NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Collaborating Centre for Women's and Children's Health (UK). Surgical Site Infection: Prevention and Treatment of Surgical Site Infection. London: RCOG Press; 2008 Oct. (NICE Clinical Guidelines, No. 74.)

Cover of Surgical Site Infection

Surgical Site Infection: Prevention and Treatment of Surgical Site Infection.

Show details

Appendix ECost-effectiveness of mupirocin nasal ointment to prevent surgical site infection caused by Staphylococcus aureus

E.1. Literature review

Two full economic analyses50,51 were identified.

A cost-effectiveness analysis conducted in the Netherlands50 compared mupirocin calcium ointment treatment with no preventative treatment in cardiothoracic surgery patients. This analysis was based on a study of 1796 patients using a historical control. The analysis was conducted from the perspective of the healthcare system (only including costs to the healthcare system), with the time frame for the analysis not stated. The outcome used was cost per SSI prevented. The authors reported that treating 1000 surgical patients with mupirocin would lead to a cost saving of $747,969, which was $16,633 saved per SSI prevented. The incidence of SSIs was 7.3% in the historical control and 2.8% with the intervention. Mupirocin led to a 62% reduction in SSIs, which was calculated to prevent 45 SSIs per 1000 patients undergoing surgery. Sensitivity analyses were carried out on the incidence of SSIs (1% to 100%), effectiveness of mupirocin (1% to 100%), SSI-attributable costs (0% to 200%) and cost of mupirocin treatment ($0 to $1000). Mupirocin treatment remained cost saving except when SSI-attributable costs dropped below $245 per patient with an SSI. No staff costs were considered for the application of mupirocin, which would make using mupirocin ointment more expensive.

A US cost-effectiveness analysis compared the following strategies:51

  1. screening patients for S. aureus colonisation with nasal culture and treating carriers with mupirocin
  2. screening no patients and treating all with mupirocin
  3. no screening and no preventative treatment.

The patient group in this analysis had multiple coexisting illnesses and underwent non-emergency cardiothoracic, neurologic, general and gynaecological surgery. The outcomes of the analysis were cost per infection avoided and cost per life year saved. The analysis was based on one large RCT for mupirocin effectiveness in 3864 surgical patients and was conducted from the perspective of society, including patient expenses as well as the costs to the healthcare system. The time frame for the analysis was 90 days. The study concluded that both mupirocin strategies were cost saving: $102 per patient undergoing surgery in the screen and treat strategy, and $88 per patient in the treat-all strategy. Mupirocin led to a 51% reduction in SSIs. If mupirocin efficacy was less than 16.1% effective, then the screen and treat strategy was no longer cost saving. If S. aureus carriage rate was greater than 42.7%, then the treat-all strategy was more cost-effective.

As neither published analysis was conducted in the UK, a new model was developed for the purpose of this guideline.

E.2. The decision tree model

A simple decision-analytic model was developed in Microsoft Excel® (see Figure E.1) to assess the cost-effectiveness of preventing SSI caused by S. aureus using mupirocin nasal ointment. Costing was calculated from the perspective of the NHS and the analysis considered a time frame of 1 year, meaning that no discounting of costs or benefits was undertaken.

Figure E.1. Decision tree for the three treatment strategies.

Figure E.1

Decision tree for the three treatment strategies.

The model compared the following three strategies:

  1. no nasal decontamination
  2. treat all patients with mupirocin
  3. screen all patients and treat patients identified as S. aureus carriers.

The analysis was based on a modelling exercise carried out in the USA51 where the population was men and women, mean age 54 years, with multiple coexisting illness who underwent non-emergency cardiothoracic, neurological, general and gynaecological surgery. The model was not applicable to orthopaedic patients or patients with few comorbidities undergoing low-risk procedures. This model looked at all healthcare-associated infections caused by S. aureus and other pathogens, including pneumonia and bacteraemia. As the scope for this guideline is SSIs, the model has been simplified to consider only these infections. This may underestimate the benefits of using mupirocin as cases of pneumonia and bacteraemia may be reduced owing to mupirocin use.

The clinical evidence (see Section 5.6) showed no statistically significant difference in the rate of SSI overall in all patients treated with mupirocin compared with placebo. In S. aureus carriers there was a reduction in SSIs caused by S. aureus when mupirocin was used, although this reduction did not achieve statistical significance at the 5% level. This model does not take into account antibiotic resistance to S. aureus, which would require a more complex model to be developed.

E.2.1. Model inputs

Table E.1Probabilities used in the mupirocin decision tree model

Prevalence of Staphylococcus aureus nasal colonisation0.230.190.55Young (2006)51
Screening sensitivity0.960.6820.98Ritchie (2007)232The base case for this model used the sensitivity and specificity for detecting MRSA, a conservative assumption
Screening specificity0.950.9450.998Ritchie (2007)232
Mortality with SSI0.0660.0570.076Coello (2005)4See hair removal model in Appendix D
Mortality without SSI0.0260.0250.027Coello (2005)4See hair removal model in Appendix D
No treatment – S. aureus carrier:
 S. aureus infection0.059Perl (2002)46Beta distribution used for PSA
 Other SSI0.058Perl (2002)46Beta distribution used for PSA
No treatment – non-carrier:
 S. aureus infection0.014Perl (2002)46Beta distribution used for PSA
 Other SSI0.062Perl (2002)46Beta distribution used for PSA
MupirocinS. aureus carrier:
 S. aureus infection0.029Perl (2002)46Beta distribution used for PSA
 Other SSI0.060Perl (2002)46Beta distribution used for PSA
Mupirocin – non-carrier:
 S. aureus infection0.019Perl (2002)46Beta distribution used for PSA
 Other SSI0.056Perl (2002)46Beta distribution used for PSA

MRSA = meticillin-resistant Staphylococcus aureus; PSA = probabilistic sensitivity analysis

Table E.2Utility values used in the mupirocin decision tree model

Patients with SSI0.570.510.64See hair removal model in Appendix D
Patients with no SSI0.640.570.71See hair removal model in Appendix D

Table E.3Costs used in the mupirocin decision tree model

Resource itemCostMinimumMaximumSourceNotes
Real-time PCR swab£5.18£7.45£19.40GDGTraditional culture nasal swab, 24–48 hours, full cost plus overheads
Time to take swab£2.55£1.28£3.83Ritchie (2007)232Costs associated with taking patient samples:
  • staff nurse (grades D–G) spending approximately 5 minutes
  • providing information to patient, taking two swab samples
  • completing related administration such as labelling samples and sending them to the lab
1 hour of nurse time£22.00£16.50£27.50Curtis (2006)233Nurse, day ward, cost per hour including qualifications
Nurse time – mupirocin (application only)£9.17Ritchie (2007)232Assuming 5 minutes per application for 5 days
Mupirocin£5.80£2.90£8.70BNF 54107Bactroban® Nasal (GSK) 3 g 0.2% mupirocin per patient
Bed day due to SSI£307£230£383Coello (2005)4
SSI treatment per patient£3,486£2,168£5,235Coello (2005)4Assuming 11.37 (minimum 9.43 days, maximum 13.66 days) additional hospital days to treat SSI

BNF = British National Formulary; PCR = polymerase chain reaction

E.2.2. Results

As is shown in Tables E.4 and E.5, treating all patients with mupirocin is the dominant strategy resulting in the least number of SSIs and the lowest cost. In the model, application of mupirocin has low costs, with five applications taking 25 minutes of a nurse’s time (£9.17) plus the cost of the ointment (£5.80). The screening is also relatively low cost, at £2.55 for the nurse’s time and £5.18 for the screening itself, but this is still higher than the cost of applying the mupirocin. However, it is because of the high ‘downstream’ costs of treating SSI that the most efficacious strategy is also the cheapest.

Table E.4. Cost per SSI prevented.

Table E.4

Cost per SSI prevented.

Table E.5. Cost per QALY.

Table E.5

Cost per QALY.

E.2.3. Sensitivity analysis

Considerable uncertainty surrounds the data inputs of the model and therefore one-way sensitivity analysis was used to assess how robust the baseline conclusions would be given different assumptions. In particular, the clinical evidence would not cause a null hypothesis that mupirocin conferred no benefit in terms of reduced SSI to be rejected at the 5% level.

Table E.6 shows the effect of assuming that mupirocin does not lead to any changes in SSI.

Table E.6. Sensitivity analysis showing cost and QALY with no treatment effect.

Table E.6

Sensitivity analysis showing cost and QALY with no treatment effect.

A sensitivity analysis with a lower SSI treatment cost is shown in Table E.7. This is an important driver of the conclusions in the baseline analysis as it is this that causes treatment to be cost saving relative to no treatment.

Table E.7. Sensitivity analysis showing incremental cost per QALY with a lower SSI treatment cost (£2,168).

Table E.7

Sensitivity analysis showing incremental cost per QALY with a lower SSI treatment cost (£2,168).

A threshold analysis showed that the cost of treating an SSI would have to fall to below £600 before the ICER for the treat all patients with mupirocin strategy exceeded £20,000 per QALY, the threshold used by NICE to determine cost-effectiveness.

In this model, there is uncertainty over more than one parameter value. One technique to address this is multi-way sensitivity analysis where a number of parameter values are varied from their baseline value simultaneously. However, in a model with many parameter values, the number of possible permutations to test can be daunting. So, instead, a probabilistic sensitivity analysis (PSA) was undertaken using Monte Carlo simulation, which is an alternative way of addressing uncertainty across many parameter values simultaneously. In the baseline deterministic model, the results are determined by the point estimates entered as parameter values. However, the point estimates of the SSI rate in different patients and with different treatments are based on a sample of patients who participated in a particular study. If that study was well designed these point estimates provide the best estimate of the true SSI rate but they are nevertheless subject to sampling error. In PSA, the parameters are made probabilistic, which involves specifying a distribution around that point estimate. A simulation exercise is then undertaken that involves ‘running’ the model many times. In each ‘run’, the parameter values are sampled from the probability distribution, which means that the model output varies on each run while still being informed by the best estimates from the evidence. It is by sampling from the probability distribution that the inherent uncertainty in the data is handled.

In this PSA for this model, only the SSI rates have been made probabilistic. In other words, the costs, the prevalence of S. aureus carriers, the accuracy of screening and the utility associated with states with and without an SSI do not change. However, to reflect the importance of treatment costs of SSI to the model, two Monte Carlo simulations were undertaken, one with treatment costs for SSI at their baseline level (£3,486) and one with a lower treatment cost for SSI (£2,168).

Each Monte Carlo simulation consisted of running the model 1000 times. For each run, the strategy that is the most cost-effective is recorded. This is straightforward where a strategy is the cheapest and most effective. However, when a strategy is more effective and more costly then its cost-effectiveness will depend on the willingness to pay for a QALY. NICE uses a willingness-to-pay threshold of £20,000 per QALY (with interventions with an ICER of less than this considered cost-effective). However, the model calculates for each run which would be the most cost-effective strategy at a range of willingness-to-pay thresholds.

The results of the PSA are shown in Figures E.2 and E.3.

Figure E.2. Cost-effectiveness acceptability curve with the cost of treating SSI = £3,486.

Figure E.2

Cost-effectiveness acceptability curve with the cost of treating SSI = £3,486.

Figure E.3. Cost-effectiveness acceptability curve with the cost of treating SSI = £2,168.

Figure E.3

Cost-effectiveness acceptability curve with the cost of treating SSI = £2,168.

E.3. Discussion

The results with the baseline analysis suggest that treating all patients with mupirocin is a cost-effective strategy. This is driven by the model inputs that assume that mupirocin does confer benefits in terms of reduced SSI and that the initial costs of treatment are offset to some extent by reduced ‘downstream’ costs of SSI treatment. Sensitivity analysis suggested that as long as treating SSI infections incurs a cost per patient of greater than £6,000, treating all patients with mupirocin would remain a good use of scarce NHS resources.

However, there are a number of caveats that need to be borne in mind when interpreting this analysis. The cost-effectiveness of mupirocin is driven by the point estimates derived from just one trial and, although SSI rates are lower with mupirocin, the difference is not statistically significant at the 5% level. Clearly, if the results are a chance finding then mupirocin will not be cost-effective. Both PSA analyses suggest that there is about a 50% chance that treating all patients with mupirocin is the most cost-effective strategy at a £20,000 per QALY willingness-to-pay threshold. In fact, treating all patients with mupirocin is usually the most cost-effective strategy regardless of the willingness-to-pay threshold. The only exception to this was in the analysis where a lower cost of treating an SSI was assumed, where no treatment was more likely to be cost-effective if the willingness to pay for a QALY was £800 or less.

Treating all patients with mupirocin carries a potential harm in that it may increase antibiotic resistance, which has public health implications and costs in the longer term. This analysis does not model any impact on increased antibiotic resistance but it may be that, even if there were genuine benefits in treating all patients with mupirocin in terms of reduced SSI, these would be outweighed by the downside of increased antibiotic resistance. It might very reasonably be decided that, although the PSA suggests that treating all patients with mupirocin is more likely to be cost-effective than the other strategies, the probability of it being so is too small given the harms and risks that have not been incorporated into the model.

In the review of the clinical evidence (Section 5.6), two studies were included and the evidence pooled in a meta-analysis. This meta-analysis did not form the basis of the point estimates entered into the model because it compared all SSIs whereas the trial data used in the model allowed SSIs to be broken down into S. aureus and non-S. aureus infections. It should be noted that, in terms of all SSIs, the point estimates of these studies contradict each other. However, both results were not statistically significant at the 5% level and are thus consistent with no treatment effect, as the forest plot of the meta-analysis suggests (Figure E.4).

Figure E.4. All infections in S. aureus carriers: mupirocin compared with placebo.

Figure E.4

All infections in S. aureus carriers: mupirocin compared with placebo.

Nevertheless, some caution may also be required in interpreting this meta-analysis. It is likely that mupirocin would only be effective in preventing S. aureus infections in S. aureus carriers. Therefore, by including all SSI infections in the analysis, any treatment effect will be diluted and the ‘noise’ will lead to wider confidence intervals. Indeed, in the trial that informed the point estimates, the effect size was closer to being statistically significant (although still not) in a comparison of S. aureus infections in S. aureus carriers. Another of the potential harms of treating all patients with mupirocin, in addition to the possible impact on antibiotic resistance, is that it may increase the patient’s susceptibility to non-S. aureus infections. In the study that informed the model, there was no evidence to support this, with non-S. aureus SSI virtually identical. However, it should be noted that the other included paper might be considered to show evidence, albeit weak, of such an effect.

Further research may be required to establish whether mupirocin does indeed reduce S. aureus SSI in S. aureus carriers and whether this is achieved at the expense of more non-S. aureus infections and/or antibiotic resistance.

Copyright © 2008, National Collaborating Centre for Women’s and Children’s Health.

No part of this publication may be reproduced, stored or transmitted in any form or by any means, without the prior written permission of the publisher or, in the case of reprographic reproduction, in accordance with the terms of licences issued by the Copyright Licensing Agency in the UK []. Enquiries concerning reproduction outside the terms stated here should be sent to the publisher at the UK address printed on this page.

The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and therefore for general use.

Bookshelf ID: NBK53728
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (5.7M)
  • Disable Glossary Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...