Systematic review of the cost‐effectiveness of preoperative antibiotic prophylaxis in reducing surgical‐site infection

Background Surgical‐site infections (SSIs) increase the length of hospital admission and costs. SSI prevention guidelines include preoperative antibiotic prophylaxis. This review assessed the reporting quality and cost‐effectiveness of preoperative antibiotics used to prevent SSI. Methods PubMed, Web of Science, Cumulative Index to Nursing and Allied Health Literature, Index of Economic Articles (EconLit), Database of Abstracts of Reviews of Effect (including the National Health Service Economic Evaluation Database) and Cochrane Central databases were searched systematically from 1970 to 2017 for articles that included costs, preoperative antibiotic prophylaxis and SSI. Included were RCTs and quasi‐experimental studies conducted in Organisation for Economic Co‐operation and Development countries with participants aged at least 18 years and published in English. Two reviewers assessed eligibility, with inter‐rater reliability determined by Cohen's κ statistic. The Consolidated Health Economic Evaluation and Reporting Standards (CHEERS) and modified Drummond checklists were used to assess reporting and economic quality. Study outcomes and characteristics were extracted, and incremental cost‐effectiveness ratios were calculated, with costs adjusted to euros (2016) (€1 = US $1·25; £1 sterling = €1·28). Results Twelve studies published between 1988 and 2014 were included from 646 records identified; nine were RCTs, two were nested within RCTs and one was a retrospective chart review. Study quality was highest in the nested studies. Cephalosporins (first, second and third generation) were the most frequent prophylactic interventions. Eleven studies demonstrated clinically effective interventions; ten were cost‐effective (the intervention was dominant); in one the intervention was dominated by the control; and in one the intervention was more effective and more expensive than the control. Conclusion Preoperative antibiotic prophylaxis does reduce SSI, costs to hospitals and health providers, but the reporting of economic methods in RCTs is not standardized. Routinely nesting economic methods in RCTs would improve economic evaluations and ensure appropriate selection of prophylactic antibiotics.


Introduction
Surgical-site infections (SSIs) occur in 1-25 per cent of surgical patients, although the occurrence and severity vary 1 -3 . These variations depend on the type, duration and time of day of the operation, and the time from infection onset to detection and successful treatment 1,3 -6 . SSI leads to longer hospital stays and higher costs to patients, hospitals and health systems 7 -11 . In Europe, a minimum estimate of increased health cost due to SSI in 2004 was €1⋅47-19⋅1 billion 12 , and more recently in the USA (2014) SSI was associated with double the costs compared with those for a patient without SSI 13 .
Jointly, the Centers for Disease Control and Prevention (CDC) in the USA, the National Institute for Health and Care Excellence in the UK and the World Health Organization developed SSI prevention guidelines 4 . These include several prevention measures: preoperative screening of patients and decolonization of nasal cavities, showering, hair removal, intraoperative skin preparation using chlorhexidine, preoperative prophylactic antibiotic administration (within 1 h before surgery), normothermia and body temperature regulation, use of incision drapes, administration of supplemental oxygen throughout the operation, control of the patient's glucose level, and postoperative use of surgical dressings and appropriate hand hygiene. The prevention measures may be implemented individually or as a bundle (3)(4)(5) interventions are grouped together). Several systematic reviews have reported on aseptic skin preparation (including surgical hand asepsis, intraoperative skin antisepsis and skin preparation with chlorhexidine) 14 -16 , dressings including wound edge protection devices 16,17 , increased oxygen supplementation 18 , glucose control 19 and thermoregulation 20 . T wo reviews have reported on the cost-effectiveness of the interventions 14,16 and the quality of health economic reporting 16 .
Despite the routine use of antibiotic prophylaxis, which is inexpensive 21 -23 , SSIs continue to occur. This suggests that implementation of SSI prevention is suboptimal -that more can be done, and done cost-effectively. To date, no cost-effectiveness review of preoperative antibiotic prophylaxis has been performed, despite the existence of clinical guidelines for antibiotic prophylaxis in surgery 21 -23 . The aim of this review was to evaluate the cost-effectiveness of preoperative antibiotic prophylaxis used to prevent SSIs, and to assess the reporting quality of clinical effectiveness and cost-effectiveness for each study.

Search strategy
Keywords and search terms were matched with database-specific medical subject heading (MeSH) terms or title fields. Keywords for four different themes were linked with AND (cost AND prophylaxis AND prevention AND surgical-site infection). Full search strategies can be found in Table S1 (supporting information). Search results were exported into EndNote ® version X7 (Thomson Reuters, New York, USA) and duplicates were removed. Manual screening of references from included articles was performed to identify additional publications not identified by the search.

Selection criteria
Systematic reviews, guidelines, conference proceedings and letters were excluded. Only articles published in English and in peer-reviewed journals were included. The studies had to define a SSI, even if it did not conform to the CDC definition 4 : an infection related to an operative procedure that occurs at or near the surgical incision within 30 days of the procedure or within 1 year if an implant is left in place. PICO (population, intervention, comparison and outcomes) were used to evaluate study eligibility. Studies were included if they were economic evaluations in RCTs or quasi-experimental studies that compared the efficacy between different antibiotic prophylaxis regimens or placebo. Economic evaluations were defined as the comparative analysis of the costs and consequences of alternative programmes 27 . Studies were excluded if they were performed in non-OECD (Organisation for Economic Co-operation and Development) countries. OECD countries were defined as high-income-earning economies 28 , and included 31 OECD members (Table S2, supporting information). Other exclusion criteria were: study participants younger than 18 years of age and surgery that did not require a general anaesthetic.

Data extraction
Data from outcomes and resource use studies were used to construct and judge the cost-effectiveness. Two reviewers independently applied the inclusion and exclusion criteria to the eligible studies. They first screened the titles, then abstracts and finally the full text. At each step their agreement was assessed using Cohen's κ statistic with a 95 per cent c.i. 29 . Cohen's κ statistic adjusts the proportion of articles for which there is agreement by the amount of agreement expected by chance alone 29,30 . Agreement strengths for Cohen's κ are defined 29,30 as: poor, κ < 0⋅00; slight, κ = 0⋅00-0⋅20; fair, κ = 0⋅21-0⋅40; moderate, κ = 0⋅41-0⋅60; substantial, κ = 0⋅61-0⋅80; and almost perfect, κ = 0⋅81-1⋅00. Disagreements were resolved by discussion, and when consensus could not be reached a third reviewer acted as referee. Reasons for exclusion were documented. All eligible articles that passed the full-text screening were included in the review.
Extracted study data were recorded in a data collection form; they included year and country of study, study design, definition of SSI, population demographics, surgical procedures, antibiotic prophylaxis (costs, dosage and mode of administration), mean hospital and patient costs, and outcome data (duration of hospital stay, mortality, incidence of SSI, bacteria identified and antimicrobial resistance).

Reporting quality assessment
The 24-item Consolidated Health Economic Evaluation and Reporting Standards (CHEERS) checklist 31 was used to assess comprehensively the quality of the clinical and methodological reporting relating to title, structured abstract, methods, results, discussion, conclusion, funding and conflicts of interest. Two of the checklist items (choice of a model and assumptions) were not included as they were not applicable to any of the studies. Each of the remaining 22 items were assigned a weighted rating 16 : 0, did not report; 1, reported poorly; 2, reported well. The overall quality rating is the proportion of items reported well: high quality, 17 or more of 22 (77 per cent or above); medium/acceptable quality, 11 or more and fewer than 17 of 22 (50 per cent or above and less than 77 per cent); and low/unacceptable quality, fewer than 11 of 22 (less than 50 per cent). There is methodological reporting overlap between the CHEERS checklist and the economic quality checklist described below.

Economic quality assessment
A modified version of the Drummond et al. checklist 27 was used to assess the quality of the economic and methodological reporting. The checklist includes ten questions, of which two have subquestions. These 12 questions enabled assessment of the following elements for each study: methods used (appropriate and accurate measurement of costs and outcomes), clinical effectiveness, limitations, uncertainty, relevance, generalizability and conclusions. Answers assigned to each question could be: 'yes', 'no' or 'not applicable'. The overall quality ratings are based on the number of questions answered as 'yes': high quality, nine or more of 12 (75 per cent or above); medium/acceptable quality, six or more and fewer than nine of 12 (50 per cent or more and less than 75 per cent); and low/unacceptable quality, fewer than six of 12 (less than 50 per cent).

Incremental cost-effectiveness ratio
When treatment effect (TE) and incremental cost-effectiveness ratios (ICERs) were not reported, they were calculated using the study data. Treatment effect is defined as the difference between the control and intervention effect (TEc − TEi). To determine the incremental cost saving of SSIs averted, the difference in mean total cost between the intervention and control prophylaxis was divided by the treatment effect. Calculated ICER costs were then adjusted to British pounds (2016) in a two-step process, using the Campbell and Cochrane Economics Methods Group-Evidence for Policy and Practice Information and Coordinating Centre cost converter web-based tool 32,33 .
Step 1 inflates the cost from the original price year to April 2016, using a Gross Domestic Product deflator index (GDPD values), obtained from the International Monetary Fund World Economic Outlook Database GDP deflator index data set 34 .
Step 2 converts the original currency to British pounds, using conversion rates based on Purchasing Power Parities for GDP (PPP values) 32,33 . Using a web-based tool, the 2016 British pound to euro conversion factor for £1 sterling is €1⋅28. When not stated, accepted standard practice to infer price year and/or currency 33 was used. The price year was assumed to be either the year the study ended or the year of publication, and the original currency to be the same as that in the study setting.

Quality assessment of reporting
The reporting quality of most of the studies was low to moderate using the CHEERS statement checklist 31 ( Table 2; Table S4, supporting information). Only one study 39 had a high reporting quality for 18 of the 22 items. Three studies 37 -39 reported economic evaluations in their titles. In most studies the objectives, methods (settings, populations and comparators) were well . ‡Values in parentheses are percentages. §Infection rate and length of stay (LOS) for cefazolin, clindamycin and cefoperazone were pooled as individual results were not stated; mean cost per patient was based on length of hospital stay (LOS). ¶Patients with infection were admitted to hospital (7 placebo, 1 intervention). #Mann-Whitney U test for LOS and χ 2 test with Yates' correction for mortality. **Intention-to-treat data; infection data were missing for six patients in the control group and seven in the intervention group. † †Per-protocol data. ‡ ‡Intention-to-treat data. § §Per-protocol data; intention-to-treat data used for mortality reported in the nested study of Itani et al. 48 . C, control; I, intervention; n.c., not calculated (insufficient data in article); n.s., not stated; AH, abdominal hysterectomy; VH, vaginal hysterectomy; CABG, coronary artery bypass graft; SDD, selective decontamination of the digestive tract. P values are those reported in the article. *Intention-to-treat data for antibiotic efficacy. †Infection data were missing for six patients in the control group and seven in the intervention group. ‡Per-protocol data; bacterial isolates and susceptibility data from nested study by Itani et al. 48 . GNB, Gram-negative bacteria; GPB, Gram-positive bacteria; MRSA, methicillin-resistant Staphylococcus aureus; SWI, surgical wound infection; UTI, urinary tract infection; CABG, coronary artery bypass graft.

Clinical effectiveness of antibiotic prophylaxis, length of hospital stay and mortality
All studies included a definition for postoperative SSI (  39 was that of Rommes et al. 51 , used in the nested study of Roos and colleagues 47 . All studies reported SSI rates and the effectiveness of the preoperative antibiotic prophylaxis. Prophylactic effectiveness was demonstrated in 11 studies [35][36][37][38][39][40][41][42][43][44]46 , although effectiveness was statistically significant in only seven 37 -40,43,44,46 . Blair and colleagues 35 demonstrated effectiveness of the intervention compared with placebo, but failed to stipulate which of the three interventions was effective (cefazolin, clindamycin or cefoperazone). Effectiveness was therefore calculated for the pooled interventions. Matkaris et al. 43 demonstrated significant effectiveness of three prophylactic antibiotics versus the no-antibiotic control, and also reported comparable differences between the three prophylactic antibiotics. The study that did not demonstrate prophylactic effectiveness for the intervention compared a single dose of ceftriaxone (third-generation cephalosporin) with three doses of cefuroxime (second generation) given three times daily, in patients undergoing CABG 45 .
Eleven studies 35 -44,46 reported length of hospital stay (LOS), although the reporting was inconsistent between treatment groups as well as between infected and non-infected patients ( Table 4). Overall LOS was reduced in the intervention group for all of the studies, although this was significant in only one study 44 . LOS was increased in the presence of infection compared with no infection in two studies 35,40 . Five studies 38 -40,42,45 reported on mortality, although none stated the day of admission when the death occurred; there was no significant difference in mortality rates between intervention and control groups in the five studies 38 -40,42,45 . There was one death from infection in each arm of the Marroni study 42 , whereas in the Sisto study 45 no death was from infection. Mortality was not reported in the paper by Wilson et al. 46 , but was reported in the nested study of Itani and co-workers 48 ; the difference was not statistically significant and was not directly related to the prophylaxis.

Bacterial isolates and antimicrobial resistance
Six studies 38,40 -42,45,46 reported and identified the bacterial pathogens responsible for SSIs; the pathogens were similar across the studies ( Table 5). Clostridium difficile, a toxic organism found in the intestine causing colitis, was identified in one study 45 after surgery following a second dose of cefuroxime. Wilson et al. 46 also reported C. difficile colitis (in 2 patients who received ertapenem) and antimicrobial resistance of the pathogens to ertapenem versus cefotetan in the nested study 48 . Resistance of pathogens to ertapenem was much lower (16 per cent) than that to cefotetan (67 per cent). Only two other studies 38,41 reported antimicrobial resistance. Dhadwal and colleagues 38 found no increase in vancomycin-resistant Enterococcus or methicillin-resistant Staphylococcus aureus (MRSA) in CABG, although Gram-positive bacteria resistant to rifampicin were identified in both control (cefuroxime) and investigation (rifampicin, vancomycin and gentamicin) groups. Jones and co-workers 41 found few pathogens (8 per cent) resistant to cefoperazone and, although no pathogens were resistant to cefotaxime, 72 per cent were inhibited by cefotaxime in several surgical procedures.

Quality assessment of economic evaluation
A modified Drummond checklist 27 was used to assess economic methodological quality for each study (  39 performed an ICER analysis.

Cost analysis of antibiotic prophylaxis
Of the included studies, nine 35 -38,40,42,43,45,46 were cost-benefit studies, two were cost-effectiveness studies 39,44 and one 41 was a cost containment study ( Table 1; Table S3, supporting information). These were all from the perspective of the healthcare provider, with costs reported as mean cost per patient or per patient episode. Sources for the cost data were reported in all studies, and costs included prophylactic antibiotic, daily hospital charge, nursing/staff time, hospital care, care after discharge, and treatment of the SSIs ( Table 7). The currencies reported were: euros 39 , British pounds 37 , US dollars 35,36,38,40 -46 , drachma 43 and pesetas 40 ; both drachma and pesetas were converted to US dollars, which was the currency used in all cost analyses. Only four studies 39,40,42,46 reported the price year for the currency conversion. Nine studies 35,36,[38][39][40][43][44][45][46] reported cost savings favouring the use of the preoperative prophylaxis intervention and two 37,42 reported cost savings favouring the control prophylaxis. Davey and colleagues 37 showed significant clinical effectiveness for cephradine and mezlocillin in abdominal and vaginal hysterectomy, but neither intervention was considered cost-effective. One study 39 reported an ICER when using selective decontamination of the digestive tract versus placebo in gastrointestinal surgery, with the prevention of at least one infection leading to a reported saving of €23 164 per patient. No study discounted costs, although Dijksman et al. 39 stated that the reason for not discounting costs included a 1-year time horizon, and they did perform a sensitivity analysis. One study 45 considered only the acquisition and delivery cost of the antibiotic prophylaxis and not the treatment failures.

Calculated incremental cost-effectiveness ratio
The calculated ICER was based on the results of each study, their reported currency and euros (2016) ( Table 7;  42 showing a negative effect in vascular prosthetic surgery. The intervention in ten studies [35][36][37][38][39][40][43][44][45][46] was dominant (more effective and cheaper than the control) and in one study 42 the intervention was dominated by the control (it was less effective and more expensive). In the remaining study 41 , the intervention was more effective and more expensive than the control. This resulted in an incremental increase of €2⋅64 per patient and a resultant ICER of €5⋅12 for the year 2016.

Discussion
This review aimed to evaluate the cost-effectiveness of preoperative antibiotic prophylaxis in preventing SSIs, including assessment of the reporting quality of the clinical and cost-effectiveness. Twelve studies published between 1988 and 2014 were identified, and included preoperative antibiotic prophylaxis as well as costs. Most of the studies had a large sample size: five had more than 500 participants, four had between 200 and 500 participants and three had fewer than 200 participants. All studies reported some measure of costs, but only two reported on incremental cost-effectiveness and none included any of the recommended economic checklists 27,31 . All identified studies reported on prophylactic effectiveness, although few included antibiotic resistance and none addressed the appropriateness of antibiotic stewardship. Prophylactic effectiveness was achieved in ten studies. The size of these effects is considered clinically important, particularly in contaminated and clean-contaminated surgery 37,39 -41,44,46 , which has a higher risk of baseline SSI compared with clean procedures 57 . Five 35,36,38,42,45 of the included studies involved clean surgical procedures, so clinical effectiveness in four 35,36,38,42 of these studies was not unexpected. Prophylactic effectiveness was also achieved even when the comparator was another antibiotic 38,40,41,46 . Most of the prophylactic interventions involved first-, second-or third-generation cephalosporins compared with either placebo or a control. Cephalosporins are safe and have a long half-life, ensuring penetration of tissues 21 . They offer cover against most S. aureus strains and some Gram-negative organisms, but not coagulase-negative staphylococci or MRSA 22 . Only two studies mentioned screening for C. difficile. Cephalosporins, especially third-generation drugs, have been linked to patients having an increased risk of colonization with C. difficile, causing toxic C. difficile colitis 22 , even when administered as a single dose 58,59 . The size and dosage of antibiotic prophylaxis is important, as single-dose administration may precipitate resistance unless the prophylactic drug has a sufficient half-life and tissue penetration. One study showed that a single dose of the intervention (cefoperazone) was less effective clinically and cost more than control prophylaxis (cefotaxime). Both of these antibiotics are third-generation cephalosporins, and both were administered as a single bolus 30 min after anaesthesia but before incision. Cefotaxime was administered again during surgery if the duration of the procedure exceeded 2 h.
Teicoplanin, a glycopeptide, may also be administered as a single dose. Its use as an intervention, however, was less effective and more expensive compared with cefazolin (a first-generation cephalosporin). Cefazolin remains the prophylactic choice in vascular surgery as it is effective against S. aureus (the most frequently isolated organism in infected vascular wounds). Cefazolin has been shown to be as effective as cefamandole and cefuroxime in prosthetic vascular surgery 60 . With the increase in MRSA, vancomycin is an alternative, but it is toxic. Teicoplanin is similar to vancomycin, but is less toxic and has a longer half-life, so may be administered once daily. Teicoplanin lacks activity against Gram-negative bacteria, however, and most infections in the teicoplanin study were caused by Gram-negative bacteria; this may have contributed to the increased costs per patient.
Combining the findings of economic evaluations with those of clinical-effectiveness trials provides healthcare policy-makers with evidence-based options for healthcare decision-making. The methodology of economic evaluations needs to be defined clearly at the study outset. This review identified low to acceptable reporting of the economic evaluations, but with great variation, whereas the reporting of clinical effectiveness was more standardized.
The most recent studies were more consistent in terminology and reporting of costs and their units. Some of the studies did not include treatment failures in their cost analysis, and this may result in an intervention that is cost-saving but not necessarily cost-effective. In addition, cost-effectiveness may be more favourable in procedures that carry a higher baseline risk of SSI when the cost of prophylaxis is the same. Length of hospital stay is a recognized factor contributing to costs 7 -9,11 , and all studies reported a reduced length of stay compared with the control regimen; however, it was difficult to determine the exact costs of the stay. It is also recognized that mean daily costs decrease with extended length of stay, with the most intensive costs incurred in the period shortly after admission 9 ; this may be perceived as a disincentive for hospitals to eliminate all SSIs 9,10 . None of the included studies reported decreasing costs of the hospital admission; all reported a daily hospital charge. Mortality also has an associated cost, and in cost-effectiveness studies is considered a permanent sequela. Only five studies and a nested study reported mortality, and none included deaths in the cost analysis.
The methodological quality of the included studies was not well reported, as evidenced by low scores on the CHEERS checklist 31 , whereas economic reporting was moderate to high, with seven studies ranking 75 per cent or above on the modified Drummond quality checklist 27 . Two of the highest-quality studies were among the most recent ones, published in 2008 and 2012. There was, however, no standard method of reporting costs, and some cost components were not always reported; discounting was not reported in any study. Consistent inclusion of standardized economic studies in clinical trials and quasi-experimental studies would allow evidence-based decision-making with respect to antibiotic efficacy and cost-effectiveness.
This review has five main limitations. First, the search terms used may not have identified all articles, as a wide variety of terms exist to describe economic evaluations, prophylaxis and infection. Second, the review was restricted to studies performed in OECD countries. The purpose of the restriction was to reduce the effect of differences in operating theatre conditions and surgical procedures on the incidence of SSI. Third, the ICER analysis is based on the published study data and, because there was heterogeneity between the studies and sensitivity analysis was not always reported, it was limited to point estimates. Fourth, in this review, an ICER was not sensitive enough to rank cost-effectiveness, as most of the interventions were dominant. For the dominant interventions using an ICER the range of difference could not be determined, and possibly a quality-adjusted life-year framework would be more suitable; however, this would require standardized reporting. Fifth, despite the importance of preventing primary antibiotic resistance, the review did not attempt to address the development of resistance or antibiotic stewardship, because no study reported on either. This also implies that the results of these studies have limited generalizability; if resistance patterns differ, a drug that is (cost-)effective in one context may not be in another. The specific findings of the studies reviewed here should therefore be treated with caution.
The strengths of this review are several. It is the first to include both clinical and economic effectiveness of preoperative prophylaxis; it included five databases, and the numerous keywords were matched with indexed terms specific to the databases. This review summarized large data sets that encompassed many surgical specialties and procedures. It is recommended 29,30,61 that more than one reviewer should screen for papers to be included in a systematic review. This review used two independent reviewers, and the κ statistic for each level of screening was at the higher end of the scale (from substantial to almost perfect).
This review of the cost-effectiveness of preoperative antibiotic prophylaxis found that most interventions were cost-effective. To ensure that preoperative prophylaxis continues to prevent SSI, there needs to be increased awareness of the prevalence of resistance within each facility and improved antibiotic stewardship to reduce the development of resistance. Antibiotic stewardship includes use of the appropriate recommended antibiotic prophylaxis based on the most common pathogens likely to cause SSI for a specific surgical procedure, following recommended timing of administration before incision to ensure maximum tissue concentration, adjusting the prophylaxis dose according to the patient's bodyweight, redosing the prophylaxis at intervals of two half-lives, and discontinuing prophylaxis after surgery within recommended time frames. New antibiotic prophylaxis regimens may be implemented when they are less effective or more expensive if economic methods are not included routinely in RCTs and quasi-experimental studies. Economic methods would improve the understanding and true economic benefit of these new regimens. The economic methods need to be standardized against recommended guidelines and incorporate sensitivity analysis, discount rates, year and date of the study, unit costs, mortality, treatment effects, antibiotic resistance and quality-of-life costs.

Disclosure
The authors declare no conflict of interest.