: screening agar + latex + disc screening agar + latex + PB2B screening agar + tube coagulase + disc screening agar + tube coagulase + PB2B chromogenic agar + latex enrichment broth + natural agar + latex + disc enrichment broth + natural agar + latex + PB2B enrichment broth + natural agar + tube coagulase + disc enrichment broth + natural agar + tube coagulase + PB2B real-time PCR. The choice of testing regimens assessed in the economic evaluation were influenced by the availability of robust clinical studies assessing the sensitivity and specificity of culture media. Only studies that assessed the sensitivity and specificity of a test that sampled directly from the screening swab were included (Section 5.3.3). For the purposes of the economic evaluation, three different types of laboratory tests were evaluated in the model: chromogenic agar + slide latex + disc screening agar + slide latex + disc real-time PCR. Laboratory turnaround times The laboratory turnaround time was defined as the time taken to report a confirmative MRSA test result, and it represents the time between laboratory receipt of a swab sample and reporting the results to clinical staff. Different combinations of testing processes vary in their turnaround times, depending on incubation times and when test reactions are observed. Minimum and maximum turnaround times for each testing regimen were estimated as detailed in Table 6-1. Table 6-1 Laboratory test turnaround times Testing regimen Chromogenic agar + slide latex + disc Screening agar + slide latex + disc Real time PCR (identifying both meca and femb genes) Turnaround time (min-max) h h 3 24 h 55

63 Hospital setting MRSA screening setting Swab screening for MRSA can occur in two settings, either in preadmission clinics or on hospital admission. In Scotland, some hospitals operate preadmission clinics for certain elective procedures such as vascular and orthopaedic surgery. Swab screening for MRSA can take place at these clinics, with the benefit being that the patient s MRSA status is established by the time they enter hospital and they can be managed accordingly. The economic model assumed that all patients presenting for elective procedures attended a preadmission clinic and thus can be screened if required by the strategy. Patients identified as MRSA positive are admitted to an isolation room if available; otherwise they are placed on an open ward. Alternatively screening for MRSA can be performed on hospital entry, as for all unplanned emergency admissions. Screening on hospital entry means an inevitable time delay, while laboratories process swab specimens and relay the results back to healthcare staff. This time delay impacts on where the patient is placed within the hospital and, if the patient is colonised with MRSA, can be associated with a transmission risk to other patients. High and low MRSA risk specialty units units, with risk determined according to the likelihood of progressing from colonisation to infection. The information used to construct and populate the patient groups was based on the hospital specialties for acute and maternity care type classifications used in the Scottish Health Care Costs 2005 (Information and Statistics Division, 2005a). Hospital specialties were categorised into high and low risk as listed in Table 6-2. The categorisation of specialty units was based on Coia et al. (2006) and was validated by external experts. The following specialties from acute and maternity care were excluded: medical paediatrics surgical paediatrics obstetrics GP MRSA screening strategies The results of the questionnaire survey showed that MRSA screening practice varies considerably across Scotland and the data were used to inform the screening strategies used in the model. There were six possible screening options using a combination of clinical risk assessment or swab screening. Patients were grouped on the basis of whether they were being admitted into a high or a lowrisk hospital specialty unit. The six screening strategies modelled comprised: The model split the overall patient group into patients entering to high and low MRSA risk hospital specialty Strategy 1: No clinical risk assessment, no swab screening of any patient. Table 6-2 Categorisation of hospital specialities High risk Cardiac surgery Coronary care unit Ear, nose and throat Haematology High dependency unit Intensive care unit Medical oncology Nephrology (renal) Neurosurgery Ophthalmology Orthopaedics Plastic surgery and burns Special care baby unit Thoracic surgery Urology Vascular surgery Low risk Accident and emergency Cardiology Care of the elderly Communicable diseases Dermatology Gastroenterology General medicine General practice General surgery (excluding vascular) Gynaecology Medical other Neurology Obstetric specialist Oral surgery and medicine Radiotherapy Rehabilitation medicine Respiratory medicine Rheumatology Spinal paralysis 56

64 Strategy 2: Strategy 3: Strategy 4: Strategy 5: Strategy 6: No clinical risk assessment, swab screening of high-risk and low-risk specialty unit patients. No clinical risk assessment, swab screening of only high-risk specialty unit patients. Clinical risk assessment of high and low-risk specialty unit patients with subsequent swab screening of all likely carriers. Clinical risk assessment of all low-risk specialty unit patients with subsequent swab screening of all likely carriers, swab screening of all high-risk specialty unit patients. Clinical risk assessment of high and low-risk specialty unit patients. Patients are then placed in pre-emptive isolation based on result of clinical risk assessment. Regardless of clinical risk assessment result, all patients undergo swab screening. This strategy assesses the value of pre-emptive isolation. Table 6-3 summarises the six screening strategies that were included in the economic model. These strategies were modelled for the three different types of laboratory tests: screening agar chromogenic agar real-time PCR Patient management The purpose of managing patients known or likely to be colonised is to control the overall prevalence of MRSA within the hospital. Patient management aims to reduce the opportunities for patient to patient interaction and thus MRSA transmission. There are various options available for the management of patients colonised or infected with MRSA including: isolation in a single-bed isolation room, with dedicated nursing staff isolation on an isolation ward, with dedicated nursing staff placement in a side room, with or without dedicated nursing staff cohorting several patients in a geographically distinct location, eg six-bed bays with or without dedicated nursing staff placement on a standard ward with no physical segregation from other patients. The economic model simplified patient management options to include either being housed in a single-bed isolation room or in a bed on an open ward. A literature search found no evidence that patient cohorting was effective in reducing MRSA transmission. Thus cohorting was excluded as a patient management option in the model. Beds that were not located within a single-bed isolation room were assumed to be located within other spacing configurations within the ward, eg side rooms and multi-bed bays. Beds in other locations were considered as bed in open ward for modelling purposes. The key property of a single-bed isolation room is that it reduces the possibility of transmission to other patients as: staff are more likely to comply with infection control procedures between patients patient to patient interaction is physically restricted airborne spread between beds will be reduced. The efficiency of using a single-bed isolation room in reducing the transmission rate was varied in the model between 100% efficiency, ie isolation reduces the transmission rate to zero, to 50% efficiency. A colonised patient located on a bed in the open ward may transmit MRSA to other patients within the ward and this transmission rate was represented by a fixed value in the model. The transmission rate was defined as the number of transmissions per source per susceptible patient day Model structure The modelling software used in this project is Similie version 4.9 which allows diagrammatic development of a model (Simulistics Limited, UK). Table 6-3 MRSA screening strategies Strategy Clinical risk Clinical risk Swab screening- Swab screening - assessment - assessment - high-risk units low-risk units high-risk units low-risk units 1 X X X X 2 X X 3 X X X 4 X X 5 X X 6 57

65 Figure 6-1 outlines the structure of the hospital and community population pool. The hospital structure shown here is specific to a tertiary referral hospital. The number of wards will differ for a large general hospital. The population is disaggregated into two age groups: people 65 years of age and over and people under 65 years of age. Both age groups mix within the community and hospital except in the wards that are specifically occupied by over 65s only. The two different age groups have different MRSA prevalence rates (see Section ). There are two routes into the hospital, either via a preadmission clinic for all elective procedures or via the emergency admissions ward for all non-elective procedures. The emergency admission ward represents a larger ward where patients are treated whilst waiting for a bed in the speciality ward. Within the hospital, specialist wards are grouped into two categories, high and low-risk speciality wards. For the tertiary referral hospital, two out of the 15 high-risk speciality wards and three out of the 19 low-risk speciality wards were assumed to be occupied by over 65s only. The remaining wards were occupied by both age groups. The following assumptions were incorporated: Colonisation occurs only in hospital, ie there is no transmission within the community pool. Loss of positive colonisation status can occur spontaneously in the community or in the hospital, and does so at the same rate in both settings. Recently discharged hospital patients experience a high risk of readmission back into the hospital, which falls over time to a constant lower risk level. The risk of colonisation experienced by a noncolonised patient is proportional to the number of non-isolated carriers on the same ward. Transitions in MRSA colonisation status occur at a constant rate Patients identified by swab screening as MRSA colonised were placed in a single-bed isolation room, capacity permitting. If single-bed isolation rooms were full, then the patient was placed in a bed on the open ward. Patients in single-bed isolation rooms transmit to patients outwith the isolation room at a lower rate than patients on the open ward. Patients in high-risk specialty wards identified as colonised by swab screening received decolonisation treatment. Isolation rooms are used solely for MRSA-colonised patients. Figure 6-1 Hospital and community population structure 58

66 The patient care pathway A patient care pathway describes the patient journey through the healthcare system and forms the basis of the economic evaluation. A patient pathway for each patient management strategy is presented in Appendix 6. The patient care pathway, with respect to the emergency admissions ward and preadmission clinic within the hospital, can be described as follows: The community contains all prospective patients, both MRSA colonised and non-colonised. Patients have two levels of readmission rate into the hospital, with readmission rate being independent of whether or not they are MRSA colonised. Patients are admitted to hospital if a bed in the speciality ward is available. If the ward is full, the admission rate is equal to the discharge rate. Patients not admitted are returned to the high readmission rate pool. There are two routes of admission to the hospital, emergency admission ward or preadmission clinic. From here, patients move into the hospital speciality wards. Depending on the strategy being modelled, patients undergo swab screening and/or clinical risk assessment for MRSA colonisation at the preadmission clinic or emergency admissions ward. Clinical risk assessment was assumed to be instantaneous and to partition patients as likely to be carriers and not likely to be carriers. Likely carriers (or all patients in the absence of clinical risk assessment) undergo swab screening for MRSA colonisation. Swab testing comprised one sample per patient (nasal). Patients with an MRSA infection that was identified at hospital entry by means other than routine screening were excluded from the model. All patients attending preadmission clinics undergo MRSA screening before admission and their screening result is known on entry to the speciality ward. Those patients not attending a preadmission clinic are screened on admission to the emergency admissions ward and wait for the results of testing. The length of stay in the emergency admission ward is 2 days during which the swab test result is reported. During their stay in the emergency admissions ward, any patient with a positive swab screening (MRSApositive patient hereafter known as an identified patient ) is isolated where possible. If an identified patient cannot be isolated because of space constraints, they are placed on the open ward. Strategy 6 specifically allows the use of temporary isolation. On entry patients undergo clinical risk assessment and regardless of the results of clinical risk assessment, all patients undergo swab screening. Patients are placed into temporary isolation (capacity permitting) based on the results of clinical risk assessment and stay there until the swab result is known. Once patients are discharged from the emergency admission ward and moved to the speciality ward, identified patients are placed in isolation, capacity permitting. If an identified patient cannot be isolated because of space constraints they are placed on the open ward within the speciality ward. All identified patients within high-risk specialty wards and identified patients in the emergency admissions ward who will move into the high-risk speciality wards undergo one course of decolonisation treatment, with subsequent laboratory tests to confirm whether decolonisation treatment was successful or not. Identified patients in low-risk specialty units do not receive decolonisation treatment. Identified patients placed in the open ward also undergo one course of decolonisation treatment. Isolated patients remain in isolation until discharge, with no change in placement being implemented because of loss of positive colonisation status. Similarly, patients undergo routine MRSA screening on admission to hospital but are not retested for colonisation whilst in hospital. MRSA colonised patients become infected at a fixed rate and, where possible, infected patients are isolated immediately. The only consequence of infection included in the model was increased length of stay. Mortality, morbidity and antibiotic use following MRSA infection were not considered. Patients with infections are detected immediately by means of clinical assessment. Further laboratory tests are not included in the model. MRSA colonised and non-colonised patients are discharged back into the community at a fixed rate, independent of colonisation or isolation status. The size of the community pool remains constant. Cost attributable to MRSA false-positive patients were identical to the cost attributable to true-positive patients. The following assumptions applied at the ward level: Transmission only occurred within a ward and not between wards. Admissions and discharges occurred from and to a common community pool, thus by default the same admission MRSA colonisation prevalence occurred in patients admitted to all wards. Clinical risk assessment, MRSA screening, isolation and decolonisation policies may differ between ward types (ie wards in high or low-risk specialty units) depending on the strategy modelled. Other than grouping according to high or low-risk specialty unit, no further distinction was made between ward types. Further information on the structure and functioning of the economic model is available on request. 59

67 Hospital setting Two types of hospital were modelled: tertiary referral hospitals large general hospitals. Hospital classifications for acute care providers, listed in the Scottish Health Service Costs 2005 (Information and Statistics Division, 2005a), were used to populate the two hospital groups. Hospital profiles for the two groups were used to calculate the dimensions of the tertiary referral hospitals and large general hospitals in terms of the total number of beds per hospital and the proportion of beds in high and low-risk specialty units. Table 6-4 shows the categories, definitions and number of hospitals used in defining the hospital setting for the model. A full list of the Scottish hospitals included in the two categories is provided in Appendix 6, Table 13-6 and Table Other hospital types such as general hospitals with less than 250 beds, sick children s hospitals and community hospitals were excluded from the analysis. Size of hospital The total number of beds in each type of hospital was calculated by averaging the number of staffed beds for each hospital type in Scotland: A tertiary referral hospital comprised an average of 840 beds. A large general hospital comprised an average of 480 beds. The proportion of beds that fell into the specialities listed in Table 6-2 was determined. There were a number of beds for non-acute care types (eg geriatric, psychiatry), which were classed as other and were excluded from the model. Table 6-5 shows the percentage of beds in high MRSA risk, low MRSA risk and other specialty units. The ratio of high-risk to low-risk beds was used in the model. Table 6-4 Scottish hospital categories Scottish health service costs (Information and Statistics Division, 2005a) Hospital in model Hospital type Definition Number of hospitals in Scotland Tertiary referral hospital A1 Major teaching hospital covering a full range of services and with special units 6 Large general hospital A2 Large general hospitals with some teaching units, usually over 250 staffed beds 19 Table 6-5 Percentage of bed in high, low-risk and other specialties Tertiary referral hospital Large general hospital (840 beds) (480 beds) High-risk specialties 43% (359 beds) 38% (182 beds) Low-risk specialties 56% (473 beds) 50% (240 beds) Other specialties 1% (9 beds) 12% (57 beds) High and low-risk specialties 99% (832 beds) 89% (422 beds) 60

68 Ward structure The number of beds per ward in Scottish hospitals varies from 18 to 30 and the model assumed 25 beds per ward (C Kilpatrick, Nurse Consultant Infection Control, HPS and Dugald Baird, Microbiologist [Retired]. Personal communication). Within each ward, a proportion of beds are in single-bed isolation rooms and the remainder are on the open ward. The number of single-bed isolation rooms was varied within the model to assess their impact on hospital MRSA prevalence levels. The assumed configuration of each hospital is detailed in Table 6-6 and Table 6-7, with bed numbers rounded where appropriate to fit a ward size of 25 beds. Hospital capacity and community population pool For the purposes of modelling, the hospital was assumed to be operating at full capacity, ie in the absence of MRSA infection, the discharge and admission rates were assumed to balance. The community population pool was adjusted to permit this balance. Of the total population, approximately 20% was aged 65 years and over. The transmission rate was fixed to ensure an equilibrium hospital MRSA prevalence of 3.6% for the under 65s and 14.5% for the over 65s (average 7.1%) in the absence of screening or isolation Model inputs Model costs This section explains the resources required, sources of data and the total cost of the following procedures: clinical risk assessment patient swabbing laboratory tests on swab samples decolonisation of patients contact precautions cost of hospital inpatient stay in isolation rooms cost of hospital inpatient stay in beds other than isolation rooms. The total costs to the health service provider were estimated and include where appropriate staffing costs, consumables and hospital overheads. The base year for all costs was March Costs are presented for a tertiary referral hospital and a large general hospital. Staff costs For all procedures, staffing resources at each staff grade were defined after consulting members of the HTA topic group and other experts. A questionnaire (resource use questionnaire), sent to infection control staff, was used to identify staff resources for clinical risk assessment and taking patient swabs. Staff salary costs were taken from 2005/2006 NHSScotland pay scales, and employer costs for superannuation and national insurance were added. Details of the distribution of staff across different grades were obtained from the ISD workforce statistics data set and were used to calculate the weighted average costs. Where cost per minute per staff was used, data on the number of working weeks per year and working hours per week were taken from PSSRU (Curtis & Netten, 2005) and included allowances for annual leave, training and sickness absence. Overheads Hospital overheads were applied to direct costs before deriving the final cost for each procedure. Data used to calculate overheads were taken from the inpatient data sets in Scottish Health Service Costs 2005 (Information and Statistics Division, 2005a). The overhead rate was calculated as the proportion of allocated cost to direct cost. Allocated costs included non-direct expenditure such as rent and rates, building maintenance, utilities, cleaning and transport. Overheads for non-laboratory procedures were calculated using the inpatient dataset. The weighted average overhead rate was calculated by weighting the hospital specific rates by hospital activity. ISD hospital classification data were used to calculate overhead rates for tertiary referral and large general hospitals. Thus separate Table 6-6 Tertiary referral hospital configuration Number of wards Total beds High-risk units Low-risk units All hospital Table 6-7 Large general hospital configuration Number of wards Total beds High-risk units Low-risk units All hospital

69 overhead rates were calculated for tertiary referral hospitals and large general hospitals and for the high and low MRSA risk specialty unit groups. Overheads for laboratory procedures were calculated from the Scottish Financial Returns 27.1, Scottish Health Service Costs 2005 (Information and Statistics Division, 2005a). The weighted average overhead rate was calculated by weighting the laboratory specific rates by the number of specimens processed. The resulting overhead rates are listed in Appendix 6, Table The cost of consumables was included where appropriate, including value added tax (VAT) payable by the healthcare provider. The cost for clinical risk assessment was based on a staff nurse (grades D G) spending approximately 10 minutes completing the assessment. The total costs are listed in Table 6-8. Full cost breakdowns with sources of information are provided in Appendix 6, Table The cost of swab sampling from patients listed on staff nurses (grades D G) spending approximately 5 minutes providing information to the patient, taking three swab samples and completing related administration such as labelling samples and sending them to the laboratory. The consumables included in the cost were three sterile transport swabs. The total cost is listed in Table 6-9. Full cost breakdowns, with sources of information are provided in Appendix 6, Table Laboratory tests The input cost per patient included laboratory tests for one swab samples. For the purposes of modelling, swabs were assumed to be analysed using the following laboratory methods: screening agar and + slide latex + disc chromogenic agar and + slide latex + disc real-time PCR. Screening agar and chromogenic agar testing Figure 13-7 in Appendix 6 outlines the processes associated with screening agar and chromogenic agar testing regimens. Staff resources required to process swab samples were estimated by visiting a number of diagnostic laboratories and observing the testing process. Staffing resources were estimated using the assumption that 30 samples would be processed in one batch, to take account of economies of scale. The final estimates of staff resources were validated by Dr Giles Edwards (Dr G Edwards, Deputy Director, MRSA Reference Laboratory, Scotland. Personal communication). Staffing resources are listed in Appendix 6, Table Laboratory consumables were estimated following advice from laboratory staff and are listed in Appendix 6, Tables to The costs for consumables were supplied by Scottish Health Care Supplies (Mr G. Cooke, Commodity Manager, Edinburgh. Personal communication) and reflect the National NHSScotland Purchasing Contract for laboratory supplies. Samples for which definitive identification is not achieved by the diagnostic laboratory are sent to the MRSA Reference Laboratory for confirmation. The MRSA Reference Laboratory processes the samples using traditional PCR methods. The cost of processing these samples was estimated in the total laboratory cost per sample. Consumables, staff resource estimates and the number of samples processed were provided by the MRSA Reference Laboratory (Dr D Morrison, Principal Clinical Scientist, Scottish MRSA Reference Laboratory. Personal communication). It was assumed that pure isolates for all MRSA-positive samples are sent to the MRSA Reference Laboratory. The costs to the diagnostic laboratory of preparing agar slopes and completing paperwork were included in the total cost. However, the cost to the MRSA Reference Laboratory for processing these samples was not included, as this forms part of their surveillance and research activity. Laboratory overheads, derived from Scottish Financial Returns 27.1, Scottish Health Care Costs 2005 (Information and Statistics Division, 2005a) and detailed in Appendix 6, Table 13-8, were applied before defining the final cost per sample. There are different stages at which a sample can be confirmed as negative (ie to confirm that the patient is not colonised or infected with MRSA), as outlined in the flow diagram in Appendix 6, Figure When Table 6-8 Costs associated with clinical risk assessment Procedure Tertiary general hospital Large general hospital Clinical risk assessment Table 6-9 Costs associated with taking patient swab samples Procedure Tertiary general hospital Large general hospital Swabbing patients

70 calculating laboratory costs, assumptions were made about the proportion of samples confirmed as MRSA negative at each stage, as detailed in Table Costs of positive and negative laboratory tests, per sample, are detailed in Table Real-time PCR Real-time PCR is not used routinely by NHSScotland for analysis of MRSA samples. Real-time PCR (SmartCylcer system with IDI MRSA Test Kit) is only licensed to analyse nasal swabs, therefore the economic evaluation assumed that only one sample per patient was analysed. The cost estimates presented here are for the SmartCylcer system using the IDI MRSA Test Kit. Each PCR analyser can process 14 samples plus two controls and a maximum of six analysers can be linked. It takes approximately one hour 20 minutes of laboratory bench time to prepare a set of 14 samples and approximately one hour to run one set of samples on the analyser. Economies of scale mean that preparing 28 samples at once takes approximately one hour 45 minutes. Prepared samples can then be stored for up to 6 hours on ice. The number of samples a laboratory can process during an eight-hour period is constrained by the number of analysers and the number of staff available to prepare samples. Estimates of staff time required to process each sample were provided by a representative from the distributor (Ms F McSharry, Product Specialist, Instrumentation Laboratory, Warrington. Personal communication). Estimates of consumables required were provided by the manufacturer (Mr Ian Hill, Sales Manager, GeneOhm Sciences, Sheffield. Personal communication). Consumables included the testing kit and cost of servicing the equipment. Laboratory overheads were included in the final cost. The annuity value of the capital cost of one unit of equipment of approximately 22,000 and equipment life of 7 years, were included in the final cost. The cost per swab sample for a real-time PCR test was estimated as and was the same for negative and positive samples. A breakdown of the costs associated with real-time PCR analysis of samples is provided in Appendix 6, Table Contact precautions When patients were placed in a single-bed isolation room, it was assumed that all healthcare staff would implement standard infection control polices and additionally adhere to contact precautions as described by Siegel et al. (2007). Contact precautions consisted of antiseptic hand washing when exiting the patient s room and wearing a plastic apron and gloves on entry into the room. Contact precautions were assumed to be additional to the normal level of care provided, and thus the cost was additional to the average daily cost of hospital stay. Three studies estimated the additional time required by healthcare staff to carry out contact precautions. Chaix et al. (1999) estimated from an observational time in motion study that an additional 20 minutes per patient per day were required by nursing staff to ensure compliance with isolation contact precautions. The additional time required by physicians was estimated at 10 minutes per patient per day. Herr et al. (2003) conducted a retrospective study to identify the additional cost of hygiene measures used with MRSA carriers in a German hospital. They estimated that on average there were 12 routine nurse visits into a patient s room and 3 minutes were required to don protective clothing. Wernitz et al. (2005) conducted a study analysing the cost of a selective screening programme for MRSA carriers in a German hospital. All patients were screened on entry and then entered prophylactic contact isolation for 2 days, whilst awaiting laboratory test results. They reported that the mean frequency of entry into the contact isolation room was 9.3 per day and estimated that an Table 6-10 Proportions assumed for negative sample confirmation Percentage of samples confirmed as negative Stage 1 Stage 2 Stage 3 Screening agar Chromogenic agar Table 6-11 Cost per sample Testing regimen Result Cost per sample Screening agar + slide latex + disc Positive 8.16 Screening agar + slide latex + disc Negative 5.06 Chromogenic agar + slide latex + disc Positive 7.45 Chromogenic agar + slide latex + disc Negative 4.34 All stages of test not required for samples found to be negative resulting in lower average cost. 63

71 additional 3 minutes was required to enter and leave the isolation room whilst complying with isolation precautions. This economic model assumed 12 patient contacts per day by healthcare staff, with 3 minutes per patient contact was needed to ensure compliance with contact precautions. For each patient contact, the consumables included in the cost analysis were one pair of gloves and one plastic apron. The total daily cost, including overheads, is listed in Table Full cost breakdowns, including information sources, are provided in Appendix 6, Table Costs associated with inpatient stay The model had two spatial options for bed management, comprising single beds in isolation rooms and beds located elsewhere in the ward. The incremental cost of providing and servicing the additional space associated with single beds in isolation rooms compared to open wards was calculated. The estimate assumed the only costs for an inpatient stay that varied with space use was the allocated costs (eg cleaning, heating, rent, rates and depreciation). Other costs such as medical and nursing staff costs were assumed to vary directly with patient numbers. Using guidance on beds and space use across Scotland from the Scottish Health Planning Note 04 (Scottish Executive Health Department, 2000), it was estimated that the average floor area for a bed in an open ward was 15.7 square metres. This included areas for shared day space and nurse workstations. Using the same source on average a single room required 21 square meters, an additional 33% space. The total allocated costs were apportioned in proportion 0.57:0.43 to capture this different floor space use. The total daily additional cost for an isolation room is listed in Table Length of inpatient stay Length of stay in hospital is a risk factor for acquiring MRSA infection. The longer a patient stays in hospital the more likely they are to be exposed to an MRSA-colonised patient or other factor that promotes transmission, potentially resulting in their colonisation and infection. ISD data on admissions and bed days occupied by under 65s and over 65s was used to calculate the mean length of stay for both age groups. Separate lengths of stay were used for tertiary referral and large general hospital types and for high and low-risk specialty risk units. A weighted average approach was used, where the weights used comprised the specialty activity values. The mean length of stay was used for patients who did not have an infection, regardless of whether they were colonised or not. Patients who develop an infection require an increased length of hospital stay. Many studies assessing additional length of stay focus on a specific patient group, eg ICU patients, or patients with a specific infection such as bacteraemia. As the economic model considered all MRSA infections, it was considered inappropriate to generalise by using data from disease specific studies. Plowman et al. (1999) carried out a cost of illness study of HAIs at an English district general hospital. They considered the additional number of days a patient stayed in hospital as a result of contracting the HAI. The different conditions assessed by Plowman comprised: urinary tract infections lower respiratory tract infections surgical wound infections bloodstream infections skin infections other infections multiple infections. Table 6-12 Costs associated with contact precautions Procedure Tertiary referral hospital Large general hospital - High risk Contact precautions Table 6-13 Additional daily costs associated with single rooms Procedure Tertiary general hospital Large general hospital Isolation room

72 Plowman developed a model to estimate additional length of stay, controlling for cofounders such as age and sex. The model showed that for one or more HAI at any site, the length of stay for an infected patient was 2.5 times that for an uninfected patient. It was noted that 37% of the HAIs observed in the study were grampositive infections. This figure from the Plowman et al. (1999) analysis was applied to the length of stay used for uninfected patients in the current model, to derive a value for MRSA infected patients. Length of stay values used in the model are detailed in Table Decolonisation treatment The model assumed that all patients in high-risk specialty units who were identified as MRSA colonised received decolonisation treatment. It was assumed that each patient received decolonisation treatment for nose, throat and skin carriage. The resource use questionnaire was used to identify staff resources for administering decolonisation treatment. Nursing staff (grades D G) administered decolonisation treatment and treatment dosage, duration and costs were estimated from British National Formulary (BNF) (British Medical Association, 2006). Daily staff resources and the decolonisation treatments administered are summarised in Table The total cost was assumed to be for each course of treatment consisting of five applications at a cost of 9.48 each Other model parameters The following parameters were calculated from linked patient data for the year ended March 2005, supplied by ISD (Mr D Clark, Mr S Oduro, Healthcare Information Group, ISD, Edinburgh. Personal communication): high risk of hospital admission rate for each age group proportion of patients attending preadmission clinics for each age group the proportion of patients entering high and low MRSA risk specialty units on admission to hospital. These and other clinical effectiveness parameters used in the model are summarised in Table 6-16 to Table Table 6-16 lists the values that are common to both tertiary referral and large general hospital settings. Table 6-17 and Table 6-18 list values specific to individual hospital settings. Values were applicable to both age groups unless specified. Table 6-14 Estimates of length of patient stay Tertiary referral Tertiary referral Large general Large general Procedure hospital hospital hospital hospital high risk low risk high risk low risk (days) (days) (days) (days) Under 65s Length of stay per uninfected patient Under 65s Length of stay per infected patient Over 65s Length of stay per uninfected patient Over 65s Length of stay per infected patient Table 6-15 Decolonisation resource requirement Daily staff time spent Location Treatment administering treatment (minutes per patient) Skin Hibiscrub/Aquasept 10 Throat Chlorhexidine 5 Nose Bactroban 5 65

73 Table 6-16 Parameter values common to both tertiary referral and large general hospital settings Parameter Value Source Readmission rate Low-risk Cooper et al. (2003) colonised patients (daily) (95% CI to ) From inpatient records for year ended 31 March 2005, (Mr D Clark, Readmission rate High-risk (Under 65s) Mr S Oduro, Healthcare Information colonised patients (daily) (Over 65s) Group, ISD, Edinburgh. Personal communication) Readmission rate Low-risk Cooper et al. (2003) non-colonised patients (daily) (95% CI to ) From inpatient records for year ended 31 March 2005, (Mr D Clark, Readmission rate High-risk, (Under 65s) Mr S Oduro, Healthcare Information non-colonised patients (daily) (Over 65s) Group, ISD, Edinburgh. Personal communication) Sensitivity of clinical risk assessment Lucet et al. (2003) False-positive rate for clinical 0.64 risk assessment Lucet et al. (2003) Screening Agar Sensitivity of swab tests Chromogenic agar 0.98 See Table 5-7 Real-time PCR 0.96 Agar False-positive rates of swab tests Chromogenic agar See Table 5-7 Real-time PCR 0.05 Agar 48 h Turnaround time for swab Chromogenic agar 24 h See Table 5-7 test results (hours) Real-time PCR 24 h Rate of spontaneous loss of positive MRSA colonisation status in Cooper et al. (2003) community (daily rate) Rate of loss of MRSA colonisation 0.1 following decolonisation (daily rate) Rohr et al. (2003) Detection rate for infected patients (daily rate) 1 Assumption From inpatient records for year ended 31 March 2005, (Mr D Clark, Proportion of patients attending 61% (Under 65s) Mr S Oduro, Healthcare Information preadmission clinics 57% (Over 65s) Group, ISD, Edinburgh. Personal communication) From inpatient records for year ended Proportion of patients not attending 39% (Under 65s) 31 March 2005, (Mr D Clark, preadmission clinic and going to 43% (Over 65s) Mr S Oduro, Healthcare Information emergency admissions ward Group, ISD, Edinburgh. Personal communication) From inpatient records for year ended 31 March 2005, (Mr D Clark, Percentage of patients going to 30% (Under 65s) Mr S Oduro, Healthcare Information high-risk speciality ward 33% (Over 65s) Group, ISD, Edinburgh. Personal communication) From inpatient records for year ended 31 Mar 2005, (Mr D Clark, Percentage of patients going to 70% (Under 65s) Mr S Oduro, Healthcare Information low-risk speciality ward 67% (Over 65s) Group, ISD, Edinburgh. Personal communication) 66

74 Parameter Value Source 3.6% (Under 65s) Hospital MRSA prevalence at start G Rao, Lewisham District General 14.5% (Over 65s) of the model Hospital, Personal Communication 7.1% Overall Screening agar 8.16 Cost of laboratory tests if result is Table 6-11 and Table Chromogenic agar 7.45 positive (per patient) (Appendix 6) Real-time PCR Screening agar 5.06 Cost of laboratory tests if result Table 6-11 and Table Chromogenic agar 4.35 is negative (per patient) (Appendix 6) Real-time PCR Decolonisation treatment per session one negative laboratory test if successful (per patient) Section Decolonisation treatment if one positive laboratory test unsuccessful (per patient) Section Table 6-17 Parameter values specific to tertiary referral hospitals Parameter Value for Value for high-risk unit low-risk unit Source Hospital discharge rate for infected 0.10 (Under 65s) 0.11 (Under 65s) Computed in model using patients (daily) (Over 65s) 0.04 (Over 65s) length of stay data Hospital discharge rate for colonised 0.26 (Under 65s) 0.29 (Under 65s) Computed in model using and non-colonised patients (daily) 0.09 (Over 65s) 0.1 (Over 65s) length of stay data Rate at which colonised patients become infected during hospital Coello et al. (1997) stay (patients per day) Transmission rate, for assumed 7.1% hospital prevalence (daily) Computed in model Number of wards Section Number of wards occupied by over 65s only 2 3 Section Number of beds per ward Section Number of isolation single-bed rooms in ward 3 3 Section Beds within each mixed age ward 17 (under 65s) 18 (under 65s) that are occupied by age group 9 (over 65s) 8 (over 65s) Section Length of stay per patient if patient 3.6 (Under 65s) 3.3 (Under 65s) is not infected (ie colonised) and 8.8 (Over 65s) 8.3 (Over 65s) non-colonised patients (days) Section Length of stay if patient is 9.0 (Under 65s) 8.2 (Under 65s) infected (days) 22.0 (Over 65s) 20.1 (Over 65s) Section Additional cost per night of hospital stay if patient is in isolation single Section bed room Swabbing (per patient) Section Clinical risk assessment (per patient) Section

75 Table 6-18 Parameter values specific to large general hospitals Parameter Value for Value for high-risk unit low-risk unit Source Hospital discharge rate for infected Computed in model using patients (daily) length of stay data Hospital discharge rate for colonised Computed in model using and non-colonised patients (daily) length of stay data Rate at which colonised patients become infected during hospital Coello et al. (1997) stay (patients per day) Transmission rate, for assumed 7.1% hospital prevalence (daily) Computed in model Number of wards 4 10 Section Number of beds per ward Section Number of isolation single bed rooms in ward 3 3 Section Number of beds in open ward Section Length of stay per patient if patient is 3.0 (Under 65s) 2.7 (Under 65s) not infected ie colonised and 8.6 (Over 65s) 8.4 (Over 65s) non-colonised patients (days) Section Length of stay if patient is infected 7.5 (Under 65s) 6.8 (Under 65s) (days) 21.5 (Over 65s) 21.0 (Over 65s) Section Additional cost per night of hospital stay if patient is in isolation single Section bed room Swabbing (per patient) Section Clinical risk assessment (per patient) Section

76 6.4 Model limitations As noted by Cooper et al. (2003), this type of economic modelling has significant limitations, which imply that it should be regarded as a qualitative descriptive tool rather than being used to make specific numerical predictions. The specific limitations include: The model only considers two heterogeneous patient group (under and over 65 years old).the likely existence of other heterogeneous patient groups is ignored. For example, some patients such as care home residents may have both a higher prevalence and a higher hospital readmission rate than the patient groups used in the model. Many inputs are subject to substantial uncertainty, for example: The prevalence of MRSA colonisation in the Scottish population is not well characterised. The literature on the sensitivity and specificity of laboratory MRSA assays comprise studies with no reference standard. MRSA decolonisation rates are not accurately known. The model does not consider variations in the prevalence of MRSA between wards. It is plausible that isolation facilities could become overloaded in some wards, even if overall provision is adequate. The model ignores the impact that MRSA containment policies in other hospitals and care facilities may have on the overall situation. The model assumes that at the time of introducing a screening strategy the hospital is at full occupancy but three isolation beds per ward are available for MRSAcolonised patients. 6.5 Results Summary of strategies Table 6-19 summarises the six screening strategies that were modelled for the three different types of laboratory tests: agar; chromogenic agar and real-time PCR. Table 6-20 describes the combinations of laboratory tests modelled for use at preadmission Results for a tertiary referral hospital The following results are based on 1,000 simulations of the model being run for each strategy/test combination following a burn-in period to allow the model to reach stability. Table 6-21 allows comparison of the predicted results of using combinations of three of the screening strategies and the four combinations of laboratory test at Years 1 and 5. The full results can be found in Appendix 8. The prevalence after introduction of screening using chromogenic agar for all laboratory tests is shown in Figure 6-2. This figure shows that all strategies are effective in reducing prevalence (from a rate of 7.1% prior to the introduction of screening), and that Strategies 2 and 6 achieve the greatest reductions in prevalence at the end of the first year. Table to Table in Appendix 8 details of the prevalence rates at Years 1, 3 and 5 for each of the screening strategies and each of the laboratory test combinations. Figure 6-3 shows the output from the Simile programme comparing Strategies 2, 3 and 5, illustrating a similar rapid fall in prevalence from the initial 7.1% for Strategies 2 and 5, but the slower predicted fall in prevalence when using Strategy 3. Table 6-19 MRSA screening stategies Clinical risk Clinical risk Strategy assessment assessment High-risk units Low-risk units Swab screening Swab screening High-risk units Low-risk units 1 X X X X 2 X X 3 X X X 4 X X 5 X X 6 Table 6-20 Laboratory test combinations Combination abbreviation Preadmission screening On-admission screening AgAg Agar Agar AgChrom Agar Chromogenic agar ChromChrom Chromogenic agar Chromogenic agar ChromPCR Chromogenic agar PCR 69

77 Table 6-21 Summary of prevalence and costs of combination of screening strategies and laboratory tests Laboratory Year 1 Year 5 test Prevalence Total cost Prevalence Total cost Strategy 2 AgarAgar ,461, ,077,939 AgChrom ,069, ,954,492 ChromChrom ,223, ,471,804 ChromPCR ,029, ,621,793 Strategy 3 AgarAgar , ,654,922 AgChrom , ,985,146 ChromChrom , ,926,629 ChromPCR , ,240,656 Strategy 4 AgarAgar ,217, ,750,077 AgChrom ,944, ,259,886 ChromChrom ,388, ,315,874 ChromPCR ,909, ,994,351 Strategy 5 AgarAgar ,684, ,193,649 AgChrom ,993, ,547,488 ChromChrom ,306, ,903,498 ChromPCR ,915, ,047,071 Strategy 6 AgarAgar ,684, ,193,649 AgChrom ,291, ,068,774 ChromChrom ,441, ,584,988 ChromPCR ,250, ,737,900 70

78 Figure 6-2 Mean prevalence (%) using chromogenic agar for all laboratory tests Figure 6-3 Change in mean prevalence (%) over time 71

79 As can be observed from Figure 6-4, using chromogenic agar for all laboratory tests is more effective in reducing prevalence at one year than the other test combinations. This pattern is reflected in the prevalence of MRSA at other time points after introduction of screening as shown in Appendix 7. It should also be noted that very low inhospital prevalences (below 1% for all strategy/test combinations) are predicted by the model if the assumptions used are accurate and the strategies are implemented to ensure complete screening of populations as described in Section 6.4. Figure 6-4 Mean prevalence (%) for Strategy 2 at Year 1 72

80 The total cumulative costs of screening are illustrated for Strategy 2 in Figure 6-5. Those strategies using chromogenic agar for all tests are considerably less expensive than the other laboratory test combinations. The total cumulative costs of the five strategies are compared in Figure 6-6. It can be seen that Strategy 3 is considerably less expensive than the other strategies over the time period modelled and at the five-year timepoint will be almost three times lower than any of the other strategies. Strategy 2 is the next least expensive, but at after one year will incur costs of 2.5 times that of Strategy 3, and at Year 5 this rises to 2.8 times that of Strategy 3. Figure 6-5 Mean costs of screening using Strategy 2 at one year Figure 6-6 Cumulative mean costs using chromogenic agar for all laboratory tests 73

81 The cost breakdown of Strategies 2, 3 and 4 is illustrated in Figure 6-7 for each of the five years following implementation. As would be expected, the laboratory and screening costs (ie staff costs for assessing carriage of MRSA and taking swab samples) remain fairly stable across the five-year period, with the cost of isolating and decolonising patients reducing as the prevalence of MRSA in the hospital decreases. The comparison of the elements of the three strategies in Figure 6-7 shows that the staff costs for screening (and therefore requirement for staff time) are much greater for Strategy 5 than the other strategies as all patients being admitted to a low-risk specialty ward will require an assessment of the likelihood of them being colonised with MRSA. In comparison, laboratory costs in Strategy 2 are the greatest proportion of total costs. Table 6-22 gives the cumulative number of infections predicted by the model over the five-year period in the absence of any screening strategy and for Strategies 2, 3 and 5. The mean number of infections predicted is also illustrated in Figure 6-8 for each of these three screening strategies. It can be seen that even for the least effective strategy (ie Strategy 3), at the end of the first year on average 236 MRSA infections would be avoided, and over 5 years, a total of 1,517 infections would not be incurred. For the most effective strategy where all admitted patients would be screened (Strategy 2), 1,592 infections would be avoided over 5 years. Using Strategy 5 would result in on average 1,585 infections being avoided. The confidence intervals indicate the level of uncertainty from the outputs of the model but these will underestimate the true uncertainty as they do not account for uncertainty in the model input values. Figure 6-7 Cost breakdown of screening strategies (mean costs) Table 6-22 Cumulative MRSA infections Year In absence of screening (95% CIs) Strategy 2 Strategy 3 Strategy 5 95% CIs) (95% CIs) (95% CIs) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 74

82 Table 6-23 provides a summary of the mean values for prevalence, cost and number of infections predicted when using chromogenic agar for all laboratory tests and using Strategies 2, 3 or 5. The costs of screening per infection avoided over the fiveyear period modelled would therefore be around 2,853 using Strategy 2, around 1,270 for Strategy 3 and approximately 3,726 for Strategy 5. The costs to the health service are not restricted to those incurred by the process of screening. There are, in addition, the opportunity costs of beds being occupied by patients who are being treated for an MRSA infection acquired during their stay in hospital. These beds are therefore unavailable to other patients requiring hospital treatment. In the stochastic model presented here, the opportunity cost has been estimated by subtracting the costs associated with the number of patients unable to enter the hospital under any given strategy from the costs associated with the number of patients turned away from the hospital if MRSA prevalence was zero. As both of these values are random variables in the model, a very large number of replicates would have to be run to ensure that confidence intervals do not include zero. This was not possible in the time available to conduct this modelling exercise, and therefore the following results should be interpreted with considerable caution. The mean opportunity cost over the five-year period modelled would be approximately 299,300 for Strategy 2, 1,460,100 for Strategy 3 and 473,300 for Strategy 5. These results would indicate that the total costs to the hospital would be of the order of 5,771,100 when using Strategy 2, 3,386,600 when using Strategy 3 and 6,376,800 for Strategy 5. These costs reflect only those incurred by the hospital during the episode of care when the infection was incurred and do not give any indication of the longer term costs to the healthcare system, or more importantly to the patient, in terms of loss of quality of life resulting from becoming infected with MRSA Results for a large general hospital The model results for a large general hospital are very similar to those of the tertiary referral hospital as can be seen in Table 6-24, with prevalence of MRSA colonisation dropping rapidly if all patients are screened. Figure 6-8 Cumulative infections over 5 years using chromogenic agar for all laboratory tests 75