Evidence review for effectiveness of nasal decolonisation in prevention of surgical site infection

Included studies

From a database of 835 studies, 70 studies were identified as being potentially relevant. Following full text review of the 70 studies, 9 RCTs were included which examined the following interventions:

• mupirocin versus placebo nasal ointment
• mupirocin versus no nasal decolonisation
• mupirocin versus 5% povidone iodine
• mupirocin in combination with chlorhexidine body wash versus no treatment (no nasal decolonisation or body wash)
• mupirocin in combination with chlorhexidine body wash versus placebo (placebo ointment in combination with placebo soap)
• nasal chlorhexidine versus placebo.

No studies of relevant study design were identified which examined the effectiveness of chlorhexidine and neomycin cream or octenisan nasal gel. No studies were identified which compared the timing of nasal decolonisation.

The included studies explored a number of different outcomes. Data on overall SSI (irrespective of pathogen) and S. aureus SSI was extracted. Where possible, data on superficial, deep and organ space occupying SSI were extracted. Data on methicillin-resistant Staphylococcus aureus (MRSA) and Methicillin-sensitive Staphylococcus aureus (MSSA) specific SSI were also extracted.

Studies included in this review also examined a number of different surgical procedures. However, sufficient evidence was not identified to conduct sub-group analyses based on wound classification, elective surgery or emergency surgery.

Comparative evidence was not identified in terms of antimicrobial resistance. However the information presented in the studies was discussed with the committee and added to the rational and impact section.

For the search strategy, see appendix C. For clinical evidence study selection flowchart, see appendix D.

Excluded studies

List of papers excluded at full text, with reasons, is given in Appendix K.

Included studies

A literature search was conducted to identify cost–utility analyses comparing nasal decolonisation interventions. Nasal decolonisation may be conditional on the results of a diagnostic test to support decision-making in some settings, and so studies that evaluated ‘screen and treat’ strategies were included. Standard health economic filters were applied to a clinical search, returning a total of 536 citations. Following review of all titles and abstracts, 25 studies were identified as being potentially relevant to this decision problem, and were ordered for full review. After reviewing the full texts, 3 studies were included as economic evidence for nasal decolonisation.

Excluded studies

Studies that were excluded upon full review are listed in Appendix K, including the primary reason for exclusion.

Courville et al. (2012)

Courville et al. (2012) developed a 1-year decision-tree model to evaluate the cost-effectiveness of preoperative nasal mupirocin, given for 5 days, to prevent SSI in people undergoing total hip or knee arthroplasty in the US. The comparators were: (1) preoperative screening cultures for all patients and treat those positive for S. aureus; (2) treat everybody without screening; and (3) do not provide any screening or treatment. Clinical inputs were sourced from a systematic literature review, with an SSI relative risk of 0.61 with mupirocin in S. aureus carriers (26% prevalence). Screening sensitivity and specificity were 0.52 and 0.85, respectively. The authors assumed that an SSI required a full hip or knee revision procedure. An SSI reduced a person’s quality of life utility value by 20%.

The ‘treat all’ mupirocin strategy was found to dominate the restricted treatment strategies in both the hip and knee arthroplasty populations, providing a small gain in quality-adjusted life-years (QALYs) per patient (0.0002 to 0.0005) and lower overall costs (saving £151 to £205) over 1 year. Treating all patients continued to dominate in the majority of univariate sensitivity analyses, unless the cost of SSI revision surgery was low (close to that of the primary arthroplasty procedure), or if the SSI relative risk with mupirocin was 0.99. In this latter scenario the ‘no treatment’ strategy becomes dominant, though it lies outside the range of values that the authors identified in the literature (0.10 to 0.64). The screening strategy was not found to be cost effective in any alternative scenario.

Wassenberg et al. (2011)

Wassenberg et al. (2011) evaluated the cost-effectiveness of preoperative nasal mupirocin with chlorhexidine soap, given for 5 days, to prevent deep SSI in people undergoing joint implant or cardiac surgery in the Netherlands. The comparators were: (1) do not provide any screening or treatment; (2) screen all patients with rapid PCR and culture, and treat those positive for S. aureus; and (3) treat everybody with mupirocin, without screening. Screening was modelled as being highly effective, with sensitivity of 0.97 and specificity of 0.99. Records from 1 hospital during the period 2001 to 2010 were reviewed to quantify hospital costs and mortality associated with a deep SSI, with data from 53 SSIs out of approximately 1,200 procedures per year. Life-expectancy was discounted by 3% per year, but SSI costs were all incurred within 1 year of the primary procedure. The deep SSI relative risk of mupirocin was 0.21 in S. aureus carriers (18% prevalence) (Bode et al., 2010).

The ‘treat all’ mupirocin strategy was found to dominate both restricted treatment strategies. Treating all patients produced an additional 0.010 discounted life-years per patient compared with the screen and treat strategy, and 0.024 more than the no treatment strategy. The total cost saving was estimated to be £41 and £114 per patient, over the screening and no treatment strategies respectively. Sensitivity analysis showed that the ‘treat all’ strategy, without screening, remained cost-saving if the SSI relative risk with mupirocin was worse than the base-case estimate (0.60 instead of 0.21).

Young & Winston (2006)

The US modelling analysis by Young & Winston (2006) was included in the original guideline. The 90-day decision-tree model compared preoperative mupirocin in elective surgery (cardiothoracic, neurologic, gynaecologic and general), given for 5 days according to the following strategies: (1) screen with nasal culture and treat screen-positive individuals; (2) treat all patients without screening; and (3) do not provide any screening or treatment. Clinical inputs were sourced from a systematic literature review, using RCTs where available. The SSI relative risk with mupirocin was 0.49 in S. aureus (23% prevalence). Screening appears to have been assumed to be 100% accurate. Direct healthcare costs were sourced from the literature review where possible, otherwise Medicare charges were used. Patient productivity loss costs were included, however, they were likely to have been an inconsequential component of total costs.

The ‘screen and treat’ mupirocin strategy was found to dominate both of the other strategies. It prevented 86 SSIs and 2 deaths per 10,000 patients compared with no treatment, and treating all patient was not more effective at reducing the incidence of SSI. Its total cost saving compared with the ‘treat all’ strategy was estimated to be $14 (£10) per patient, or $8 (£6) per patient if only hospital costs are included, due to avoiding unnecessary treatment costs. One-way sensitivity analysis found that mupirocin efficacy was the only parameter likely to influence cost-effectiveness results; an SSI relative risk with mupirocin of 0.92 makes the ‘treat all’ strategy incur higher costs than ‘no treatment’, with a cost per life-year gained of $389,782 (£277,000).

Model methods

A decision-tree model was developed, adopting a very similar structure to the model developed for the initial guideline. The model captures the short-term decision about whether to use nasal decontamination. At model entry a patient has just undergone a surgical procedure, from which they are subject to a risk of SSI and mortality. For the purpose of this decision problem, the model focuses on SSIs caused by S. aureus. After the perioperative period, the model applies age-related life expectancy to surviving patients. In this way, the full impact of any SSI-related mortality on health gains are captured. The model takes a patient perspective for outcomes and an NHS and PSS perspective for costs, in line with the NICE manual for guideline development (NICE 2014). Long-term outcomes are discounted at a rate of 3.5% per year.

The model includes 3 comparators: universal decolonisation using mupirocin; decolonisation with mupirocin in people who are screened positive for nasal carriage of S. aureus; and no decolonisation at all (standard care only). Standard care is assumed to capture general infection control measures, such as a chlorhexidine body wash. Mupirocin is the only active intervention included, as this was the focus of almost the entire body of clinical evidence. The screen-and-treat strategy was included as the committee felt that the decision to use nasal decontamination of S. aureus is intrinsically linked with knowledge about whether the person is a carrier of S. aureus or not.

Baseline S. aureus SSI rates were obtained from 2 UK sources: a surveillance study in an English hospital in the base-case analysis (Jenks et al., 2014), and a national registry (PHE, 2017) in a scenario analysis. The committee expressed a strong preference for the hospital study, advising that its higher SSI rates (5.1% overall, compared with 1.3%) are more representative of outcomes in current practice and their own experiences. In this study, 33% of SSIs were caused by S. aureus, compared with 11% in the PHE SSI surveillance service data (although this estimate relates to inpatient-detected SSIs only; for inpatient and readmission-detected SSIs, the proportion is 20%). These data were used to inform baseline S. aureus SSI rates without nasal decontamination, for 17 different types of surgery as well as a pooled ‘all surgery’ cohort. As the screening strategy will provide information about whether a person is a S. aureus carrier or non-carrier, it was necessary to adjust the baseline infection rate to reflect this distinction, as the Jenks and PHE data are in general surgical populations composed of some carriers and some non-carriers. To do so, we used RCT control-arm data to obtain odds ratios for S. aureus SSI incidence in carriers and non-carriers compared with the general surgical population (2.4 and 0.6 respectively; Kalmeijer et al., 2006; Perl et al., 2002).

Treatment effects were informed by pooling the 5 mupirocin RCTs identified in the clinical review that report S. aureus SSI rates in carriers, comparing mupirocin with placebo or no nasal decontamination (Bode et al., 2010; Kalmeijer et al., 2002; Konvalinka et al., 2006; Perl et al., 2002; Tai et al., 2013). The mupirocin pooled odds ratio was found to be 0.47 (0.31 to 0.70) in a fixed-effect analysis, and 0.47 (0.30–0.73) in a random-effects analysis. Excluding an RCT in the highly-specialised Mohs surgery setting (Tai et al., 2013) and another that was unclear about its use of a chlorhexidine wash on the control arm (Kalmeijer et al., 2006) had little impact on these odds ratios (0.50 and 0.53). The mupirocin odds ratios were applied in the model to treated S. aureus carriers, whose prevalence was set to 25% of the surgical population, informed by a UK observational study (den Heijer et al., 2013). On the universal mupirocin model arm, non-carriers treated with mupirocin were assumed to experience no treatment effect on their risk of S. aureus SSI. On the screening arm, only correctly-identified carriers (true-positive results) received the treatment effect. A full diagnostic accuracy literature search was not conducted for this review; therefore the inputs used in the model for the initial guideline were used for these data. In the base-case analysis, a nasal swab and culture with sensitivity of 0.68 and specificity of 0.95 was applied. A scenario analysis applied the superior, but more expensive, polymerase chain reaction (PCR) test (sensitivity 0.98, specificity >0.99).

Resource use inputs include the cost associated with intervention: mupirocin (£4.24; NHS Drug Tariff VIIIA, May 2018) and screening (£10–29 including nurse time; guideline committee and original guideline model). The committee advised that mupirocin is self-administered by the person due to undergo surgery; therefore no administration cost was applied. For SSI costs, data on excess bed days attributable to an SSI across different types of surgery was obtained from the English hospital study (Jenks et al., 2014), and was costed using NHS reference costs 2016–17. The average SSI cost is £3,123, ranging from £823 to £9,056.

Recent evidence suggests there might not be a significant mortality effect attributable to SSI (Badia et al., 2017); however, the committee felt that such an effect is very likely to exist, but that it is difficult to quantify. We estimated a mortality odds ratio associated with SSI compared with no SSI of 1.5 (0.8 to 4.1), based on the UK SSI surveillance data used in the original guideline model (Coello et al., 2005). This was applied to baseline mortality rates by surgery type, from the PHE (2017) data, to estimate separate mortality rates in people who experience an SSI and those who do not. People who survive are assumed to experience typical, age-related life expectancy and quality of life after the surgical period, informed by UK population norms (ONS, 2017; Kind et al., 1999). The impact of an infection on quality of life is informed by a UK EQ-5D study in people undergoing laparotomy (Pinkney et al., 2013), which found that people will an SSI report a bigger utility loss at 7 days (−35% vs. −33%) and 30 days (−17% vs. −6%). We linearly interpolated that it will take people who have an SSI 22 additional days to fully recover to baseline quality of life, compared with 6 additional days for people who do not have an SSI. These effects are applied as a one-off QALY loss.

Model results

Base-case model results, across all types of surgery for a cohort aged 70 and 42% male (PHE, 2017), suggest that universal nasal decolonisation of S. aureus with mupirocin is cost effective, dominating both the screen-and-treat and no treatment strategies (Table ). It produces the highest number of QALYs as it ensures that all S. aureus carriers receive nasal decontamination, and there is no negative health effect caused by treating non-carriers. The cost of treating everybody is more than offset by not screening people and reducing the incidence of SSI. Probabilistic sensitivity analysis from 1,000 model runs indicates that universal mupirocin has a 99.6% probability of being cost-effective when QALYs are valued at £20,000 each.

Providing nasal decontamination only to people who screen positive for S. aureus has a 0% probability of being optimal. One-way sensitivity analysis showed that results are sensitive to baseline S. aureus SSI rates; if much lower infection rates are used, from the PHE (2017) registry, universal mupirocin is no longer cost-effective compared with providing no nasal decontamination (Figure 1). No other individual model parameter or setting – for example, reducing mupirocin efficacy to its 95% confidence interval limit (odds ratio: 0.70), or assuming that mupirocin requires nurse-led administration – affects the base case model result.

In subgroup analysis, the base case result was found to be largely consistent across different types of surgery. The only specialties in which universal mupirocin was not the dominant strategy were breast surgery (ICER vs. standard care: £849 per QALY gained) and cranial surgery (ICER vs. standard care: £13,089 per QALY gained). The respective probabilities of its ICER being £20,000 or better are 76% and 65%. This probability is 67% in spinal surgery, though mupirocin was dominant in the deterministic result. Across all other types of surgery the probability of mupirocin being cost-effective, when a QALY is valued at £20,000, is 89% or higher.

Clinical evidence

Evidence identified in the review explored a number of different surgical procedures and timing of decolonisation. Where possible, evidence statements were constructed to reflect these characteristics. Evidence statements were stratified based on different populations.

Economic evidence

• Three partially applicable economic evaluations with potentially serious limitations compared providing a 5-day course of preoperative intranasal mupirocin, to all patients and to patients screened positive for S. aureus, in various surgical settings, with providing no nasal decolonisation. Two studies found the universal treatment strategy to be more effective and cost-saving compared with the other options, while 1 study found the screen-and-treat strategy to be dominant. Conclusions were largely robust to one-way sensitivity analysis.
• A directly applicable economic model with minor limitations compared the use of mupirocin nasal ointment in all surgical patients with its use only in patients screened positive for S. aureus and with no nasal decolonisation. It found that universal mupirocin dominates other strategies in most types of surgery. Its ICER is better than £20,000 per QALY gained in all types, with likelihoods ranging from 65% to 100%. Results are sensitive to the baseline incidence of SSI; mupirocin is less likely to be cost effective if the underlying risk of SSI is very low.

Interpreting the evidence

Cost effectiveness and resource use

The committee discussed the cost-effectiveness evidence for nasal decolonisation of S. aureus. It was noted that the 3 published studies included all found in favour of nasal decolonisation with mupirocin; however, all 3 were non-UK studies, limiting their applicability to NHS practice. It was agreed that this is particularly important for the estimated costs of treating an SSI, which are likely to be very high in the US studies. The committee agreed that the additional cost attributable to an SSI is likely to be much lower in the NHS. The committee discussed whether the Young & Winston (2006) study provides evidence that a strategy of screening people for S. aureus, and treating only those screen positive, is cost effective compared with giving mupirocin to all patients. It was noted that the study applied 100% diagnostic accuracy of screening, such that there is no health loss associated with failing to identify S. aureus carriers, and that this is therefore not strong evidence with which to recommend a strategy of nasal decolonisation only in S. aureus carriers.

The committee went on to discuss the economic model developed for this guideline, adopting a UK perspective, with resource use and cost inputs from directly relevant data-sources. It agreed that the resource use data, and the use of a UK EQ-5D study to inform the quality of life impact of SSI, made this analysis more applicable to the decision problem than the published studies. The committee agreed that the model should include a ‘screen-and-treat’ strategy, as knowledge about whether the person is a carrier of S. aureus will directly influence the decision regarding whether to provide nasal decolonisation of S. aureus.

The committee discussed the most appropriate way to characterise baseline SSI rates in the model, aware that the 2 main data-sources report disparate SSI rates. It heard that the PHE (2017) data suggest a relatively low underlying risk of SSI, compared with a considerably higher risk of SSI from the English hospital surveillance study (Jenks et al., 2014). The committee also recognised that European registry data appear to suggest SSI incidence rates closer to the PHE data than the Jenks et al. (2014) data. However, the committee expressed a strong preference for using the hospital surveillance data to inform baseline SSI risk, advising that these values are much more representative of clinical experience. The committee highlighted evidence in abdominal surgery trial control arms – which benefit from extensive trial follow-up – suggesting the SSI rate is over 20%, and so the Jenks data might still underestimate baseline SSI incidence (though being more accurate than the PHE data). It advised that a thorough hospital surveillance study is more likely to capture all SSIs across surgical specialties, and that this benefit of the Jenks study offsets its smaller sample size.

The committee reviewed the cost-effectiveness results from the new model, noting that universal nasal mupirocin – alongside standard infection control, including a chlorhexidine body wash – is highly likely to be an efficient use of resources in most surgical specialties. These results would have been even more strongly in favour of universal mupirocin if, based on committee experience, baseline SSI rates are higher than those reported in the Jenks et al. (2014) hospital surveillance data. In this respect the model results may in fact be conservative estimates of the cost effectiveness of mupirocin. The committee noted that the base-case result was largely robust to one-way sensitivity analysis and scenario analysis, including extreme values of S. aureus carriage prevalence. The only parameter of influence was the use of lower baseline SSI rates, from the PHE (2017) registry, instead of the hospital surveillance study. As noted above, the committee felt that the PHE values significantly underestimate the incidence of SSI, and that the Jenks et al. (2014) are likely to be more representative of current NHS outcomes.

The committee discussed whether it is plausible that a universal nasal decolonisation strategy would always provide the highest number of QALYs. The committee accepted that this was plausible, after agreeing that nasal mupirocin is not associated with important negative side effects, and screening measures are subject to imperfect sensitivity.

The committee discussed the results presented by type of surgery, observing that in 15 out of 17 specialties universal mupirocin remained dominant, consistent with the base-case ‘all surgery’ population. Mupirocin did not have an ICER worse than £20,000 per QALY gained in any of the 17 surgical subgroups. The most uncertain results were in cranial and spinal surgery, with 65% and 67% probabilities of mupirocin being cost effective respectively, due to their low underlying SSI risk and, in the case of cranial surgery, a low estimated additional resource use associated with SSI. The committee advised that, based on clinical experience, nasal decolonisation is sometimes currently used in breast surgery practice, which would make the baseline SSI incidence in breast surgery lower than if no nasal decolonisation were used. This may mean the non-dominant mupirocin ICER in breast surgery (£849 per QALY gained) is a conservative estimate. The committee also noted that the screen-and-treat strategy is very unlikely to be the optimal choice, even if the superior PCR test is used. This is consistent with the 2 published economic evaluations that assumed imperfect diagnostic accuracy.

The committee discussed the relative effectiveness evidence used in the model, which came from 5 pooled mupirocin RCTs which reported S. aureus SSI rates in S. aureus carriers. The committee advised that Mohs surgery is a highly specialised setting that is less representative of general surgical practice. It saw that excluding this trial (Tai et al., 2013) from the pooled measure of effect did not influence cost-effectiveness conclusions. The committee noted that most of the clinical evidence included in this review, which had been separated in different ways, did not indicate a significant benefit associated with mupirocin, and agreed that this warrants a cautious approach to interpreting model results. Given this, the committee recommended that nasal decolonisation with mupirocin is considered, alongside a chlorhexidine body wash, replacing a previous recommendation advising against its use.

Other factors the committee took into account

The secondary aim in this evidence review was to examine the timing of nasal decolonisation for the prevention of SSI. The committee noted that typically, mupirocin and chlorhexidine bundle is applied 2 days prior to surgery to 3 days after surgery. However no evidence was identified which explored timing of nasal decolonisation. Furthermore, timing of nasal decolonisation in the included studies ranged from five days before surgery to the day before surgery. Due to the lack of evidence, the committee were unable to comment on when mupirocin and chlorhexidine body wash should be administered. The committee did identify this as an important area which required further research as timing of decolonisation can vary and made a research recommendation to reflect this. Furthermore, as repetitive and prolonged use of antibiotics and antiseptics is not advised when taking into consideration antimicrobial resistance, understanding timing of decolonisation is crucial.

The PICO in Table 1 outlines that this review examined the effectiveness of a number of different interventions including chlorhexidine and neomycin cream (Naseptin) and octenisan nasal gel. However, no studies of relevant study design were identified which examined the effectiveness of these interventions. Therefore the committee were unable to make recommendations on the use of these interventions. The committee did consider these interventions in the three additional research recommendations made.

In this review evidence on people who had been identified as S. aureus carriers through screening and the whole population (in whom screening was not conducted) was identified. However, studies examining the combined use of mupirocin with chlorhexidine body wash only included people who had been identified as S. aureus carriers through screening. Due to the lack of information on the whole population, the committee identified this as an important area of further research.

The committee further noted that studies included in the review did not provide data on children. Due to the lack of evidence in this population, specific recommendations could not be made for this population group. Additionally, the committee identified risks associated with the use of chlorhexidine in preterm neonates. As data on alternative body washes was not identified, no recommendations could be made, however the committee identified this as an important area for research and made a research recommendation to examine the effectiveness of other nasal decolonisation protocols. It was also noted that children may find it difficult to tolerate nasal decolonisation. However, the recommendations state that use of nasal mupirocin should be used taking into account surgical procedure, patient risk factors and impact of infection. Therefore, clinicians should take age into their decision making process.

The committee also discussed that as decolonisation with mupirocin and chlorhexidine takes place prior to surgery, people may be required to self-administer the treatment. While some manufacturers may provide a guide to aid people with the procedure, people with learning disabilities, English as their second language or the elderly who live alone may find the application of the intervention difficult. However, it was noted that in some centres, people may be invited to the centres a day before surgery to assist in washing. Chlorohexidine wipes are also available however these are more expensive so their use is subject to funding with some NHS trusts being unable to fund their use.

Quality assessment

Individual systematic reviews were quality assessed using the ROBIS tool, with each classified into one of the following three groups:

• High quality – It is unlikely that additional relevant and important data would be identified from primary studies compared to that reported in the review, and unlikely that any relevant and important studies have been missed by the review.
• Moderate quality – It is possible that additional relevant and important data would be identified from primary studies compared to that reported in the review, but unlikely that any relevant and important studies have been missed by the review.
• Low quality – It is possible that relevant and important studies have been missed by the review.

Each individual systematic review was also classified into one of three groups for its applicability as a source of data, based on how closely the review matches the specified review protocol in the guideline. Studies were rated as follows:

• Fully applicable – The identified review fully covers the review protocol in the guideline.
• Partially applicable – The identified review fully covers a discrete subsection of the review protocol in the guideline.
• Not applicable – The identified review, despite including studies relevant to the review question, does not fully cover any discrete subsection of the review protocol in the guideline.

Using systematic reviews as a source of data

If systematic reviews were identified as being sufficiently applicable and high quality, and were identified sufficiently early in the review process (for example, from the surveillance review or early in the database search), they were used as the primary source of data, rather than extracting information from primary studies. The extent to which this was done depended on the quality and applicability of the review, as defined in Table 4. When systematic reviews were used as a source of primary data, any unpublished or additional data included in the review which is not in the primary studies was also included. Data from these systematic reviews was then quality assessed and presented in GRADE tables as described below, in the same way as if data had been extracted from primary studies. In questions where data was extracted from both systematic reviews and primary studies, these were cross-referenced to ensure none of the data had been double counted through this process.

Quality assessment

Individual RCTs were quality assessed using the Cochrane Risk of Bias Tool. Other study were quality assessed using the ROBINS-I tool. Each individual study was classified into one of the following three groups:

• Low risk of bias – The true effect size for the study is likely to be close to the estimated effect size.
• Moderate risk of bias – There is a possibility the true effect size for the study is substantially different to the estimated effect size.
• High risk of bias – It is likely the true effect size for the study is substantially different to the estimated effect size.

Each individual study was also classified into one of three groups for directness, based on if there were concerns about the population, intervention, comparator and/or outcomes in the study and how directly these variables could address the specified review question. Studies were rated as follows:

• Direct – No important deviations from the protocol in population, intervention, comparator and/or outcomes.
• Partially indirect – Important deviations from the protocol in one of the population, intervention, comparator and/or outcomes.
• Indirect – Important deviations from the protocol in at least two of the following areas: population, intervention, comparator and/or outcomes.

Methods for combining intervention evidence

Meta-analyses of interventional data were conducted with reference to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al. 2011).

Where different studies presented continuous data measuring the same outcome but using different numerical scales (e.g. a 0–10 and a 0–100 visual analogue scale), these outcomes were all converted to the same scale before meta-analysis was conducted on the mean differences. Where outcomes measured the same underlying construct but used different instruments/metrics, data were analysed using standardised mean differences (Hedges’ g).

A pooled relative risk was calculated for dichotomous outcomes (using the Mantel–Haenszel method). Both relative and absolute risks were presented, with absolute risks calculated by applying the relative risk to the pooled risk in the comparator arm of the meta-analysis.

Fixed- and random-effects models (der Simonian and Laird) were fitted where appropriate, with the presented analysis dependent on the degree of heterogeneity in the assembled evidence. Fixed-effects models were the preferred choice to report, but in situations where the assumption of a shared mean for fixed-effects model were clearly not met, even after appropriate pre-specified subgroup analyses were conducted, random-effects results are presented. Fixed-effects models were deemed to be inappropriate if one or both of the following conditions was met:

• Significant between study heterogeneity in methodology, population, intervention or comparator was identified by the reviewer in advance of data analysis. This decision was made and recorded before any data analysis was undertaken.
• The presence of significant statistical heterogeneity in the meta-analysis, defined as I²≥50%.

In any meta-analyses where some (but not all) of the data came from studies at high risk of bias, a sensitivity analysis was conducted, excluding those studies from the analysis. Results from both the full and restricted meta-analyses are reported. Similarly, in any meta-analyses where some (but not all) of the data came from indirect studies, a sensitivity analysis was conducted, excluding those studies from the analysis.

Meta-analyses were performed in Cochrane Review Manager v5.3.

Minimal clinically important differences (MIDs)

The Core Outcome Measures in Effectiveness Trials (COMET) database was searched to identify published minimal clinically important difference thresholds relevant to this guideline. Identified MIDs were assessed to ensure they had been developed and validated in a methodologically rigorous way, and were applicable to the populations, interventions and outcomes specified in this guideline. In addition, the Guideline Committee were asked to prospectively specify any outcomes where they felt a consensus MID could be defined from their experience. In particular, any questions looking to evaluate non-inferiority (that one treatment is not meaningfully worse than another) required an MID to be defined to act as a non-inferiority margin.

No MIDs were identified. Therefore, a default MID interval for dichotomous outcomes of 0.8 to 1.25 was used.

When decisions were made in situations where MIDs were not available, the ‘Evidence to Recommendations’ section of that review should make explicit the committee’s view of the expected clinical importance and relevance of the findings. In particular, this includes consideration of whether the whole effect of a treatment (which may be felt across multiple independent outcome domains) would be likely to be clinically meaningful, rather than simply whether each individual sub outcome might be meaningful in isolation.

GRADE for pairwise meta-analyses of interventional evidence

GRADE was used to assess the quality of evidence for the selected outcomes as specified in ‘Developing NICE guidelines: the manual (2014)’. Data from all study designs was initially rated as high quality and the quality of the evidence for each outcome was downgraded or not from this initial point, based on the criteria given in Table 5.

The quality of evidence for each outcome was upgraded if any of the following three conditions were met:

• Data from non-randomised studies showing an effect size sufficiently large that it cannot be explained by confounding alone.
• Data showing a dose-response gradient.
• Data where all plausible residual confounding is likely to increase our confidence in the effect estimate.

Publication bias

Publication bias was assessed in two ways. First, if evidence of conducted but unpublished studies was identified during the review (e.g. conference abstracts, trial protocols or trial records without accompanying published data), available information on these unpublished studies was reported as part of the review. Secondly, where 10 or more studies were included as part of a single meta-analysis, a funnel plot was produced to graphically assess the potential for publication bias.

Evidence statements

Evidence statements for pairwise intervention data are classified in to one of four categories:

• Situations where the data are only consistent, at a 95% confidence level, with an effect in one direction (i.e. one that is ‘statistically significant’), and the magnitude of that effect is most likely to meet or exceed the MID (i.e. the point estimate is not in the zone of equivalence). In such cases, we state that the evidence showed that there is an effect.
• Situations where the data are only consistent, at a 95% confidence level, with an effect in one direction (i.e. one that is ‘statistically significant’), but the magnitude of that effect is most likely to be less than the MID (i.e. the point estimate is in the zone of equivalence). In such cases, we state that the evidence could not demonstrate a meaningful difference.
• Situations where the data are consistent, at a 95% confidence level, with an effect in either direction (i.e. one that is not ‘statistically significant’) but the confidence limits are smaller than the MIDs in both directions. In such cases, we state that the evidence demonstrates that there is no difference.
• In all other cases, we state that the evidence could not differentiate between the comparators.

For outcomes without a defined MID or where the MID is set as the line of no effect, evidence statements are divided into 2 groups as follows:

• We state that the evidence showed that there is an effect if the 95% CI does not cross the line of no effect.
• The evidence could not differentiate between comparators if the 95% CI crosses the line of no effect.

Economic evaluations

1. Economics/
2. exp “Costs and Cost Analysis”/
3. Economics, Dental/
4. exp Economics, Hospital/
5. exp Economics, Medical/
6. Economics, Nursing/
7. Economics, Pharmaceutical/
8. Budgets/
9. exp Models, Economic/
10. Markov Chains/
11. Monte Carlo Method/
12. Decision Trees/
13. econom$.tw.
14. cba.tw.
15. cea.tw.
16. cua.tw.
17. markov$.tw.
18. (monte adj carlo).tw.
19. (decision adj3 (tree$ or analys$)).tw.
20. (cost or costs or costing$ or costly or costed).tw.
21. (price$ or pricing$).tw.
22. budget$.tw.
23. expenditure$.tw.
24. (value adj3 (money or monetary)).tw.
25. (pharmacoeconomic$ or (pharmaco adj economic$)).tw.
26. or/1–25

Quality of Life

1. “Quality of Life”/
2. quality of life.tw.
3. “Value of Life”/
4. Quality-Adjusted Life Years/
5. quality adjusted life.tw.
6. (qaly$ or qald$ or qale$ or qtime$).tw.
7. disability adjusted life.tw.
8. daly$.tw.
9. Health Status Indicators/
10. (sf36 or sf 36 or short form 36 or shortform 36 or sf thirtysix or sf thirty six or shortform thirtysix or shortform thirty six or short form thirtysix or short form thirty six).tw.
11. (sf6 or sf 6 or short form 6 or shortform 6 or sf six or sfsix or shortform six or short form six).tw.
12. (sf12 or sf 12 or short form 12 or shortform 12 or sf twelve or sftwelve or shortform twelve or short form twelve).tw.
13. (sf16 or sf 16 or short form 16 or shortform 16 or sf sixteen or sfsixteen or shortform sixteen or short form sixteen).tw.
14. (sf20 or sf 20 or short form 20 or shortform 20 or sf twenty or sftwenty or shortform twenty or short form twenty).tw.
15. (euroqol or euro qol or eq5d or eq 5d).tw.
16. (qol or hql or hqol or hrqol).tw.
17. (hye or hyes).tw.
18. health$ year$ equivalent$.tw.
19. utilit$.tw.
20. (hui or hui1 or hui2 or hui3).tw.
21. disutili$.tw.
22. rosser.tw.
23. quality of wellbeing.tw.
24. quality of well-being.tw.
25. qwb.tw.
26. willingness to pay.tw.
27. standard gamble$.tw.
28. time trade off.tw.
29. time tradeoff.tw.
30. tto.tw.
31. or/1–30

Outcomes in whole population

Overall SSI

Follow up: 30 days

Overall superficial SSI

Follow up: 30 days

Overall deep SSI

Follow up: 30 days

S. aureus SSI

Follow up: 30 days

Overall nosocomial infections

Follow up: 30 days

S. aureus nosocomial infections

Follow up: 30 days

Hospital readmission

Follow up: 30 days

Mean hospital stay

Follow up: 30 days

Outcomes in S. aureus Carriers

Overall SSI

Overall superficial SSI

Follow up: Within 8 weeks

Overall deep SSI

Follow up: Within 8 weeks

Overall deep space occupying SSI

Follow up: Within 8 weeks

S. aureus SSI

Overall nosocomial infections

S. aureus nosocomial infections

Follow up: 30 days

Mortality

Follow up: Within 8 weeks

Outcomes in whole population

Overall SSI

Follow up: 30 days

Overall superficial SSI

Follow up: 30 days

Overall deep SSI

Follow up: 30 days

S. aureus SSI

Follow up: 30 days

Overall nosocomial infections

Follow up: 30 days

Outcomes in whole population

Overall deep SSI

Follow up: within 3 months

S. aureus deep SSI

Follow up: within 3 months

MRSA deep SSI

Follow up: within 3 months

MSSA deep SSI

Follow up: within 3 month

S. aureus carriers

Overall deep SSI (PJI)

Follow up: at 1 year

S. aureus deep SSI (PJI)

Follow up: at 1 year

S. aureus SSI

Follow up: postoperative period

MRSA SSI

Follow up: postoperative period

MSSA SSI

Follow up: postoperative period

S. aureus carriers

S. aureus SSI

Follow up: until 6 weeks after discharge

S. aureus superficial SSI

Follow up: until 6 weeks after discharge

S. aureus deep SSI

Follow up: until 6 weeks after discharge

S. aureus nosocomial infections

Follow up: until 6 weeks after discharge

Mortality

Follow up: until 6 weeks after discharge

Mortality in carriers with S. aureus infections

Follow up: until 6 weeks after discharge

S. aureus carriers

S. aureus SSI

Whole population

Overall SSI

Follow up: at 30 days

Overall deep SSI

Follow up: at 30 days

S. aureus SSI

Follow up: at 30 days

Overall nosocomial infection

Follow up: at 30 days

Nosocomial infection: Lower respiratory tract infection (LRTI)

Follow up: at 30 days

Nosocomial infection: urinary tract infection (UTI)

Follow up: at 30 days

Nosocomial infection: bacteraemia

Follow up: at 30 days

Mortality

Follow up: at 30 day

Mean hospital stay

Follow up: at 30 days

Hospital readmission

Follow up: at 30 days

Outcomes in whole population

Outcomes in S. aureus carriers

Outcomes in whole population

Outcomes in whole population

Outcomes in S. aureus carriers

Outcomes in S. aureus carriers

Outcomes in S. aureus carriers

In whole population