Diabetic foot prevention and management certainly result in high economic burden for the Peruvian health system and society as well. The overall COI for prevention and complications management with standard care strategy would amount to a total savings of roughly US$2.7 million and avert 791 deaths. Our findings suggest that secondary prevention strategies are cost-effective for the individuals and the government of Peru if properly implemented. We note that ICER values per disability adjusted life year (DALY) are usually compared with three times the Gross Domestic Product (GDP) per capita of a country, which for Peru was US$6,568 in 2012 . Our ICER using deaths averted is potentially more conservative given that disability benefits are not accounted for. ICER fell under three times the GDP per capita (US$16,124), we are confident it would be considered cost-effective .
Furthermore, standard care alone is even cost-saving, i.e. total costs decrease and deaths averted increase in comparison to sub-optimal care. When adding the indirect costs of lost productivity, the ICER for standard care versus sub-optimal care remains cost-saving and the standard care plus temperature monitoring versus sub-optimal care becomes more favorable, at US$9,405 per death averted. The sensitivity analysis showed that for the standard care only, the ICERs still fell under three times the GDP after varying the key parameters by +/−30 %.
From this analysis, a preventative strategy such as standard care not only improves the current level of care, but also leads to economic benefits in terms of societal cost-savings. Since prevention must be provided to a large number of patients at high-risk of ulceration, total costs of the prevention strategies are high; however, in the case of standard care plus temperature monitoring, the costs of thermometers would decline over time after the initial investment. For this strategy, even if the cost of the thermometer was zero, the strategy of standard care plus temperature monitoring would be cost-effective but not cost-saving. In the future, temperature monitoring technology may be integrated with electronic and mobile health prevention platforms that could further reduce the costs of personnel and phone services and make the intervention even more attractive.
Comparison with the literature
Direct annual costs per patient calculated for management of foot ulcers and other diabetic foot complications are lower in Peru than costs estimated in other studies for developed countries [22–25]. Ulcers that require amputation can cost, per admission, from US$15,790 [26, 27] to as high as US$45,870 [23, 27]. In contrast, a study in Brazil  used a hypothetical cohort including all the Brazilian citizens with Type 2 diabetes and estimated a total annual cost of hospital admissions per patient starting from US$2,151.
The reported costs of diabetic foot prevention and treatment in developing and developed nations are difficult to compare, further emphasizing the importance of studies such as ours. Previous studies in cost-effective interventions to prevent and control diabetes agree with our estimates that comprehensive and multidisciplinary foot care leads to cost-savings [28–33]. One systematic review  included two studies that dealt with diabetic foot prevention, and one of these studies used an optimal care strategy similar to our standard care strategy . An intensive prevention strategy including patient education, foot care and footwear was found to be cost-saving if the risk for foot ulcers and lower extremity amputations were reduced by 25 %  among those patients at risk of ulceration. In our study we assumed a reduction of ulceration rate of 45 % in line with lower bound estimates shown by the Centers for Disease Control and Prevention for potential reductions in amputation as a result of comprehensive foot care program [14, 15]. A retrospective cohort study from Austria using a Markov model, compared a dedicated screening program with conventional preventive care and concluded that the screening program would reduce costs by 29.8 % for mild ulcers and by 49.7 % for severe ulcers, primarily due to lower amputation rates . Ollendorf et al.  estimated an increasing economic benefit associated to educational interventions, multidisciplinary teams, and the therapeutic shoe coverage. They did not, however, include the costs of the underlying intervention strategies.
Recent published studies indicate that the use of temperature monitoring is an effective way to prevent diabetic foot ulceration; however the cost-effectiveness of temperature monitoring had not been yet studied [29, 30]. Our study provides relevant findings about the potential of standard care plus temperature monitoring as a cost-effective prevention strategy even in a short-term horizon. However, standard care as recommended by the IDF is still more cost-effective and provides net savings for the society.
This study was performed from a societal perspective considering all direct healthcare costs, regardless of who paid; however, we did not include more assumptions about travel costs and waiting time given the lack of reliable information for those estimates. In this study, we tried to be as conservative as possible when considering the most appropriate health outcome indicator or whether to include indirect costs into the analysis. We have only considered productivity losses as indirect cost calculated under the "human capital" approach. This approach generates large estimates, but we presented our results of the ICER both including and excluding the productivity costs. For the purpose of this cost-effectiveness study, we chose two tangible outcomes as health effect indicators: amputations and deaths. Further cost-utility analysis may extend this to looking at a summary health outcome that incorporates morbidity such as the quality adjusted life year (QALY) or DALY. The estimates in this study are also limited by the lack of available data on clinical outcomes of preventative strategies specific to diabetic foot in Peru. As a result, we at times used secondary information from other country settings while aiming to use data from developing countries.
Another limitation is the model assumption that healthcare utilization was 100 % and constant for the entire population. We considered this assumption given the severe condition of these kinds of patients and given the lack of information on this issue. However, we expect that a lower utilization rate would lead to more major complications and more fatal outcomes of those who do not receive timely medical care.
We also assume that the parameters apply for the whole country and do not make a distinction between rural and urban populations. It is likely that this decision contributes to an underestimation of the total COI because the rural population will most likely incur larger indirect costs due to travel and time to reach a hospital in urban area. Our study calculated an ulcer rate using the data from a clinical study that occurred over an 18-month period , because longitudinal data is not yet available.
We only investigated the secondary prevention methods for diabetic foot, and given the lack of data for the burden of Type 1 diabetes, we did not include a comprehensive overhaul of diabetes monitoring and prevention in order to not overestimate and attribute costs that are also related with other complications (retinopathy, kidney diseases, among others). We chose to provide a detailed breakdown of costs, yearly costs, and unit costs, each presented separately. To ensure the wide applicability of our study, we have offered a breakdown of the likely routes for resource allocation in a given year. We are aware that there are intangible benefits for the patients and their families that are impossible to measure in monetary terms. We also only looked at the economic burden of the disease for the first year assuming that each prevention strategy achieved full scale-up.
While aiming to produce a conservative estimate of the cost of ulceration and amputation, we left out the added risk for recurrent ulcers and new amputations, which could be included in the future development of a probabilistic Markov model.
Implications and further research
Diabetes prevalence in Peru is rising  and there is a low level of disease awareness and control . Thus monitoring and treatment protocols are urgently needed. The health and economic impacts of diabetic foot are exacerbated in Peru because of insufficient preventative measures. We recommend validating costs and collecting better epidemiological data for diabetic foot disease in Peru.