Background
Type 2 diabetes mellitus is a chronic metabolic disease which results not only in significant direct medical costs but indirect productivity losses due to disability and early mortality. For this reason, the future growth of type 2 diabetes is an increasing concern to researchers worldwide: in the U.S. for instance, type 2 diabetes is projected to almost double from a disease population of 23.7 million in 2009 to 44.1 million in 2034, resulting in a more-than-proportional tripling of diabetes-related spending from US$113 billion to US$336 billion [
1]. However, the composition of social costs varies significantly by population subgroups: while many analyses focus on the direct cost burden for older and retired individuals, in the working-age population, lost productivity can far exceed disease-related spending [
2].
To date, there have been few comprehensive analyses of current and future diabetes-related social costs in Asia, notably for Singapore. Globally, studies of national economic burden such as the American Diabetes Association (ADA) studies have typically been cross-sectional, focusing on the retrospective assessment of costs in a given year. At the same time, efforts to incorporate projection of costs from a societal perspective for the diabetes population have been observed in studies conducted in Australia, Canada, China, Colombia, Iran, IMS Health as well as the latest study from U.K. [
3‐
7]. In Singapore, the best available current prediction from the International Diabetes Federation (IDF) suggests health expenditures of US$0.83–1.48 billion in 2030 for Singapore. However, this figure is an aggregate estimate based on a cross-country model with no insight into potential indirect costs [
8]. The lack of comprehensive social cost studies for Asia is critical for regional decision makers, since comparison of national health expenditure is complicated by differences in disease epidemiology as well as healthcare financing systems [
9].
It is also critical to understand the changing dynamics of disease and costs: for instance, Asian populations are currently at higher risk of developing type 2 diabetes compared to others [
10], but are also seeing trends of earlier disease onset that may have strong implications for overall economic growth and employment.
In this study, we focus our attention on estimating the current and future total economic burden of diagnosed diabetes for the working-age population (i.e., 20–69 years old) in Singapore, which has the highest proportion of younger patients in the region and is also aging rapidly [
11]. We provide estimates of the composition of indirect and excess direct costs among this population, to assess potential justification for interventions within as well as beyond the health sector. We then project the estimated costs into the future by adopting the latest methodology described by the ADA, as well as using a published local micro-simulation disease model, the Demographic Epidemiological Model of Singapore (DEMOS) [
12,
13].
Discussion
Despite the growing epidemic of type 2 diabetes, to our knowledge, this study is the first to predict the diabetes-related social cost of the working-age population not just for Singapore, but possibly for any country in Asia from 2010 to 2050 [
35]. In 2010, Singapore’s GDP was US$223 billion, bringing the total diabetes-related economic cost among the working-age population to about 0.35 % of Singapore’s GDP [
36]. Considering that Singapore currently spends only 4 % of its GDP on healthcare expenditure, even in our conservative scenario, diabetes has imposed a significant economic burden on the national health care system and will continue to do so in the next four decades [
37].
Among this population, indirect costs are a significant driver of the total burden, and will increase in importance over time. Working-age males, especially those aged 50–59 in our analysis, were found to incur a higher total economic cost than females primarily because of the higher labor force participation rate of males in the workforce and a higher income than their female counterparts in the same age group [
16].
The results also showed that projected trends in healthcare spending are likely to continue, both in terms of total cost as well as the shifting emphasis on indirect costs. This suggests that prevention efforts are relevant not only to policymakers in the health sector, but also to policymakers and employers concerned with labor force productivity who will bear the “unseen” majority of diabetes cost in the future.
Finally, this study illustrates that differences in underlying disease and cost drivers as well as methodology can complicate comparisons across time. Although we have largely adopted the ADA methodology in our analysis, direct comparisons are not easily facilitated as our findings vary for both methodological and substantive reasons [
12]. For instance, firstly, Singapore’s total cost estimate of 0.35 % of GDP is low relative to the ADA estimate of 1.52 % for the U.S. [
38], and in the U.S., 72 % of the total economic burden for diabetes patients is direct medical costs, compared to 42 %. However, this disparity is partly due to the use of total versus excess direct medical costs, and also due to different levels of healthcare cost in Singapore versus the U.S. Secondly, excluding costs due to disability which were not available for Singapore, presenteeism was the highest cost driver in the U.S. [
12]. This trend was not observed in our study where the cost of lost capacity from premature mortality was the highest contributor, possibly due to the due to a higher proportion of older diabetes patients with an increasing mean age (58.2 years old in 2010 to 66.9 years old in 2050) and a higher average wage compared to other age groups.
There are several important limitations to this study, some of which are inherent to the ADA methodology which was adapted for comparability. For example, the methodology used to estimate the lost productivity from early mortality due to diabetes is not in concept consistent with the costing for the other components (which represent actual flows realized during the year of accounting) and gives estimates that are extremely sensitive to the assumptions. Ideally, productivity loss from premature mortality should be computed as the foregone value of productivity in a given year from individuals who died from diabetes in all prior years but would otherwise have been alive. However, the ADA (and our research team) did not implement this approach due to practical difficulty. This could yield very significant differences in calculated losses from diabetes attributed mortality from small changes in the discount rate applied. Most critically, the analysis relies on assumptions which were drawn from multiple sources, a number of which are based on assumptions adopted in the U.S. (and used in other international settings). Unfortunately, locally available primary data or secondary research are not yet sufficiently well developed in most cases to supply the necessary parameters for our analysis. We have addressed these limitations by estimating alternative scenarios where the data are available. In our baseline scenario, we adopt a consistently conservative approach, such that our figures may be viewed as lower bounds for costs. For example, the undiagnosed rate for diabetes was assumed to be constant in our analysis due to conflicting trends projected in literature [
1,
6,
39]. In addition, the assumption for excess direct medical cost here was it did not differ by age and gender based on results from a local study [
22] despite two previous studies which suggested otherwise [
40,
41]. Furthermore, we did not include the economic burden of those with undiagnosed diabetes as well as the cost of informal care which may contribute to a significant portion of indirect cost as observed by Hex et al. [
5]. This would in particular underestimate the cost of diabetes for women as previous studies have found that informal care is typically provided by women [
42].
In addition to being the first analysis of this kind for Asia to our knowledge, a key strength of the study is the generation of age- and gender-specific estimates, allowing us to assess the evolution of sub-group differences. A further key strength is the ability to assess the relative importance of the different indirect cost components. Both of these features allow for more nuanced policy interpretations for all stakeholders (including those outside the health sector) such as prioritizing the need to target specific groups, or specific interventions (e.g., reducing workplace presenteeism or addressing medical costs). Finally, in principle, this methodology can be replicated in other countries as well to generate regional estimates.
Work in progress in Singapore includes the gathering of disaggregated cost data (e.g., by ethnic and socioeconomic groups) to enable more accurate estimates of diabetes-related expenditures due to complications; obtaining better estimates of diabetes-related mortality and incorporating the comparative cost effectiveness of interventions like drug interventions or active lifestyle programmes into the model [
43]. Planned future work includes the collection of population-based data to allow better estimation of the effects of diabetes on employment and productivity. The inclusion of this data as well as the development of these improved estimates would help enhance the accuracy of the cost estimations presented in this study as well as help policymakers anticipate future costs of diabetes and determine effective public health interventions.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
MEP, JY and HLW conceived the study and its design. MEP and JY collected the data. MEP, TPP and JY contributed to the data management and analysis. MEP and JY were responsible for drafting the manuscript. All authors read, commented and approved the final version of this manuscript.