A cost‐effectiveness analysis of optimal care for diabetic foot ulcers in Australia

Abstract

In addition to affecting quality of life, diabetic foot ulcers (DFUs) impose an economic burden on both patients and the health system. This study developed a Markov model to analyse the cost‐effectiveness of implementing optimal care in comparison with the continuation of usual care for diabetic patients at high risk of DFUs in the Australian setting. The model results demonstrated overall 5‐year cost savings (AUD 9100·11 for those aged 35–54, $9391·60 for those aged 55–74 and $12 394·97 for those aged 75 or older) and improved health benefits measured in quality‐adjusted life years (QALYs) (0·13 QALYs, 0·13 QALYs and 0·16 QALYs, respectively) for high‐risk patients receiving optimal care for DFUs compared with usual care. Total cost savings for Australia were estimated at AUD 2·7 billion over 5 years. Probabilistic sensitivity analysis showed that optimal care always had a higher probability of costing less and generating more health benefits. This study provides important evidence to inform Australian policy decisions on the efficient use of health resources and supports the implementation of evidence‐based optimal care in Australia. Furthermore, this information is of great importance for comparable developed countries that could reap similar benefits from investing in these well‐known evidence‐based strategies.

Introduction

The number of people living with diabetes globally was estimated to be 422 million in 2014 1. One common complication of diabetes is diabetic foot ulcers (DFUs). Among patients diagnosed with diabetes, the annual incidence of developing foot ulcers ranges from 1% to 4% 2. DFUs not only significantly increase the risk of poor quality of life, morbidity and mortality of affected individuals 2, 3, 4 but also impose a large economic burden on patients and the health care system. Recent studies from developed countries report substantial estimated national annual costs incurred from DFU care, ranging from $547 million in hospitalisation costs in Canada 5 and £580 million in health system costs in England 6, to US$9–13 billion costs incurred by Medicare and private health insurers in the USA 7. In other developed nations where evidence‐based strategies for DFUs have been systematically implemented, these national costs have been significantly reduced 8, 9, 10, 11. Although there are national and international evidence‐based guidelines on the best practice management and prevention of DFUs 12, 13, many patients with diabetes do not receive this best practice care 14, 15, 16.

This appears to be especially true in Australia. With over 1 million Australians registered as having diagnosed diabetes 17 and around 24·1% of patients with known diabetes at high risk of foot ulceration 18, the annual direct costs of hospitalisation were estimated to be US$238 million in 2012 19. It is clear that the implementation of evidence‐based care in regions of Australia also considerably reduces DFU‐related hospitalisation, amputation and overall burden 20, 21, but many Australians with DFUs do not receive evidence‐based care 22, 23. Reasons include lack of knowledge among health care providers and decision makers, low patient compliance and high costs and lack of reimbursement for recommended interventions 23. Australia's failure to successfully implement evidence‐based recommendations for DFUs has coincided with a national diabetes‐related amputation rate that increased by 30% between 1998 and 2011 and was reported to be one of the worst in the developed world 24, 25.

The cost‐effectiveness of guideline‐based optimal prevention and treatment of DFUs has been established by studies from Sweden and the Netherlands 26, 27. In Sweden, Ragnarson Tennvall and Apelqvist 26 demonstrated that an intensified prevention strategy, including patient education, foot care and footwear, was cost‐effective if the risk of ulceration and amputation could be reduced by 25%. Ortegon et al. 27 also found that guideline‐based care resulted in better health outcomes and was cost‐effective and even cost saving compared with standard care in a Dutch setting. However, both studies are more than 10 years old, and the estimated cost savings and quality of life gains may not be generalisable to Australia where the population, guidelines on managing DFUs and reimbursement system are different.

To date, the cost‐effectiveness of implementing guideline‐based care has not been investigated in Australia. Thus, the aim of this study was to examine the costs and health outcomes associated with implementing optimal guideline‐based care compared with usual care in people at high risk of DFUs in Australia. It is expected that improved access to wound management expertise and evidence‐based wound care with adequate reimbursement for consumables such as dressings and offloading devices will lead to faster healing and fewer complications. In addition, by reducing recurrence and instances of infections and amputations, these changes can be expected to ease the economic burden. It is expected that this study will provide contemporary evidence for decision makers on the most efficient use of scarce health care resources to improve health outcomes for patients affected by DFUs in Australia and other developed countries. The results of this study are expected to be translated to policies that can promote the wider use of evidence‐based guidelines on the management and prevention of DFUs. Furthermore, if this study can demonstrate that a reduction in ulceration and amputation rates by implementing and funding best practice systems to manage diabetes foot complications is a ‘cost‐saving strategy’ in Australia, then this information could be of great importance for comparable countries that could reap similar benefits from investing in these well‐known evidence‐based strategies.

Methods

Description of the competing alternatives

Two policy alternatives were simulated using economic modelling, and their associated health outcomes and costs incurred were estimated and compared. The first alternative was the continuation of current practice, referred to as ‘usual care’. Under current practice, individuals receive a mix of largely uncoordinated set of services in the community, often facing substantial out‐of‐pocket costs. The amalgamation of these services represents usual care. The second alternative refers to a situation where all individuals at risk of developing DFUs receive optimal care, with full Medicare Benefits Schedule (MBS) or Pharmaceutical Benefits Scheme (PBS) reimbursement linked to services, devices and consumables. ‘Optimal care’ was defined as care that follows the set of recommendations from the National Guideline for the Prevention, Identification and Management of Foot Complications in Diabetes 12. According to the national guidelines, an optimal care programme includes the following parts:

1. Initial foot examination, with documentation of foot risk status and grading of ulcer severity in those with DFUs.

2. Debridement of callus and non‐ischaemic DFUs.

3. The use of topical hydrogel and appropriate dressings in people with DFUs.

4. Access to and use of appropriate footwear; pressure offloading using an irremovable device for those with DFUs or a removable device under particular conditions.

5. Infection management for those with DFUs.

6. Multidisciplinary care for patients with DFUs (including podiatrist visits six times per year for those without a DFU/healed DFU), including patient education.

Markov cohort simulation

A Markov model 28, 29 was developed for the natural history of DFUs. The advantage of using such models is that they are flexible, particularly for modelling chronic diseases and dealing with ongoing risks and events that might happen more than once over time. They can be adapted to compare the costs and health benefits of competing decisions. The first step was to define DFUs in terms of mutually exclusive 28 states, including all relevant health outcomes, with movement between these states based on transition probabilities. Using a decision‐analytical model provides a structure within which evidence from a range of sources can be directed at a specific decision problem for a defined population and context 30, which in this case is whether optimal care provides a cost‐effective means of treating patients with DFUs in Australia. The model was built on previous models developed for the economic evaluations of treatments for DFUs 26, 27, 31, 32, 33. In this study, as in previous models, the amputations were divided into minor and major amputations as there were significant quality of life and cost differences between major and minor amputation health states 34. Minor amputations were defined as those amputation procedures at or below the ankle, and major amputations were defined as those above the ankle 35.

In this study, the Markov model contained seven possible health states (Figure 1). Patients entered the model in the ‘complicated DFU with infection’ (31%) or ‘uncomplicated DFU’ state (69%) 36. The 5‐year model ran on 1‐month cycles for a total of 60 cycles (further details regarding the model and transitions between health states are available in the Appendix). The main outcome measures analysed in the model included expected costs and quality‐adjusted life years (QALYs) associated with usual care versus optimal care. Transition probabilities, costs and QALYs with each health state were determined with information from published literature and expert opinion (Table 1 and Table 2). Each model simulation represented a hypothetical cohort of diabetic patients at high risk of developing DFUs, and separate simulations were conducted for the following three age groups: 35–54 years, 55–74 years and 75 years and older. For the purposes of this study, high risk of developing DFUs is defined as having had a previous DFU or amputation.

Figure 1.

Markov model of diabetic foot ulcer. States are identical for usual and optimal care.

Table 1.

Transition probabilities between health states for usual and optimal care (1‐month cycle)

From	To	Age 35–54 Usual care	Optimal care	Age 55–74 Usual care	Optimal care	Age 75+ Usual care	Optimal care	References
No DFU	No DFU	0·9438	0·9621	0·9437	0·9620	0·9426	0·9609
	Uncomplicated DFU	0·0561	0·0378	0·0561	0·0378	0·0561	0·0378	36, 37, expert opinion
	Death	0·0001	0·0001	0·0002	0·0002	0·0013	0·0013	40, 41
Uncomplicated DFU	Uncomplicated DFU	0·7837	0·4954	0·78	0·4936	0·7733	0·4891
	Complicated DFU with infection	0·0079	0·0045	0·0114	0·0061	0·0165	0·009	20
	Healed DFU/No DFU	0·2083	0·5	0·2083	0·5	0·2083	0·5	38
	Death	0·0001	0·0001	0·0003	0·0003	0·0019	0·0019	40, 41, 42
Complicated DFU with infection	Complicated DFU with infection	0·8995	0·0824	0·9034	0·083	0·9011	0·0828
	Uncomplicated DFU	0·0816	0·9	0·0816	0·9	0·0816	0·9	26
	Post minor amputation	0·0146	0·0145	0·0102	0·0117	0·0083	0·0093	20
	Post major amputation	0·0042	0·003	0·0045	0·005	0·0071	0·006	20
	Death	0·0001	0·0001	0·0003	0·0003	0·0019	0·0019	40, 41, 42
Post minor amputation	Post minor amputation	0·992	0·9954	0·9883	0·9936	0·9816	0·9891
	Infected post minor amputation	0·0079	0·0045	0·0114	0·0061	0·0165	0·009	20
	Death	0·0001	0·0001	0·0003	0·0003	0·0019	0·0019	40, 41, 42
Infected post minor amputation	Infected post minor amputation	0·9141	0·0969	0·9136	0·0947	0·9094	0·0921
	Post minor amputation	0·0816	0·9	0·0816	0·9	0·0816	0·9	26, 39, expert opinion
	Post major amputation	0·0042	0·003	0·0045	0·005	0·0071	0·006	20
	Death	0·0001	0·0001	0·0003	0·0003	0·0019	0·0019	40, 41, 42
Post major amputation	Post major amputation	0·9999	0·9999	0·9997	0·9997	0·9981	0·9981
	Death	0·0001	0·0001	0·0003	0·0003	0·0019	0·0019	40, 41, 42

DFU, diabetic foot ulcers.

Table 2.

Health state costs, transition costs and quality of life weights (1‐month cycle, AUD, 2013 prices)

Variables		References
Costs
Ongoing costs for health states (community)
No DFU	AUD 45·8 for optimal care AUD 0 for usual care	MBS for podiatrist and GP; Market prices for products
Uncomplicated DFU	AUD 504·8 for optimal care AUD 302·64 for usual care	MBS for podiatrist and GP; Market prices for products
Complicated DFU with infection	AUD 829·59 for optimal care AUD 315·83 for usual care	PBS for antimicrobials; Market prices for products
Post minor amputation	AUD 1843·3 for optimal care AUD 1797·5 for usual care	MBS for podiatrist and GP; Market prices for products 26;
Post major amputation	AUD 4934·3	26
Infected post minor amputation	AUD 829·59 for optimal care AUD 315·83 for usual care	PBS for antimicrobials; Market prices for products
Initial costs for health states (community)
Uncomplicated DFU	AUD 296·8 for optimal care AUD 67·05 for usual care	MBS for podiatrist and GP Market prices for products
Complicated DFU with infection	AUD 769·9 for optimal care AUD 100·8 for usual care	MBS for pathology and radiology
Transition costs (hospital)
Minor amputation	AUD 10 640	AR‐DRG code F13B 46
Major amputation	AUD 23 921	AR‐DRG code F11B 46
Infected DFU	AUD 16 354	AR‐DRG code K01B 46
Infected post minor amputation	AUD 25 108	AR‐DRG code F13A 46
Quality of life	Utility
No DFU	0·84	48
Uncomplicated DFU	0·75	48
Complicated DFU with infection	0·7	48
Post minor amputation	0·68	48
Post major amputation	0·62	48
Infected post minor amputation	0·59	48

DFU, diabetic foot ulcers; AUD, Australian dollar; AR‐DRG, Australian Refined Diagnosis Related Group; GP, general practitioner; MBS, Medicare Benefits Schedule; PBS, Pharmaceutical Benefits Scheme.

Epidemiological parameters

Transition probabilities used in the model are presented in Table 1. For patients with diabetes at high risk of foot ulcers, the annual incidence of DFU was assumed to be the same as the risk of developing a recurrent ulcer after healing of a previous ulcer. For optimal care, an annual recurrence rate of 37% was assumed based on the Australian data 36, while for usual care, 50% annual recurrence was assumed based on expert opinion and systematic reviews 37. Probabilities of healing were informed by the proportion of ulcers completely healed at the 1‐month time point, with the probability of healing markedly improved with the implementation of evidence‐based care (50% for optimal care and 20·83% for usual care) 38. Probabilities of transitioning to the state of complicated DFU requiring hospitalisation as well as minor and major amputation were derived from an Australian study that determined the change in the annual incidence of foot‐related hospitalisation and amputation after the introduction of an evidence‐based, multifaceted foot‐related complication management strategy for patients with diabetes 20. The probability of infection healing for usual care was informed by a Swedish study 26, while for optimal care, it was based on expert opinion and practice guidelines 39. The probability of death for diabetes patients was calculated by dividing the number of diabetes deaths by the number of people diagnosed with diabetes. The age‐specific number of diabetes deaths in Australia in 2014 was derived from the WHO Mortality Database 40, and the age‐specific diabetes population was reported by the Australian Health Survey, 2011–2012 41. As a history of DFU would increase the risk of mortality among diabetic patients 42, 43, the probability of death of patients in uncomplicated and complicated DFU states in this model was assumed to be higher (hazard ratio of 1·47) 42 compared with ‘no DFU’ state. We assumed that the probability of death for individuals in the amputation states was the same as that for individuals in the DFU states as there was insufficient evidence for an increased risk of mortality after amputation 44. Yearly probabilities were transformed to monthly probabilities by the formula: tp = 1 – (1 − tp _t)^1/t 45. Because the probabilities of moving between states in each cycle must sum up to 1, the probability of staying in a state was simply 1 minus the sum of the probabilities of leaving the state.

Estimating the costs of providing usual and optimal care

Details of resource use by health state for optimal care and usual care are provided in the Appendix. All costs associated with health states and transition costs in the Markov model were measured in Australian dollars for 2013, as presented in Table 2. The community health services cost items included were all consultations with a general practitioner (GP) and podiatrist. Community health services were valued in line with Australian Federal Government reimbursements through MBS and medications in line with PBS. A review of market prices for all relevant products was used to value all consumable items (which is available in spreadsheet form from the authors on request). For patients undergoing amputation procedures, a cost for transitioning from ‘complicated DFU with infection’ to ‘post minor amputation’ or ‘post major amputation’ was informed by an Australian refined diagnosis‐related group (AR‐DRG) used in the National Hospital Cost Data Collection Australian Public Hospitals Cost Report 2012–2013, Round 17 46. Similarly, hospitalisation costs of ‘complicated DFU with infection’ and ‘infected post minor amputation’ states were also based on AR‐DRG cost weight 46. Costs of ‘post minor amputation’ and ‘post major amputation’ included home care, prostheses, inpatient and outpatient care. As there are currently no Australian data available on the cost of post minor amputation and post major amputation care in the community, these costs were informed by a Swedish study 26 and were converted to AUD (average exchange rates in 1998: 1 SEK = 0·2 AUD) and then inflated to 2013 prices 47. This economic evaluation was carried out from a health system perspective, and patients' out‐of‐pocket costs associated with either optimal or usual care were not considered. The cost estimates were also interpreted at a population level by multiplying the difference in costs by the number of individuals at high risk of developing DFUs. By 31 December 2015, there were 1 192 927 people with diagnosed diabetes registered on the National Diabetes Services Scheme in Australia 17, and 24·1% of them are at high risk of DFUs 18. The distribution of diabetes population among age groups was informed by the Australian Health Survey, 2011–2012 41.

Quality of life estimates

Quality of life assessments for the six possible health states were based on the generic EuroQol instrument and obtained from a previous published study, as presented in Table 2 48. Redekop et al. used a time trade‐off method to estimate the utility weights associated with a range of health states related to DFUs and their complications. This source has also been used previously to provide the quality of life estimates for a cost‐effectiveness analysis of a negative pressure device for the treatment of DFUs 32. QALYs were calculated by multiplying the quality of life utility weight for one health state by the number of years staying in that state.

Discounting

Future costs and health effects were discounted at 5% in baseline analyses in line with guidelines for submission to the Medical Services Advisory Committee and the Pharmaceutical Benefits Advisory Committee in Australia 49, 50.

Scenario analysis

Scenario analyses were conducted to determine how different choices of discounting rate, starting distribution of DFU patients with infection, topical antimicrobials and systemic antimicrobials could have influenced the model results. Parameters used in scenario analysis are presented in Table A1 of the Appendix.

Probabilistic sensitivity analysis

Parameter uncertainty was quantified using probabilistic sensitivity analysis (PSA) to give the decision‐maker insight into its impact on the final results. Point estimates (mean values) for transition probabilities, mortality risks and cost and health outcomes do not provide information about the uncertainty in estimates. Hence, statistical distributions were used to describe variability in the parameters. The probability distributions were chosen based on the type of parameter under consideration. For example, transition probabilities, which must be between 0 and 1, had a beta distribution 28. Beta distributions were also specified for estimated mean QALYs and variances. Gamma distribution using the method of moments 28 reflected the skew typically found in cost data.

Re‐samplings were performed 10 000 times using Monte Carlo simulation. Each time, the model parameters took random values from the fitted distribution, and the economic outcomes of change to costs (ΔC) and change to health benefits or QALYs (ΔE) were calculated. A decision rule for cost‐effectiveness was given by $\frac{\Delta C}{\Delta E}<\lambda$, where λ is the decision‐maker's maximum willingness‐to‐pay threshold for an additional unit of health benefit gains. A value of $64 000 per QALY was used for willingness‐to‐pay in this study, in line with recent studies in an Australian setting 51. The results of PSA were used to estimate the probability that optimal care was cost‐effective as the cost‐effectiveness threshold varied. This analysis was presented using a cost‐effectiveness plane and cost‐effectiveness acceptability curve. The probability that optimal care was cost‐effective at certain willingness‐to‐pay thresholds was derived by counting the number of times out of 10 000 that the ratio of ΔC to ΔE is less than λ, and was summarised in cost‐effectiveness acceptability curves.

Results

Baseline analysis results

Expected costs and QALYs per patient within 5 years for the three age groups were summarised in Table 3. Model results demonstrated an overall 5‐year cost saving ($9100·11 for 35–54 years, $9391·60 for 55–74 years and $12 394·97 for 75+ years) and improved health benefits (0·13 QALYs for 35–54 years, 0·13 QALYs for 55–74 years and 0·16 QALYs for 75+ years) for high‐risk DFU patients treated with optimal care compared with usual care. Patients with older ages incurred higher costs but had lower quality of life, no matter which type of care they received. Total cost savings for the Australian health system were estimated. If all patients at high risk of developing DFUs in Australia were to receive optimal care instead of usual care for DFUs, then the cost savings nationwide are likely to be AUD 2·7 billion over 5 years.

Table 3.

Markov model results for usual care and optimal care. Total expected 5‐year costs and QALYs per patient with 95% credible interval (AUD 2013 prices, discount rate 5 %)

Age group	Care received	Costs (95%CI)	QALYs (95%CI)	Incremental costs (95%CI)	Incremental QALYs (95%CI)
35–54	Usual Care	$15 780·77 ($5513·90, $34 706·98)	3·54 (2·28, 4·23)
	Optimal care	$6680·66 ($2111·21, $15 488·51)	3·67 (2·09, 4·37)	−$9100·11 (−$27 627·33, $3818·86)	0·13 (−0·42, 0·70)	Dominating*
55–74	Usual Care	$16 334·76 ($5961·58, $36 096·41)	3·53 (2·27, 4·21)
	Optimal care	$6943·16 ($2352·70, $16 057·59)	3·66 (2·02, 4·41)	−$9391·6 (−$28 711·66, $3777·27)	0·13 (−0·42, 0·71)	Dominating*
75+	Usual Care	$19 461·33 ($6604·40,$ 43 385·07)	3·38 (2·20, 4·04)
	Optimal care	$7066·36 ($2357·93, 16 299·90)	3·54 (2·66, 4·20)	−$12 394·97 (−$35 445·21, $2265·41)	0·16 (−0·41, 0·71)	Dominating*

^*The optimal care alternative yields both lower costs and higher QALYs and is cost‐effective

Scenario analysis

Results from the scenario analysis are presented in Table A2 in the Appendix. Different choices of discounting rate, topical antimicrobials and systemic antimicrobials had no impact on the dominance of optimal care over usual care. Increasing the proportion of patients with infection entering the model to 70% changed the main results for the age group 55–74. Optimal care costs $39 705 more to gain an extra QALY, which is still considered cost‐effective at the $64 000/QALY willingness‐to‐pay threshold.

Probabilistic sensitivity analysis results

Results of the Monte Carlo simulation are presented in cost‐effectiveness planes (Figure 2), where each point represents one calculation of change to costs (ΔC) and change to QALYs (ΔE) when model variables take random values from specified distributions. Points below the willingness‐to‐pay threshold line suggest that optimal care is cost‐effective. Cost‐effectiveness acceptability curves (Figure 3) showed that optimal care always has a higher probability (around 85–95%) of being cost‐effective than usual care regardless of the value of the willingness‐to‐pay threshold. Cost‐effectiveness planes and acceptability curves for age groups 55–74 and 75+ are displayed in the Appendix (Figures A1, A2, A3, A4).

Figure 2.

Cost‐effectiveness plane for optimal care over usual care (age 35–54; WTP, willingness‐to‐pay).

Figure 3.

Cost‐effectiveness acceptability curve of optimal care and usual care (age 35–54; CE cost‐effectiveness).

Discussion

This is the first study in Australia investigating the costs and health benefit gains of implementing optimal care compared with usual care for DFUs among high‐risk patients. Results suggest that for high‐risk patients, implementing optimal care for DFUs adhering to national evidence‐based guidelines yields improved health outcomes and significant cost savings for the health system. We found that the additional costs of optimal care in our model could be offset in all three age cohorts of high‐risk patients by the reduced costs from the avoided DFUs, DFU‐related hospitalisations and amputations. These cost savings ranged from $9100 to $12 400 for patients of different ages and translated to a national saving of AUD 2·7 billion over 5 years even after factoring in the costs of delivering optimal care. PSA also showed that the probability that optimal care is cost‐effective is always higher than that of usual care regardless of the willingness‐to‐pay threshold.

The implementation of evidence‐based diabetes‐related foot management has been found to improve patient health outcomes in other countries. One study in the UK reported that the incidence of major amputations reduced by 62%, from 7·4 to 2·8 per 100 000 of the general population, over the 11‐year period following improvements in foot care services, including multidisciplinary team work 52. In the USA, the enrolment in a comprehensive diabetic lower‐extremity amputation prevention programme led to a reduction in foot‐related ulcer days, hospitalisations and lower‐extremity amputations among a low‐income African‐American population in Louisiana 11. Economic modelling has also been used to evaluate the cost‐effectiveness of implementing evidence‐based care for DFUs in other nations. In Sweden, the additional prevention costs associated with funding and implementing evidence‐based practice to manage diabetes foot‐related complications were found to be offset by reduced costs associated with a 25% reduction in the incidence of foot ulcers and amputations 26. Similarly, in the Netherlands, adopting international standards to prevent and treat DFUs versus current Dutch care was found to be cost‐effective, with reductions in DFUs and lower‐extremity amputations 27. Therefore, the combined findings of our study and these other national studies suggest that implementing best practice guideline recommendations on the management of diabetes complications is highly likely to significantly lower national costs, reduce associated morbidity and mortality rates and improve the quality of life of people living with diabetes.

This study was conducted from a health system perspective, and we assumed that all costs incurred were borne by the health system. However, in Australia, the costs of evidence‐based care recommended in the national guideline for diabetes complications are only partially covered by the MBS or PBS 24. Although the cost of GP consultations is covered, DFU care consumables such as dressings and offloading devices are not subsidised under the PBS or MBS outside of tertiary care, and thus, patients can incur high personal financial costs 53. In addition, all MBS‐rebatable allied health consultations (including podiatry) are capped at five per year 24, 53. Therefore, guideline‐recommended routine podiatry treatments ‘compete’ with other allied health consultations. All these result in the inability of patients to afford ongoing care – delaying (or preventing) DFU healing in Australia 53. In this study, results from the model have shown that if the health system could invest in guideline‐based care and cover all additional costs of consumables and devices, there would be huge cost‐savings in the near future. Thus, it is expected that this study could bring changes to the reimbursement system in Australia by informing policy makers on the most efficient use of scarce health care resources. By examining the current DFU care and possible changes in Australia, it is also expected that this study will draw more attention to evidence‐based guidelines on managing DFUs, both nationally and internationally.

Limitations

A factor not adjusted for in our study that could influence the cost‐effectiveness of optimal care is patient compliance. We assumed that all patients under optimal care system would follow the national guidelines 100% of the time. However, it is possible that patients could, for example, only have access to pressure offloading using an irremovable device for 70% of the time. In such circumstances, the cost of delivering optimal care in the model would be an overestimate and the clinical effect on DFU outcomes an underestimate. Additionally, in our simplified model, we assumed that for usual care, treatment would be provided only by a GP, whereas in Australia, wound management is complex and most often involves a multitude of uncoordinated health care providers and treatment arrangements, including nurses and medical specialists, and may include self‐management 53. Again for optimal care, we assumed that the multidisciplinary team consists of a GP and podiatrist, while optimal treatment could be provided by a GP, podiatrist, medical specialist or nurse practitioner with wound management training. Furthermore, the additional cost of education and training of wound care providers in optimal wound care was not included in this model.

A model simulation is always associated with some degree of uncertainty. Our model was simplified in some ways that may have influenced the results. Firstly, the model did not follow the total remaining lifetime of the patients; it followed only 5 years. Secondly, there is only one health state for infection, while in the real world, patients could develop a minor or moderate infection prior to severe infection. Moreover, we did not include re‐ulceration as a separate state after minor amputation in the model and assumed that re‐ulceration occurs along with an infection that requires hospitalisation. We plan to extend our model to include more health states when local patient‐level data become available. Thirdly, we assumed that individuals could not have more than one minor amputation or major amputation; thus, the cost of amputation may be underestimated. Fourthly, incidences of hospitalisation and amputation were derived from a study that included all diabetic patients with different risks of developing foot ulcer, while our study focused on patients with a high risk of foot ulcers. As a result, the probabilities of being admitted to hospital and undergoing an amputation used in the model could be underestimated. We will examine the reduction in amputation and hospitalisation further when local data of high‐risk patients become available. Finally, in the absence of local data, health utilities for diabetic foot patients were based on a Dutch study, and costs of post minor and post major amputations were informed by a Swedish study. As the costs of post minor and major amputations are not only from another country but also another era in health care, it is possible that the long‐term costs of living with an amputation may have been underestimated in the current study. Compared with the Swedish study that reported costs of post minor and major amputations 26, one Australian study reported lower costs of amputation (minor and major amputations combined) 54, while one US study reported higher costs of post minor and major amputations 55. To account for the great uncertainty around cost estimates, we conducted a PSA that allowed costs to take random values from a wide range, and results showed that optimal care still had a high probability of being cost‐effective, even when using costs of amputation that are lower than in the Swedish study. Bearing these limitations in mind, the current model appears appropriate for simulating the effects of optimal care for DFUs on long‐term costs associated with foot ulcers and amputations. The model can be used for future cost‐effectiveness and cost‐utility analyses of DFU prevention and treatment as new information on outcomes, costs and quality of life becomes available.

Conclusions

In conclusion, our findings clearly suggest that implementing optimal care by adhering to the Australian diabetic foot complications guidelines is cost‐effective in people at high risk of DFUs because of a reduction in DFUs, infections, hospitalisation and lower‐extremity amputations. Therefore, we recommend high‐level policy development and investment in DFU optimal care to improve affordability and support evidence‐based access to health professionals and multidisciplinary teams. Incentivising cost‐effective evidence‐based wound care within MBS and listing evidence‐based wound products on MBS/PBS will not only ease patients' financial burden but also save considerable costs for Australia's health system. Given the current economic climate and recent government focus on funding only health services found to be both effective and cost‐effective, this research is important, relevant and timely.

Methods

Markov model transitions

Patients entered the model in the ‘complicated DFU with infection’ (31%) or ‘uncomplicated DFU’ state (69%). Individuals in the ‘uncomplicated DFU’ state were assumed to be receiving either optimal care in the community or usual care in the community and could transition back to ‘No DFU’ state by becoming healed or develop ‘complicated DFU with infection’ that requires hospital care followed by additional community care. Patients who developed ‘complicated DFU with infection’ might undergo minor amputation or major amputation. Following a minor amputation, once healed, individuals could move to a ‘Post minor amputation’ state and are at risk of transitioning to an ‘Infected post minor amputation’ state which again requires hospitalisation. For the purposes of this study it was assumed that individuals could not have more than one minor amputation, and so patients in the ‘infected post minor amputation’ state either: (i) stayed in this state, (ii) were healed and returned to the ‘post minor amputation’ state, (iii) went on to have a major amputation or (iv) died. Again for the purposes of this study it was assumed that individuals could have only one major amputation and once healed moved to a ‘Post major amputation’ state. Patients in the ‘Post major amputation’ health state could either remain in that state or transition to the dead state. The model was simplified with the assumption that all amputations were preceded by complicated DFUs with infection. In addition, in all health states, individuals had a probability of death and enter into an absorbing state of ‘Dead’. Both usual and optimal care were assumed to start in the first cycle and continued in all cycles during the simulation period of 5 years. The pathway between health states in the model was identical for usual and optimal care, but with different probabilities for the incidence of foot ulcers, recurrence, infection and amputations.

Resource use by health state for optimal care and usual care

Usual care in the community for uncomplicated DFU, no ulcer states and complicated DFU with infection

Under usual care, patients with an uncomplicated DFU were assumed to receive a one‐off initial assessment by GP for risk of amputation and then undergo medical checks by a GP twice a week. Patients with uncomplicated DFU in usual care were also assumed to receive absorbent dressing changes twice a week and post‐operative boots. Once the ulcer healed, patients would receive no further care. If patients developed complicated DFU with infection, they were assumed to receive pathology services to determine the pathogen that caused infection and received systemic antimicrobials.

Optimal care in the community for uncomplicated DFU, no ulcer states and complicated DFU with infection

Optimal care was defined according to the National Guidelines for the Prevention, Identification and Management of Foot Complications in Diabetes 50 and the following use of resources: (i) all patients with an uncomplicated DFU receive a one‐off initial assessment to grade the severity of the DFU by both podiatrist and GP; (ii) patients receive debridement each week; (iii) patients receive dressings consisting of soft‐gelling cellulose fibre and polyurethane foam twice a week; (iv) patients receive an irremovable pressure offloading device during the treatment; (v) patients receive at least weekly multidisciplinary care from both podiatrist and GP trained in wound management. After the ulcer heals, for the prevention of further complications, patients would visit podiatrist every 2 months and receive one pair of extra‐depth footwear per year. Patient education was included in the podiatrist visits. If patients in optimal care developed complicated DFU with infection, they were assumed to receive pathology services to determine the pathogen that caused infection and received topical and systemic antimicrobials. Diagnostic imaging would be applied to evaluate suspected osteomyelitis.

Table A1.

Parameters included in scenario analysis

	Baseline value	Choices in scenario analysis
Discounting rate	5%	3%; 10%
Starting distribution of patients with complicated DFU with infection	31%	5%; 70%
Topical antimicrobials	Silver impregnated	Silver impregnated alginate; Silver impregnated soft‐gelling cellulose fibre
Systemic antimicrobials	amoxycillin+clavulanate	cephalxin + metronidazole; ciprofloxacin + clindamycin

Results

Table A2.

Results from scenario analyses (incremental cost/QALY for optimal care of DFU over usual care)

	Age 35–54 years	Age 55–74 years	Age 75+ years
Baseline (5% discounting rate, 31% DFU start with infection, topical antimicrobial option 1, systemic antimicrobial option 1)	Dominating	Dominating	Dominating
3% discounting rate	Dominating	Dominating	Dominating
10% discounting rate	Dominating	Dominating	Dominating
5% DFU start with infection	Dominating	Dominating	Dominating
70% DFU start with infection	Dominating	$39 704·73	Dominating
Topical antimicrobial option 2	Dominating	Dominating	Dominating
Topical antimicrobial option 3	Dominating	Dominating	Dominating
Systemic antimicrobial option 2	Dominating	Dominating	Dominating
Systemic antimicrobial option 3	Dominating	Dominating	Dominating

DFU, diabetic foot ulcers; dominating, the optimal care alternative yields both lower costs and higher QALYs and is cost‐effective.

Figure A1.

Cost‐effectiveness plane for optimal care over usual care (age 55–74; WTP, willingness‐to‐pay).

Figure A2.

Cost‐effectiveness plane for optimal care over usual care (age 75+; WTP, willingness‐to‐pay).

Figure A3.

Cost‐effectiveness acceptability curve of optimal care and usual care (age 55–74; CE cost‐effectiveness).

Figure A4.

Cost‐effectiveness acceptability curve of optimal care and usual care (age 75+; CE cost‐effectiveness).