Journal of Hospital Infection (2002) 51: 189±200 doi:10.1053/jhin.2002.1247, available online at http://www.idealibrary.com on

Cost-effectiveness of different MRSA screening methods T. Kunori*, B. Cookson*y, J. A. Roberts*, S. Stonez and C. Kibblerz *Public Health and Policy Department, London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT; yLaboratory of Hospital Infection, 61, Colindale Ave., London NW9 5HT; and zRoyal Free Hospital, Pond Street, London, NW3 2QG, UK Summary: We describe a model to examine the cost-effectiveness of various laboratory-screening approaches to detect methicillin-resistant Staphylococcus aureus (MRSA). A critical literature review was used to derive relevant data on the sensitivity (X), specificity (S) and time to result (T) of different tests. Additional cost information was provided by a hospital. Tests were considered in four interactive groups based on a hierarchy of procedures used in laboratories. X, S and Ts of screening tests were then used in formulae to calculate effectiveness for the various tests. The model was developed to explore the effects on MRSA infection acquisition of differing X, S and T for the different tests in detecting MRSA colonized patients admitted to a high-risk unit such as an intensive care unit. It was concluded that taking a sample from the nose alone and inoculating directly on to Ciprofloxacin Baird-Parker agar without broth incubation and confirmation by a Pastorex Staph-Plus test without any methicillin resistance confirmation test was the most cost-effective approach. The complexity of designing this apparently simple scenario is apparent, and we describe many other factors that would need to be considered to refine this model further. However, this and other models should aid the debate and development of more cost-effective screening strategies given the lack of standardization or agreement concerning so many of the variables within the UK and elsewhere. & 2002 The Hospital Infection Society

Keywords: MRSA; methicillin resistant; screening; cost-effectiveness; mathematical modelling; nosocomial infection.

Introduction Methicillin-resistant Staphylococcus aureus (MRSA) is now one of the most important nosocomial infections.1 A variety of MRSA control measures has been developed. Pivotal to these has been the use of various isolation and screening measures. Their use will vary depending on numerous factors outlined in the UK guidelines.2 Isolation of affected patients reduces direct and airborne MRSA transmission. In one paper, for instance, isolation resulted in

Received 30 October 2001; revised manuscript accepted 8 May 2002; published online 28 June 2002. Author for correspondence: B. Cookson, Fax: 0208 200 7449; E-mail: [email protected]

0195±6701/02/070189 ‡ 12 $35.00

a reduction of MRSA infection from 56% to 25%.3 MRSA screening is used in many ways but particularly for the detection of asymptomatic colonized patients during outbreaks and others transferred or re-admitted to hospitals. One report found that 25% of patients were already colonized at admission.4 Although many screening methods have been proposed, few have described their cost savings. Although some studies have compared performance, we have been unable to identify studies that have compared the cost savings between the different methods or any attempt to discuss this within the context of MRSA outbreak modeling. There are many variables that have to be considered when considering such studies. Laboratory processing of MRSA screening specimens, for & 2002 The Hospital Infection Society

190

instance, can be influenced by many factors. These can include, for example, the number of mannitol saltpositive, methicillin-resistant, coagulase-negative staphylococci in the microbial flora of the screened population, whether the MRSA are positive for rapid coagulase tests from salt-containing media, the occurrence of other resistances, such as quinolone resistance (QR), in MRSA, methicillin-sensitive S. aureus (MSSA) and coagulase-negative staphylococci (CNS). Laboratories might prefer to use a method or combination of methods irrespective of the type of problem to avoid confusion in the laboratory. In addition, during large outbreaks of known types of MRSA, a pragmatic approach might be used with just one type of MRSA selective medium, if it has been shown to be effective for that strain. We thus decided to calculate the costeffectiveness ratio (CER) of available screening methods and to determine the most cost-effective method using mathematical modelling based on the published data relating to MRSA screening tests. We anticipate that this model, which can be adapted readily to other clinical settings and laboratory policies, will aid the choice of screening tests. The initial scenario we chose to examine was on an intensive care unit (ICU). Methods Mathematical formulae were developed to calculate the CER. These were based on estimates of the effectiveness of each test identified in a critical review of the literature (see below). Costs of resources were extracted from data provided by the study hospital (the Royal Free). Development of the mathematical model Definition of cost-effectiveness The benefit (cost-effectiveness) of a screening test can be calculated by subtracting from the `baseline' cost of MRSA control without screening, the lower costs of MRSA control from the decreased transmissions due to isolation of MRSA-positive patient isolation following screening. The effectiveness of a screening test in reducing costs is related to three factors: its sensitivity (X), specificity (S) and the time (T) taken to obtain a result. Mathematical formulae were developed to calculate the CER. These were based on estimates of the effectiveness (i.e., S, X, T) of each test identified in our critical review of the

T. Kunori et al.

literature. Costs were also extracted from data provided by the Royal Free Hospital Pathology, Pharmacy, Stores and Financial Accounts Departments. Modelling of MRSA A mathematical model was designed that considered these three factors for the different screening tests and then calculated the cost-effectiveness in controlling MRSA. Certain assumptions were made in this model for screened patients (Figure 1). The screening test was to be performed on patients just admitted to an ICU of a hospital with limited isolation facilities. They were barrier nursed on the open ward and were either clear (negative) of MRSA or were asymptomatic MRSA carriers. All patients then found to be positive were isolated immediately. The average duration of stay (Ln) was 4.05 days, the average for the Royal Free hospital. The days of isolation of these screened patients was 4.05 days minus the time to MRSA detection in days (Ln ÿ T (h)/24). We performed sensitivity analysis with duration of stays of two and 10 days. Transmission from admitted colonized patients to others occurred, but in this model there was no screening of other patients on the ward. Reliance was placed on MRSA being detected in clinical specimens from infected cases. Any MRSA patient placed in a side-room or cohorted in an isolation facility within the ICU, would reduce the numbers of secondary infected cases from 0.2715 per primary colonized patients (Rn) to 0.017 per primary colonized patients per day (Ri).5 We also performed sensitivity analysis with Rns of 0.13 and 0.54. In this model an ideal `gold standard' laboratory test with 100% sensitivity and specificity was assumed to be carried out as soon as infection had developed in these secondary infected cases. They were then isolated and treated immediately for an average of 12 days (Li: the average at the Royal Free Hospital). Isolation costs were estimated as £306.93 per patient per day (Ci): summated from the Royal Free cost of gloves (£0.02) and a paper gown (£1.95) and the additional labour cost for 1 min of time5 needed to don these items (£0.161) and assumed health personnel entered a patient room 144 times a day.6 The additional costs for a patient who developed MRSA infection were £11 788.61 per person (Ct) estimated by summing the additional cost before the extension of ICU stay (Ln: 4.05 days) and after the extension (Li ÿ Ln: 12±4.05 days). Additional costs

Cost-effectiveness of MRSA screening

191

Figure 1 Modelling of primary and secondary cases. 2 Cases: secondary, cases; *transmission rate (Rn) is 0.2715 per colonized person per day; # Transmission rate (Ri) is 0.0170 per colonized person per day (see reference 5 and text). See the text for explanations, and Table I for formulae, relating to b1, b2, b3, b4 and b5.

before this extension were estimated by summing the cost of vancomycin (£26.82 per day) and the isolation costs (Ci: £306.93 per day), and after extension by adding the cost of the side-room (overhead, medical and nursing staff time, isolation, diagnosis and other treatment; £1286 per day) and the cost of vancomycin treatment. The costs of bed days and personnel were derived from data provided by the Finance Department of the Royal Free Hospital. The data related to length of stay (Ln, Li) and other cost information were collected from the ICU of Royal Free Hospital. To simplify the model, no test was performed for the approximately 70%5 (1-D) of patients to whom MRSA was transmitted but who did not become infected, neither were these patients isolated. We did not consider transmission from any secondary cases for this part of the exercise, as this would have required a very complex and dynamic model. We assumed that the MRSA were resistant to quinolones as is true for the majority of MRSA in the UK at present.2

Mathematical modelling The various formulae used in the study, such as the MRSA control costs without screening, are described in Table I. The benefit of the screening test was calculated by subtracting from the `baseline' cost of MRSA control without screening (`a' in Table I), the MRSA control costs when using a particular screening test calculated by adding the costs `b1', `b2', `b3', `b4' and `b5' shown in Table I and Figure 1.

CER estimation ˆ [NX(Ln ÿ T/24){DCt(Rn ÿ Ri) ÿ Ci} ÿ (1 ÿ S)(P ÿ N)(Ln ÿ T/24)Ci]/PC To compare the CER of each screening test we used the same data for the number (N), which depended on the prevalence rate of the screened population and

192

T. Kunori et al.

Table I Relevant formulae used in mathematical modelling Terms

Explanation

Estimation

a

Absolute cost of controlling MRSA without a screening test Cost for secondary cases who are infected from primary colonized cases before the screening result appeared Cost for secondary cases who are infected from false-negative, primary-colonized cases Cost for secondary cases who are infected from true-positive, primary-colonized cases Isolation cost for primary colonized cases Unnecessary isolation cost for false-positive, colonized cases Cost-effectiveness ratio {a ÿ (b1 ‡ b2 ‡ b3 ‡ b4 ‡ b5)}/PC CER (when N ˆ 1, P ˆ 50, Ci ˆ 306.93, Ct ˆ 11,788.61, Ln ˆ 4.05, Ri ˆ 0.017, Rn ˆ 0.2715 D ˆ 0.3) See text: Calculation related to {a ÿ (b1 ‡ b2 ‡ b3 ‡ b4 ‡ b5)} ˆ PC

NDRnLnCt

b1 b2 b3 b4 b5 CERs

Borderline prevalence rate

NDRnCtT/24 NDRnCt(1 ÿ X)(Ln ÿ T/24) NDXRi Ct(Ln ÿ T/24) NXCi(Ln ÿ T/24) Ci(1 ÿ S)(P ÿ N)(Ln ÿ T/24) [NX(Ln ÿ T/24){DCt(Rn ÿ Ri) ÿ Ci} ÿ (1 ÿ S)(P ÿ N)(Ln ÿ T/24)Ci]/PC {70.8X ÿ 0.729XT ÿ 1218(1 ÿ S) ‡ 12.5T(1 ÿ S)}/C {Ci (1 ÿ S)(Ln ÿ T/24) ‡ c1 ‡ c3 ‡ c6 ‡ (c4 ‡ c5)(1 ÿ s3)}/ [{DXCt(Rn ÿ Ri) ÿ SCi}(Ln ÿ T/24) ÿ {(x1)(x3) ‡ s3 ÿ 1}(c4 ‡ c5)]

C: overall screening cost (£); N: number of primary MRSA colonized patients; P: number of patients screened; S: overall specificity; T: overall time (h) until a positive screen result is available; X: overall sensitivity; Ci: daily cost of isolation (£): £306.93 (See text for the detail of calculation); Ct: additional cost for secondary infected patient (£): £11 788.61 (See text for the detail of calculation); Ln: length of stay for non-MRSA infected patient (day): 4.05 days; Ri: daily rate of transmission from isolated MRSA primary patient (number of transmitted secondary cases from one primary case per day): 0.0170 per colonized person per day (from ref 5); Rn: daily rate of transmission from non-isolated MRSA primary patient (number of transmitted secondary cases from one primary case per day): 0.2715 per infected person per day (from ref 5); D: The expected number of MRSA infection among secondary colonized patients: 0.3 (from ref 5), see Figure 1 for various b categories and Figure 2 for `x, s, c', categories and the text for transmission rates and isolation costs; CER: cost-effectiveness ratio.

P the scale of the screening programme. Thus for N ˆ l, P ˆ 50: CER estimation ˆ {71.2X ÿ 0.733XT ÿ 1218(1 ÿ S) ‡ 12.5T(1 ÿ S)}/C

Exploration of different sensitivity, specificity, time, cost data for each laboratory test The figures for sensitivity (X), specificity (S), time (T) and screening cost (C) used in the formulae had to be those for the whole process of that screening test from the taking of the samples to the reporting of the results. The latter was difficult to identify in many of the papers. These usually mentioned the specificity, sensitivity and time for one of the processes in a screening related test such as S. aureus identification. Therefore, in order to calculate an overall figure for the three factors, each was integrated by the formulae described in Figure 2 using the most cost-effective method for the previous stage of the test.

Overall sensitivities, specificities, times and screening cost in the study were divided into four categories according to the reported timing after a methicillin-susceptibility test (Group A), after a S. aureus identification test (Group B), after detection by selective incubation media (Group C) and after direct identification, such as with a multiplex PCR (Group D). We then compared CERs for several laboratory-screening methods within and between each of these different groups. When the comparison of CER within the same group was undertaken, common sensitivity, specificity, time and cost variable data of previous test stages were used as coefficients (see Figure 2) and only the effect of last stage of the laboratory test was then compared. For example, only the methicillin-susceptibility test stage differed in Group A. In each group, the sensitivity, specificity, time and cost of the most cost-effective test in the previous stage from our final result were used as coefficients (Table II). The most cost-effective collecting site derived from our final result (Table III) was also used for the calculation as a coefficient (`x1' in Figure 2). It was impossible to collect information about sensitivity of the swab taking techniques and swab storage

Cost-effectiveness of MRSA screening

193

Sample taking

Figure 2 Calculations for overall sensitivity (X), specificity (S), time (T) and screening cost (C) for each of the groups of methods. The figures for sensitivity, specificity, time and cost of most cost-effective test of each stage were used when the calculations of the next stage were undertaken. For *1: the nose; for *2: 18 incubation on CBP without broth; for *3: Pastorex Staph-Plus. y These figures were from Royal Free Hospital; N: number of primary MRSA colonized patients; P: number of patients screened.

conditions from the literature, and both were thus assumed to be 100%. The coefficients used in the calculations are shown in Figure 2. The study formulae were also used to compare the CER of different sampling sites. In this calculation, the BBL Crystal MRSA ID test in Group A was used as a point of reference to take into account other factors which can affect overall sensitivity, specificity, time and cost of a result. Another assumption made was that the cost for two or three sites was two or three times that for a single site. The overall screening cost (C) was affected by sensitivity and specificity of the selective incubation media and sampling site. In Groups A and B, a S. aureus identification test and a methicillinsusceptibility test were not undertaken for all screened cases, but only for those where suspected MRSA colonies were grown on selective incubation media. The cost of S. aureus identification and methicillin susceptibility was thus modified by sensitivity and specificity of the selective incubation media and sampling site.

In order to compare the CER of several laboratory tests by using these formulae, specific variables for sensitivity, specificity, time and cost for each laboratory test were needed. The methods for obtaining this information are described below.

Screening cost calculation The laboratory test costs were based on information collected from material and labour costs of each laboratory test derived from 1998±1999 costing returns in the Department of Medical Microbiology and the Infection Control Team at the Royal Free NHS Trust. The cost for one sample was much less if many samples were tested at the same time. For example, the incubation media and DNAse test costs were reduced fourfold as a quadrant of an agar plate was used for each of four samples and the costs of PCR and of the methicillin-susceptibility tests were processed more cost-effectively in batches of 20 tests performed at the same time.

194

T. Kunori et al.

Table II Cost-effectiveness ratios and borderline prevalence rate of each test Cost-effectiveness ratio

Prevalence (%) calculated from cost-effectiveness data

Test groups*

Details of the relevant testsy

A

BBL Crystal MRSA ID (Beckton Dickinson, Cowley, UK) Automated API ATB (bioMeÂrieux, Basingstoke, UK) Agar screening Disc diffusion (NCCLS, USA) Agar dilution (NCCLS, USA) E-test (Cambridge Diagnostics, Cambridge, UK) Broth microdilution (NCCLS, USA)

4.195 3.744 3.732 2.916 2.382 2.019 1.983

0.482 0.533 0.536 0.740 0.981 0.988 0.998

B

Pastorex Staph-Plus (Sanofi Diagnostic Pasteur, Paris, France) Staph Latex (Amer. Microsan, Mahwah, NJ, USA) Staphyloslide (Beckton Dickinson, Cowley, UK) Staphaurex (Murex Diagnostic Limited, Kent, England) Slidex Staph (bioMeÂrieux, Basingstoke, UK) Staphylochorome (Beckton Dickinson, Cowley, UK) Prolex Staph Latex (Pro-Lab Diagnostics, Wirral, UK) Bacto Staph (Difco Laboratories, Detroit, MI, USA) Slide Coagulase (at source, Royal Free) Thermonuclease (Remmal Lab., Kansas, USA) Tube coagulase (at source, Royal Free)

5.466 5.437 5.410 5.366 5.200 5.197 5.024 4.765 4.323 3.925 3.642

0.386 0.371 0.389 0.418 0.500 0.384 0.509 0.453 0.519 0.509 0.555

C

CBP 18 h incubation without broth (Oxoid, Basingstoke, UK) CBP 42 h incubation without broth CBP 42 h incubation ‡ broth subculture CBP 18 h incubation ‡ broth subculture MMSA 18 h incubation without broth (Oxoid, Basingstoke, UK) MMSA 42 h incubation ‡ broth subculture MMSA 18 h incubation ‡ broth subculture MMSA 42 h incubation without broth MSA 42 h incubation ‡ broth subculture (Oxoid, Basingstoke, UK) MSA 42 h incubation without broth MSA 18 h incubation without broth MSA 18 h incubation ‡ broth subculture

ÿ0.056 ÿ0.460 ÿ2.520 ÿ3.211 ÿ28.074 ÿ29.628 ÿ40.762 ÿ46.187 ÿ54.220 ÿ92.671 ÿ95.430 ÿ95.580

2.362 2.656 4.537 3.811 12.022 21.032 17.826 21.018 32.856 36.242 31.154 35.194

D

Modified PCR (Promega, Southampton, UK) PCR multiplex (Promega, Southampton, UK)

5.035 ÿ12.170

0.394 6.854

*A: methicillin-susceptibility test; B: S. aureus identification test; C: selective incubation media; D: direct identification test (see Figure 2). ySee text and Figure 2 for explanation of the context of the relevant tests and formulae. Broth subculture. 24 h previous incubation in 2 mL 1% Tryptone-T broth with 60 g/L sodium chloride. CBP: Ciprofloxacin Baird-Parker agar; MSA: Mannitol salt agar; MMSA: Mannitol salt agar with methicillin; NCCLS: National Committee for Clinical Laboratory Standards, USA. Table III Cost-effectiveness ratios of the different screening sites Site Single site Nose Perineum Skin Axilla Urine Stool Throat Combination of sites Nose and wound Nose and perineum Nose and throat Nose, throat and perineum

Average sensitivity (%)*

Cost-effectiveness ratio

64.17 56.15 38.00 25.00 22.00 18.68 14.65

8.36 7.32 4.95 3.21 2.87 2.55 2.43

100.00 93.40 85.60 98.30

1.91 6.52 4.27 6.09

Literature review A literature review was performed to detect key variable data from the past literature using MEDLINE between the years 1960 and 1999. The full review strategy is available from the corresponding author. However, in brief, the keywords `MRSA, EMRSA, methicillin, oxacillin, resistant, resistance, Staphylococcus and Staphylococci' were used to yield 2919 papers. In parallel, 25 keywords such as `agar, screen(s), identifying, isolates, detection' and variants of these terms yielded 1 368 237 papers. Combining these two strategies resulted in 291 papers. Restriction to the English language produced 243 papers. Further selection used the three variables, sensitivity, specificity and time. There were

Cost-effectiveness of MRSA screening

eventually 152 papers of which only 17 contained the key data and provided a total of 108 relevant laboratory test results. There were 56 for methicillinsusceptibility tests,7±14 36 for S. aureus identification tests,15±19 six for detection only by selective incubation media without broth,20 six for detection only by incubation media with broth,20 and four for direct identification.21±23 A similar strategy was used for sampling sites. This yielded 43 papers of which only six mentioned sensitivity of different sites in colonized patients.24±29 Analysis of cost-effectiveness data The CER was calculated by dividing the benefit (savings) by the cost of each laboratory test (see Table I). The average CER for identical or similar tests was calculated. For each screening test a `borderline prevalence rate' was calculated, this being the prevalence of MRSA-positive patients (N) in the number of screened patients (P) at which the benefit (savings) was equal to the cost of that screening test (see Table I). Results In our review of the literature, we found that every paper had used different experimental conditions even where the same laboratory method had ostensibly been used. This created some potential problems in comparing different tests. For example, some methicillin-susceptibility tests used isolates with low-level methicillin resistance. Culture conditions, such as the incubation temperature or the concentration of sodium chloride, also varied. It was impossible to adjust for some of these imponderables, so an average of all these similar types of tests was used. Table III shows comparison of CERs for different sampling sites. The most cost-effective single sampling site was the nose, although this was exceeded by any combination of sampling methods. A combination of nose and wound was optimal. We could find no data on sampling from other sites recommended in the guidelines2 such as intravenous lines and pressure sores. We decided to use the nasal site as a sole screening site for this model. The cost-effectiveness of different methicillinsusceptibility tests (Group A) are shown in Table II. The most cost-effective of the 11 S. aureus identification tests evaluated was the Pastorex Staph Plus test followed by Staph Latex. The CER of the tube

195

coagulase test was low despite its 100% specificity and recognition as a gold standard, because it was so time-consuming to perform. This was even more so for the thermonuclease test. All mannitol salt agar (MSA) and methicillin MSA (MMSA) media (Table II) performed poorly with negative CERs. The Ciprofoxacin Baird-Parker (CBP) media was the most cost-effective medium of the three. An 18 h incubation time was more costeffective than 42 h for both CBP and MMSA, if no subculture by broth was performed. Overall, the most cost-effective test in Group C (selective incubation media) was 18 h incubation using CBP without broth. Subculture from broth was only more cost-effective when MSA or MMSA media were incubated for 42 h. When we reviewed papers relating to Group D (direct identification), most had used the tube coagulase test as the standard for specificity. However, this failed to identify three PCR and hyper-selective medium positive isolates in one paper (see reference No. 21). Despite this result the CER for the multiplex PCR method using mecA and femA in Table II were calculated on the condition that the conventional phenotypic method was the most accurate. Although the modified PCR method had 100% sensitivity and specificity and a rapid identification time, its CER was lower than six of the tests used in Group B. This was because the test was very expensive even without the capital equipment and training costs. Reporting after the confirmation by a S. aureus identification test increased the CER in comparison with reporting after the selective incubation media. This was because the CERs of all S. aureus identification tests were greater than the CERs of CBP with 18 h incubation without broth (the relevant test required in the hierarchy of methods used; see Figure 2). Reporting after the confirmation by a methicillinsusceptibility test reduced the CER in comparison with reporting after the confirmation by a S. aureus identification test, because all CERs of methicillinsusceptibility tests were smaller than the CER of Pastorex Staph-Plus (see Table II). The reason for the use of the figure for the Pastorex Staph-Plus as a comparison was that data for sensitivity, specificity, time and cost of the test were used as coefficients when the cost-effectiveness calculation of methicilliinsusceptibility tests was undertaken (Figure 2). This meant that, although the methicillin-susceptibility test improved specificity, it was outweighed by the cost and time involved.

196

T. Kunori et al.

These results suggest that the most cost-effective test in the model was to use CBP with 18 h incubation without broth and to add a Pastorex Staph-Plus test as a confirmatory test without the addition of a methicillin-susceptibility test. As can be seen in Table II, if the prevalence of MRSA in the screening population is more than 0.998%, the money saved on MRSA control measures more than covered the cost of the screening programmes for Groups A and B. Sensitivity analyses The effects of changing lengths of stay and transmission rates are shown in Tables IV and V. In summary,

when CERs were positive, the CERs greatly increased with increased length of stay and, where these were negative, they were further decreased as length of stay decreased. There were also some changes in the rank order (see Table IV). Some tests take longer than two days, and so would not provide a timely result for the affected patients, although they would provide useful general infection control data. Where the transmission rate increased, so did the CER of the tests. There were also some changes in the rank order in Test Groups B and C (See Table V); as the transmission rates decreased, so the importance of specificity increased. This was particularly noticeable in the change in rank order for the Prolex and Slidex identification tests.

Table IV Sensitivity analysis exploring how the cost-effectiveness ratio (CER) was affected by changes in the average intensive care unit stay Length of Stay 10 days

4.05 days

2 days*

Tests performed in each category (see Table 2)

CER

Rank order

CER

Rank order

CER

Rank order

BBL Crystal MRSA ID (Beckton Dickinson, Cowley, UK) Automated API ATB (bioMeÂrieux, Basingstoke, UK) Agar screening Disc diffusion (NCCLS, USA) Agar dilution (NCCLS, USA) E-test (Cambridge Diagnostics, Cambridge, UK) Broth microdilution (NCCLS, USA) Pastorex Staph-Plus (Sanofi Diagnostic Pasteur, Paris, France) Staph Latex (Amer. Microsan, Mahwah, NJ, USA) Staphyloslide (Beckton Dickinson, Cowley, UK) Staphaurex (Murex Diagnostic Limited, Kent, England) Slidex Staph (bioMeÂrieux, Basingstoke, UK) Staphylochorome (Beckton Dickinson, Cowley, UK) Prolex Staph Latex (Pro-Lab Diagnostics, Wirral, UK) Bacto Staph (Difco Laboratories, Detroit, MI, USA) Slide Coagulase (at source, Royal Free) Thermonuclease (Remmal Lab., Kansas, USA) Tube coagulase (at source, Royal Free) CBP 18 h incubation without broth (Oxoid, Basingstoke, UK) CBP 42 h incubation without broth CBP 42 h incubation ‡ broth subculture CBP 18 h incubation ‡ broth subculture MMSA 18 h incubation without broth (Oxoid, Basingstoke, UK) MMSA 42 h incubation ‡ broth subculture MMSA 18 h incubation ‡ broth subculture MMSA 42 h incubation without broth MSA 42 h incubation ‡ broth subculture (Oxoid, Basingstoke, UK) MSA 42 h incubation without broth MSA 18 h incubation without broth MSA 18 h incubation ‡ broth subculture PCR multiplex (Promega, Southampton, UK) Modified PCR (Promega, Southampton, UK)

12.3 13.5 13.5 12.4 10.4 11.4 11.2 15.4 15.3 15.2 15.1 14.6 14.6 14.1 13.4 12.2 14.2 12.4 ÿ0.159 ÿ1.66 ÿ14.2 ÿ11.6 ÿ79.0 ÿ167 ÿ147 ÿ167 ÿ306 ÿ335 ÿ269 ÿ345 ÿ13.1 ÿ31.3

3 1 2 4 7 5 6 1 2 3 4 5 6 8 9 11 7 10 1 2 4 3 5 8 6 7 10 11 9 12 1 2

4.195 3.744 3.732 2.916 2.382 2.019 1.983 5.466 5.437 5.41 5.366 5.2 5.197 5.024 4.765 4.323 3.925 3.642 ÿ0.056 ÿ0.46 ÿ2.52 ÿ3.211 ÿ28.074 ÿ29.628 ÿ40.762 ÿ46.187 ÿ54.22 ÿ92.671 ÿ95.43 ÿ95.58 5.075 ÿ12.17

1 2 3 4 5 6 7 1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11 12 1 2

1.38 0.376 0.375 0.203 0.152 0 0 2.048 2.037 2.027 2.011 1.948 1.948 1.882 1.785 1.62 0.395 0.616 ÿ0.021 ÿ0.046 0 ÿ0.323 ÿ10.523 0 ÿ4.099 ÿ4.644 0 ÿ9.318 ÿ35.771 ÿ9.61 2.322 ÿ5.568

1 2 3 4 5 Ð Ð 1 2 3 4 5 6 7 8 9 11 10 1 2 Ð 3 8 Ð 4 5 Ð 6 9 7 1 2

*A missing value indicates that the test takes longer than two days (see text).

Cost-effectiveness of MRSA screening

197

Table V Sensitivity analysis exploring how the cost-effectiveness ratio was affected by changes in the transmission rate Transmission rate (/primary patient/day) 0.54 Tests performed in each category (see Table 2)

CER

0.27 Order

CER

0.13 Order

CER

Order

BBL Crystal MRSA ID (Beckton Dickinson, Cowley, UK) Automated API ATB (bioMeÂrieux, Basingstoke, UK) Agar screening Disc diffusion (NCCLS, USA) Agar dilution (NCCLS, USA) E-test (Cambridge Diagnostics, Cambridge, UK) Broth microdilution (NCCLS, USA)

10.912 9.745 9.712 7.604 6.204 5.252 5.161

1 2 3 4 5 6 7

4.195 3.744 3.732 2.916 2.382 2.019 1.983

1 2 3 4 5 6 7

0.656 0.581 0.581 0.445 0.368 0.316 0.308

1 2 3 4 5 6 7

Pastorex Staph-Plus (Sanofi Diagnostic Pasteur, Paris, France) Staph Latex (Amer. Microsan, Mahwah, NJ, USA) Staphyloslide (Beckton Dickinson, Cowley, UK) Staphaurex (Murex Diagnostic Limited, Kent, England) Slidex Staph (bioMeÂrieux, Basingstoke, UK) Staphylochorome (Beckton Dickinson, Cowley, UK) Prolex Staph Latex (Pro-Lab Diagnostics, Wirral, UK) Bacto Staph (Difco Laboratories, Detroit, MI, USA) Slide Coagulase (at source, Royal Free) Thermonuclease (Remmal Lab., Kansas, USA) Tube coagulase (at source, Royal Free)

14.328 14.16 14.18 14.195 14.159 13.518 13.655 12.557 11.262 10.209 9.473

1 4 3 2 5 7 6 8 9 10 11

5.466 5.437 5.41 5.366 5.2 5.197 5.024 4.765 4.323 3.925 3.642

1 2 3 4 5 6 7 8 9 10 11

0.796 0.84 0.788 0.713 0.479 0.812 0.475 0.658 0.666 0.614 0.569

3 1 4 5 10 2 11 7 6 8 9

9.089 6.505 1.753 3.984 ÿ19.87 ÿ25.527 ÿ33.787 ÿ39.88 ÿ50.34 ÿ87.054 ÿ88.107 ÿ89.511

1 2 4 3 5 6 7 8 9 10 11 12

ÿ0.056 ÿ0.46 ÿ2.52 ÿ3.211 ÿ28.074 ÿ29.628 ÿ40.762 ÿ46.187 ÿ54.22 ÿ92.671 ÿ95.43 ÿ95.58

1 2 3 4 5 6 7 8 9 10 11 12

ÿ4.876 ÿ4.131 ÿ4.772 ÿ7.003 ÿ32.397 ÿ31.79 ÿ44.438 ÿ49.511 ÿ56.264 ÿ95.631 ÿ99.289 ÿ98.779

3 1 2 4 6 5 7 8 9 10 12 11

13.199 ÿ4.046

1 2

5.075 ÿ12.17

1 2

0.793 ÿ16.451

1 2

CBP 18 h incubation without broth (Oxoid, Basingstoke, UK) CBP 42 h incubation without broth CBP 42 h incubation ‡ broth subculture CBP 18 h incubation ‡ broth subculture MMSA 18 h incubation without broth (Oxoid, Basingstoke, UK) MMSA 42 h incubation ‡ broth subculture MMSA 18 h incubation ‡ broth subculture MMSA 42 h incubation without broth MSA 42 h incubation ‡ broth subculture (Oxoid, Basingstoke, UK) MSA 42 h incubation without broth MSA 18 h incubation without broth MSA 18 h incubation ‡ broth subculture PCR multiplex (Promega, Southampton, UK) Modified PCR (Promega, Southampton, UK)

Discussion From our theoretical model, the most cost-effective screening method was to take a single sample from the nose and directly inoculate it on to CBP without the use of broth and confirmed by a Pastorex Staph-Plus (a staphylococcal latex test) but without any methicillin-resistance confirmatory test. The model found that specificity was not a major determinant of CER, as it was almost 100% in most combinations of selective incubation media and confirmation of identification testing. However, sensitivity was highly affected by the sampled site and the selective incubation media. Raising the sensitivity of these would make the most significant improvement in overall sensitivity. For example, after the Pastorex Staph-Plus test, with an overall

specificity of almost 100 (99.98)%, the overall sensitivity was only 45.63%, mainly due to the poor sensitivity of the sampling site (64.17%) and selective incubation media (73%). We addressed the complexity of the testing process by dividing it into several stages, producing what we consider to be a very useful estimate of the overall CER of each stage and combination of the various tests. In addition, the model enabled the comparison of reporting times. Even if the combination of each stage was changed, the model could estimate the overall CER with small changes to the formulae. Our model has, of course, several limitations. The literature review revealed many laboratory methodological differences, which made interpretation and analysis very difficult. Standardization of the approaches used in publications would make this task

198

much easier and results more applicable, as well as simplifying future models. Problems were encountered with sampling sites in that there were very limited data, and data for some important sampling sites were not found in the literature. We used the most sensitive single site in this model, but laboratories could add to this model using their local data. It was based on the assumption that there was no interaction between each stage of any test. However, there are several possible ways in which such interactions may occur. For example, Davies and co-workers30 showed a difference in the effects of culture media on the rapid slide coagulation test.3 Colonies from MSA, unlike those from CBP, led to a false-negative result. Allen and co-workers31 showed an interaction between the different microbial flora at a sampling site and the incubation media. For example, coliforms from perineal samples reduced the sensitivity for methicillin-blood agar, but not that of methicillin-milk agar or Chapman's medium.31 The model was based on a scenario where all patients admitted to a high-risk ward, such as an ICU, were screened. A recent survey of ICUs found this to be common practice (Ms M. Carter, personal communication). Although isolation was recommended in previous UK guidelines on all wards, with the increasing occurrence of MRSA few hospitals now have this capability and a risk assessment process is recommended.2 In addition, ICUs have limited numbers of side-rooms and there are quality of patient care issues when certain patients are isolated in this manner. The patient is more commonly barrier nursed, and this can be carried out with variable compliance resulting in MRSA transmission as explored in this model. We need further data on the resource implications for different types of isolation. We applied the data fron two studies in this work.5,6 This MRSA transmission modelling was typical of most outbreaks. However, we considered only primary screening and the infected secondary transmission events. As can be seen, this was not a simple matter and the decision at this stage was made not to consider the complexities of colonized secondary cases or tertiary transmissions. The model can of course be adapted for other scenarios within and without the ICU. Dynamic models will also be needed. We did explore the effects of lengths of stay and different transmission data. This work showed the importance of validating the approach locally as CERs did change for most tests and, to

T. Kunori et al.

a lesser extent, so did the rank order for some categories of tests. However, the overall conclusions were the same. Our screening cost estimation method also has some limitations. First, the estimated cost only included incremental costs. It did not contain overhead costs for electricity, water, and maintenance. However, these were not considered to be affected by the different MRSA tests and attribution of these costs to particular tests is complex. In addition, they were difficult to estimate, might be regarded as relatively fixed, and were not, apart from PCR tests, thought to be a large part of the total cost. A full costing study would include all items.32 There were complex issues relating to when to apportion the costs of susceptibility and identification tests to other tests. For example, we chose not to add the cost of a methicillin-confirmation test to the selective incubation media. This is variably used in laboratories (e.g., from all screened sites or infected sites), and can be combined with other susceptibility tests for medico-legal, treatment or epidemiological typing purposes. The approach we have used should be adapted and applied to meet local circumstances and the results interpreted to guide local practices. Another issue relates to the type of MRSA encountered. The CBP media, unlike the other media, will not detect quinolone-sensitive (QS) MRSA. The majority of current UK and Northern European MRSA are quinolone resistant (QR), indeed, the paper describing the media assumed an incidence of QR MRSA of 98%.20 However, this can change and new QS MRSA strains can be important locally. CBP does have the added advantage of preventing many QS CNS and MSSA complicating the detection of QR MRSA growing on selection media. From discussions with laboratories several now adopt an approach where CBP is combined in many screening scenarios with other selective media, e.g., MMSA (Dr Barry Cookson, personal communication). Other models and data are required to provide costing information on which to base these decisions. Work underway with rather more complex models has also focused on the interactions between the incidence of QR MRSA, MSSA and CNS and the necessity to perform the various confirmatory tests for the reasons mentioned above. It has found, thus far, that the rate of QR MSSA and QR CNS has little effect on the cost-effectiveness if additional S. aureus identification and methicillinsusceptibility tests are undertaken. This is because

Cost-effectiveness of MRSA screening

there is little change in the overall specificity when these other QR staphylococci are also present in screened samples. The main factor that affects costeffectiveness is, as one might have expected, the rate of QR MRSA. The CER calculated in this study was not a perfect estimate of the real cost-effectiveness of these tests. However, the modelling used was comparatively simple, and we found it to be a very useful way to compare different laboratory tests that are carried out in several stages. The accuracy of this and other models depends on the application of these screening tests and the quality of data. The use of this model, however, should enable infection control teams and microbiology departments to develop more costeffective screening strategies given the lack of standardization of so many of the variables within the UK and elsewhere.

Acknowledgements We would like to thank to Ms Yvonne Carter, Infection Control, and Ms Sheila Ainscough in the Department of Medical Microbiology, and Mrs Sheena Thorne in the Financial Department of the Royal Free Hospital and her colleagues in pharmacy, pathology and stores departments for assistance and collecting the cost information, and Dr Ben Cooper for helpful discussions and facilitating access to his data.

References 1. Boyce JM, Landry M, Deetz TR, DuPont HL. Epidemiologic studies of an outbreak of nosocomial methicillin resistant Staphylococcus aureus infections. Infect Control 1981; 2: 110±116. 2. Ayliffe GAJ, Buckles A, Casewell MW et al. Reviewed guidelines for the control of methicillin-resistance Staphylococcus aureus infection in hospitals. J Hosp Infect 1998; 39: 253±290. 3. Selkon JB, Stokes ER, Ingram HR. The role of an isolation unit in the control of hospital infection with methicillin-resistant staphylococci. J Hosp Infect 1980; 1: 41±46. 4. Bradley BF, Terpenning MS, Ramsey MA et al. Methicillin-resistant Staphylococcus aureus: colonization and infection is a long-term care facility. Ann Intern Med 1991; 115: 417±422. 5. Jernigan JA, Clemence MA, Stott GA et al. Control of methicillin-resistant Staphylococcus aureus at a university hospital: one decade later. Infect Control Hosp Epidemiol 1995; 16: 686±696.

199 6. Cooper BS. The transmission dynamics of methicillinresistant Staphylococcus aureus and vancomycinresistant enterococci in hospital wards. PhD thesis, University of Warwick, UK. 2000. 7. Kubina M, Jaulhac B, Delabranche X, Lindenmann C, Piemont Y, Monteil H. Oxacillin susceptibility testing of staphylococci directly from Bactec Plus blood cultures by the BBL Crystal MRSA ID system. J Clin Microbiol 1999; 37: 2034±2036. 8. Kampf G, Lecke C, Cimbal AK, Weist K, Ruden H. Evaluation of mannitol salt agar for detection of oxacillin resistance in Staphylococcus aureus by disk diffusion and agar screening. J Clin Microbiol 1998; 36: 2254±2257. 9. Weller TM, Crook DW, Crow MR, Ibrahim W, Pennington TH, Selkon JB. Methicillin susceptibility testing of staphylococci by Etest and comparison with agar dilution and mecA detection. J Antimicrob Chemother 1997; 39: 251±253. 10. Zambardi G, Fleurette J, Schito GC et al. European multicentre evaluation of a commercial system for identification of methicillin-resistant Staphylococcus aureus. Eur J Clin Microbiol Infect Dis 1996; 15: 747±749. 11. Wallet F, Roussel Delvallez M, Courcol RJ. Choice of a routine method for detecting methicillin-resistance in staphylococci. J Antimicrob Chemother 1996; 37: 901±909. 12. Bignardi GE, Woodford N, Chapman A, Johnson AP, Speller DC. Detection of the mec-A gene and phenotypic detection of resistance in Staphylococcus aureus isolates with borderline or low-level methicillin resistance. J Antimicrob Chemother 1996; 37: 53±63. 13. Richard P, Meyran M, Carpentier E, Thabaut A, Drugeon HB. Comparison of phenotypic methods and DNA hybridization for detection of methicillinresistant Staphylococcus aureus. J Clin Microbiol 1994; 32: 613±617. 14. Huang MB, Gay TE, Baker CN, Banerjee SN, Tenover FC. Two percent sodium chloride is required for susceptibility testing of staphylococci with oxacillin when using agar-based dilution methods. J Clin Microbiol 1993; 31: 2683±2688. 15. Gupta H, McKinnon N, Louie L, Louie M, Simor AE. Comparison of six rapid agglutination tests for the identification of Staphylococcus aureus, including methicillin-resistant strains. Diagn Microbiol Infect Dis 1998; 31: 333±336. 16. Smole SC, Aronson E, Durbin A, Brecher SM, Arbeit RD. Sensitivity and specificity of an improved rapid latex agglutination test for identification of methicillinsensitive and resistant Staphylococcus aureus isolates. J Clin Microbiol 1998; 36: 1109±1112. 17. Tveten Y. Evaluation of new agglutination test for identification of oxacillin-susceptible and oxacillinresistant Staphylococcus aureus. J Clin Microbiol 1995; 33: 1333±1334. 18. Qadri SM, Akhter J, Qadri SG. Latex agglutination and hemagglutination tests for the rapid identification of methicillin sensitive and methicillin resistant Staphylococcus aureus. J Hyg Epidemiol Microbiol Immunol 1991; 35: 65±71.

200 19. Berke A, Tilton RC. Evaluation of rapid coagulase methods for the identification of Staphylococcus aureus. J Clin Microbiol 1986; 23: 916±919. 20. Davies S, Zadik PM. Comparison of methods for the isolation of methicillin resistant Staphylococcus aureus. J Clin Pathol 1997; 50: 257±258. 21. Vannuffel P, Laterre PF, Bouyer M et al. Rapid and specific molecular identification of methicillinresistant Staphylococcus aureus in endotracheal aspirates from mechanically ventilated patients. J Clin Microbiol 1998; 36: 2366±2368. 22. Lawrence C, Cosseron M, Mimoz O et al. Use of the coagulase gene typing method for detection of carriers of methicillin-resistant Staphylococcus aureus. J Antimicrob Chemother 1996; 37: 687±696. 23. Mir N, Sanchez M, Baquero F, Lopez B, Calderon C, Canton R. Soft salt-mannitol agar-cloxacillin test: a highly specific bedside screening test for detection of colonization with methicillin-resistant Staphylococcus aureus. J Clin Microbiol 1998; 36: 986±989. 24. Sanford MD, Widmer AF, Bale MJ, Jones RN, Wenzel RP. Efficient detection and long-term persistence of the carriage of methicillin-resistant Staphylococcus aureus. Clin Infect Dis 1994; 19: 1123±1128. 25. Harbarth S, Dharan S, Liassine N, Herrault P, Auckenthaler R, Pittet D. Randomized, placebocontrolled, double-blind trial to evaluate the efficacy of mupirocin for eradicating carriage of

T. Kunori et al.

26.

27.

28. 29.

30. 31.

32.

methicillin-resistant Staphylococcus aureus. Antimicrob Agents Chemother 1999; 43: 1412±1416. Coello R, Jimenez J, Garcia M et al. Prospective study of infection, colonization and carriage of methicillinresistant Staphylococcus aureus in an outbreak affecting 990 patients. Eur J Clin Microbiol Infect Dis 1994; 13: 74±81. Hicks NR, Moore EP, Williams EW. Carriage and community treatment of methicillin-resistant Staphylococcus aureus: what happens to colonized patients after discharge? J Hosp Infect 1991; 19: 17±24. Brady LM, Thomson M, Palmer MA, Harkness JL. Successful control of endemic MRSA in a cardiothoracic surgical unit. Med J Aust 1990; 152: 240±245. Lee YL, Cesario T, Gupta G et al. Surveillance of colonization and infection with Staphylococcus aureus susceptible or resistant to methicillin in a community skilled-nursing facility. Am J Infect Control 1997; 25: 312±321. Davies S. Detection of methicillin-resistant Staphylococcus aureus: the evaluation of rapid agglutination methods. Br J Biomed Sci 1997; 54: 13±15. Allen JL, Cowan ME, Cockroft PM. A comparison of three semi-selective media for the isolation of methicillin-resistant Staphylococcus aureus. J Med Microbiol 1994; 40: 98±101. Plowman R, Graves N, Griffin M et al. SocioEconomic burden of hospital infection. J Hosp Infect 2001; 47: 198±209.