Economic evaluations considering costs and outcomes of diabetic foot ulcer infections: A systematic review

Background Diabetic foot ulcer (DFU) is a severe complication of diabetes and particularly susceptible to infection. DFU infection intervention efficacy is declining due to antimicrobial resistance and a systematic review of economic evaluations considering their economic feasibility is timely and required. Aim To obtain and critically appraise all available full economic evaluations jointly considering costs and outcomes of infected DFUs. Methods A literature search was conducted across MedLine, CINAHL, Scopus and Cochrane Database seeking evaluations published from inception to 2019 using specific key concepts. Eligibility criteria were defined to guide study selection. Articles were identified by screening of titles and abstracts, followed by a full-text review before inclusion. We identified 352 papers that report economic analysis of the costs and outcomes of interventions aimed at diabetic foot ulcer infections. Key characteristics of eligible economic evaluations were extracted, and their quality assessed against the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist. Results 542 records were screened and 39 full-texts assessed for eligibility. A total of 19 papers were included in the final analysis. All studies except one identified cost-saving or cost-effective interventions. The evaluations included in the final analysis were so heterogeneous that comparison of them was not possible. All studies were of “excellent”, “very good” or “good” quality when assessed against the CHEERS checklist. Conclusions Consistent identification of cost-effective and cost-saving interventions may help to reduce the DFU healthcare burden. Future research should involve clinical implementation of interventions with parallel economic evaluation rather than model-based evaluations.


Aim
To obtain and critically appraise all available full economic evaluations jointly considering costs and outcomes of infected DFUs.

Methods
A literature search was conducted across MedLine, CINAHL, Scopus and Cochrane Database seeking evaluations published from inception to 2019 using specific key concepts. Eligibility criteria were defined to guide study selection. Articles were identified by screening of titles and abstracts, followed by a full-text review before inclusion. We identified 352 papers that report economic analysis of the costs and outcomes of interventions aimed at diabetic foot ulcer infections. Key characteristics of eligible economic evaluations were extracted, and their quality assessed against the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist.
Results 542 records were screened and 39 full-texts assessed for eligibility. A total of 19 papers were included in the final analysis. All studies except one identified cost-saving or cost-effective interventions. The evaluations included in the final analysis were so heterogeneous that comparison of them was not possible. All studies were of "excellent", "very good" or "good" quality when assessed against the CHEERS checklist.

Introduction
Diabetic foot ulcers (DFU) are a common and severe complication of diabetes mellitus characterised by a deep tissue lesion. [1] Factors underlying the development of DFU are peripheral sensory neuropathy, foot deformity, minor foot trauma and peripheral arterial disease. [2] It is estimated that the annual incidence of DFU is 2-4% [3,4] in developed countries with only two-thirds of cases healing within 12 months. [5] A significant consequence for those DFUs which do not heal is infection which has an incidence of 40.1%. [6] DFU infection is a well-recognised risk-factor for lower extremity amputation which occurs in 8% [5] of cases and carries a 5-year mortality of 74%. [7] Not only is DFU a source of significant patient suffering, it also brings significant costs to the individual and healthcare system. The cost burden of DFU requires 6 days to 5.7 years of patient income to cover treatment cost with variation based on setting and treatment strategy. [8] The annual cost of DFU treatment is significantly greater than non-diabetic foot ulcer treatment, estimated at $1.38 billion versus $0.13 billion. [9] DFU infection places an additional burden on the healthcare system. Cost per admission among patients with DFU infection versus without was significantly higher in those with infected DFU ($11,290 versus $8,145). [9] Cost-effective DFU interventions have been identified in previous systematic reviews of economic evaluations, [10][11][12][13] however these reviews do not focus on those evaluations where infection is part of the clinical presentation of the DFU or model pathway. Given the high prevalence of infection in DFU and the accompanying economic burden, it is essential to illuminate potentially cost-saving or cost-effective interventions to reduce the burden of DFU infection. We therefore seek to obtain all available economic evaluations that jointly consider the costs and outcomes of DFU with infection considered as part of the clinical situation and critically appraise this literature.

Search strategy
A literature search was conducted using MedLine, CINAHL, Scopus and Cochrane Databases seeking articles published in English from inception to 2020. Terms including diabetic foot, economic evaluation and infection were used as Medical Subject Headings (MeSH) and textwords to capture the outcomes of interest. The MedLine search strategy, adapted for use in other databases, is presented in Table 1. The last database search was 31 Jan 2020.
3. the study population was diabetic with an infected foot ulcer, 4. they were published in the English language in peer reviewed journals between inception and 2020.
Studies were excluded if: 1. costs and outcomes were not considered and/or compared, 2. study population was not over the age of 18, 3. they were theory papers, letters, editorials, reviews, theses, or dissertations and studies where full texts could not be obtained.
This systematic review was conducted according to the Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the checklist is provided as supporting information (S1 Checklist). [14] Articles were identified by screening titles and abstracts, followed by assessment of full-texts for eligibility.

Data extraction and quality assessment
Key characteristics of the economic evaluations were identified and extracted including study design and perspective, study population, intervention and comparator(s), time horizon and discount rate, methods or model used, costs included, reporting of costs, outcomes measuring health benefit and cost-effectiveness and overall economic evaluation result.
Quality assessment of the reporting of identified studies was performed according to the 24-item Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist. [15] Two reviewers independently assessed articles against the criteria, calculating a score out of 24. Each item was assigned one-point, partial marks were awarded if the study did not completely fulfill the criteria, for example if perspective or discount rate choices were not 3 (diabet � adj3 (foot or feet or wound � or ulcer � or lesion � or lower limb � or lower extremit � )).tw,kf. 4 or/1-3 5 economics/ or "costs and cost analysis"/ or "cost allocation"/ or cost-benefit analysis/ or "cost control"/ or "cost of illness"/ or health care costs/ or direct service costs/ or hospital costs/ or health expenditures/ or economics, hospital/ or hospital charges/ or economics, medical/ or fees, medical/ or economics, nursing/ or economics, pharmaceutical/ explained. Any differences in marks awarded were discussed by reviewers to reach consensus. Calculation of a percentage score was performed. Given the absence of a largely accepted method for quality appraisal, set categories were based on published literature. [16][17][18][19] Studies scoring 85% or higher were of "excellent" quality, studies scoring between 70-<85% of "very good" quality, studies scoring 55-<70% were rated to have "good" quality and studies scoring below 55% were classified as "poor" quality.

Study selection
PRISMA guidelines [14] were followed in the study selection process (Fig 1). Database searches identified 527 studies and an additional 63 records were identified through screening of referencing lists; 93 duplicates were removed. Titles and abstracts of 542 articles were screened for eligibility; 503 did not meet the criteria. During this stage, a second reviewer independently assessed 20% of these articles for eligibility and interrater agreement was calculated using Cohen's kappa statistic. [20] Thirty-nine full-texts were assessed for eligibility. Two papers did not consider DFU infection, [21,22] three did not report participant age, [23][24][25] one did not consider DFU, [26] two were review papers [27,28] and one was not published in a peer-reviewed journal. [29] Eleven papers did not consider
Study perspectives. The economic perspective taken by each study determines the cost and benefits included. [17] The societal perspective was taken by Cárdenas et al. [41] and Redekop et al. [54] The perspective of the healthcare system was taken in six studies [42,43,50,55,57,59] and the payers perspective was taken in five [45,46,49,56,58] (Table 2). Both the societal and payers perspectives were taken by Guo et al. [48] Perspective was not reported in five studies, [44,47,[51][52][53] however Ortegon et al. [51] discussed some results from the policy and clinical perspective.
Intervention and comparator. A multitude of interventions were used to manage DFUs, typically antimicrobials or wound care strategies, reflecting components of the standard treatment of DFUs [2,60] (Table 2). Most economic evaluations assessed adjuncts to standard wound care strategies. [41,47,[54][55][56] Two studies assessed becaplermin gel plus good wound care (GWC) [52,58] and two studies assessed hyperbaric oxygen therapy (HBOT). [43,48] The overall wound care strategy was compared in three studies, each comparing what was locally considered as standard versus optimal wound care. [42,53,59] Wound dressings were assessed in four.
Time horizon. Time horizons should be specified and cover the provision of the intervention and tracking of costs and consequences/benefits. Ideally, they should reflect actual clinical practice. One-year [41,45,54,58] and five-year time horizons were most common. [42,44,53,55,56,59] The longest time horizon was 12-years by Chuck et al. [43] Short time horizons were taken by Gilligan et al. [46] and Guest et al. [47] of three and four months respectively. Two time horizons were taken in two papers, Tesar et al. [56] used 5-and 10-year time horizons and Lobmann et al. [49] used 20-week and 100-week time horizons. Persson et al. time horizon was between one and two years. [52] Four studies did not explicitly report the time horizon. [48,50,51,57] In McKinnon et al., [50] time horizon was ambiguous, reported as being from the beginning of study-drug commencement to the completion of study-drug or secondary treatment following clinical failure. Overall, reported time horizons ranged from 12 weeks to 12 years.
Discount rates. Discount rates allow economic evaluations to account for changes in the value of money over time. Four studies did not report a discount rate. [43,44,49,57] Three studies used a 3% discount rate, however none of them justified this choice. [48,51,53] Four studies used a 5% discount. [42,52,55,59] All studies with a time horizon of one year of less did not discount costs. [41, 45-47, 50, 54, 58] Study designs and models used. Eleven studies were CEAs, [41-43, 45, 47, 49, 52, 54, 55, 58, 59] another five were CUAs, [44,48,51,53,56] while three studies found the interventions were equally efficacious, therefore costs were directly compared as in a CMA. [46,50,57]    Two studies did not use an economic model to simulate the impact of interventions on DFU. [57,58] Tice et al. performed a direct cost comparison as each intervention was assumed to be equally efficacious. [57] Waycaster et al. used wound surface area reduction rates to predict costs associated with DFU healing. [58] Seven studies used a decision-tree analytical model. [41,43,44,47,48,50,59] The Markov model was the most common choice and was used in ten studies. [42,45,46,49,[51][52][53][54][55][56] Costs included. There was variation between cost inclusions in each study depending on setting, perspective and interventions investigated. Some studies kept a narrow scope of direct costs associated with the intervention [47,56,57] or ulcer state, [44] but most studies had a broad scope of inclusions that captured DFU intervention, rehabilitation and patient management. [41, 48, 49, 52-55, 58, 59] Indirect costs were only itemised in one study. [41] Overall economic evaluation results. All evaluations except one concluded the intervention assessed was cost-effective or cost saving. This means all interventions provided more health benefit at a lower incremental cost most of the time. All CMAs showed the intervention achieved equal health benefit at lower costs versus the comparator.
While the evaluations are incomparable due to heterogeneous methods and analyses, the intervention dominated the comparator in nine studies by providing greater health benefits at lower cost. [42-45, 47, 49, 52, 54, 59] The intervention was dominated by the comparator in one study due to the unit cost of the adjunct. [56] Six studies found the intervention to be costeffective [41,48,51,53,55,58] and three found the intervention was cost saving as the health benefits were equivalent. [46,50,57] Ragnarson et al. [53] found the intervention was dominant or cost-effective for higher risk patients in all age groups, but was dominated by the comparator in the lowest risk group.
Adjuncts were dominant or cost-effective interventions in nine of ten papers. McKinnon et al. [50] and Tice et al. [57] evaluated antimicrobials, both finding the intervention was cost

PLOS ONE
Economic evaluations of diabetic foot ulcer infection intervention: A systematic review saving. In the four studies which assessed dressings, three found the intervention was dominant [44,45,49] and one found the intervention was cost-saving. [46] Three papers compared standard wound care to optimal wound care strategies. [42,53,59] Cheng et al. [42] and Wu et al. [59] found optimal wound care was dominant while Ragnarson et al. [53] found optimal wound care was dominant or cost-effective only in higher risk groups.

Quality assessment of economic evaluations
The reporting quality of each paper was assessed against the 24-item CHEERS checklist. [15] Studies were allocated one mark for each criterion met in full (represented by p ), 0.5 marks if the criterion was partially met (represented by 6 ¼) or if the criterion was not met, 0 marks (represented by ×) ( Table 3). The total possible score was reduced by one point for all criteria that were not applicable (N/A) to a single study. For example, studies that were not model-based could not be assessed by criteria 15 or 16 (model justification and assumptions). Six studies were of "excellent" quality (scoring >85%). [41,45,49,55,58,59] Nine studies were of "very good" quality (scoring 70-<85%) [42, 43, 46-48, 50, 52, 54, 56] and four studies were of "good" quality (scoring 55-<70%). [44,51,53,57] The best addressed criterion was findings and limitations; conversely, the least addressed areas were study perspective, time horizons and discount rates. Many studies failed to report these and where reported, justification of their relevance was absent. Similarly, discussion of choice of outcomes was rarely related to the particular health state or intervention.

Discussion
Whilst the need for effective DFU interventions increases, few have been subject to economic evaluation. All interventions examined in these evaluations were cost-effective or cost-saving in a clinical situation involving DFU infection. Collectively, they suggested that short-and long-term implementation of such interventions could reduce the burden of DFU infection on healthcare systems while providing optimal patient management. Although the evaluations captured the standard care for DFUs and associated costs, other considerations were made on the issue. These included assessments of antibiotic efficacy, route and setting of administration and the overall strategies embodied in a variety of guidelines and recommendations. Seventeen evaluations were model-based and did not implement the intervention in clinical settings. [41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56]59] They relied on published data and literature to build a model that simulated the intervention and transitions between health states. Decision tree and Markov models were used, and model justification and assumptions were generally well-reported. Although models are widely accepted methods for informing policy-making decisions, [19] future research would benefit from implementing the interventions in a clinical setting.
Together all evaluations assessed at least one component of the DFU management strategy, with adjuncts assessed in most papers. Because the treatment of DFUs has multiple components including antimicrobials and standard wound care, it was to be expected that these economic evaluations would be too heterogeneous for comparison.
Quality assessment of studies against the CHEERS checklist [15] found all studies to be of "excellent", [41,45,49,55,58,59] "very good" [42, 43, 46-48, 50, 52, 54, 56] and "good" [44,51,53,57] quality, adding strength to the conclusions drawn in this review. Furthermore, some studies included were published before the conception of the CHEERS checklist in 2013. [15] This shows that high-quality studies with earlier publication dates conform with the guidelines crystallised in the checklist and supports the comprehensiveness of the search strategy. There are some limitations to our study. The results of this review rely exclusively upon studies published in English, which may not represent all research. Grey literature that is unpublished or published but non-commercially available was not searched due to time constraints. Full texts that could be obtained were limited to those accessible by the Flinders University library system. Although few studies were parallel evaluations to clinical trials, the majority of studies were model-based so the results relied heavily upon simulation of intervention effects rather than clinical application.

Conclusion
In conclusion, economic evaluations have considered all aspects of DFU intervention, finding there is potential to select more cost-saving and cost-effective alternative to reduce the burden of DFU. Instead of model-based evaluations, future research should be directed toward actual implementation of interventions in clinical settings with economic evaluations in parallel.
Supporting information S1 Checklist. Preferred Reporting Items of Systematic Reviews and Meta-Analyses (PRISMA) checklist. (DOCX)

Author Contributions
Conceptualization: Taylor