The global economic burden of diabetes for 190 countries 2021-50: a macroeconomic modelling study Authors: Jinxi Li, Ruxu Zhang, Ruixi Li, Qihua Song, Yanhong Gong, Chichen Zhang, Xiaoxv Yin Published: 2026 ABSTRACT Diabetes imposes a substantial health burden, yet its macroeconomic consequences remain incompletely quantified. Here we use a health-augmented macroeconomic model to estimate output losses attributable to diabetes across 190 countries and territories during 2021–2050. We project that diabetes could reduce cumulative global output by 5.177 trillion international dollars over this period, with substantial heterogeneity in relative losses as a share of GDP across countries. Country-level burdens vary markedly, reflecting heterogeneity in demographic trends, macroeconomic conditions, diabetes prevalence, health-system capacity and treatment costs. Most projected output loss is attributable to diabetes-related disability rather than premature mortality. In this work, we provide comparable cross-country estimates of the long-run macroeconomic burden of diabetes, informing prioritization of prevention, treatment policies, and related investments. Introduction Diabetes has become a global public health problem affecting 500 million people worldwide1. The global incidence of diabetes has increased every year since 1980 due to factors such as nutritional insecurity, physical inactivity, and air pollution2,3. Diabetes could increase all-cause mortality mainly through cardiovascular and renal diseases, and could also lead to a variety of other diseases such as blindness, loss of limbs, chronic pain and disability4. Deaths from diabetes have also been increasing over the last four decades, and diabetes is one of the leading ten causes of death globally5. Diabetes causes 1.66 million deaths in 2021, with the largest death toll in India, followed by China, Mexico, and the United States (US)3. The incidence, prevalence, and mortality for diabetes are shown in detail in (Supplementary Fig. 1-3). In addition to causing a massive health burden, diabetes also inflicts a large economic toll through reduced productivity, unemployment, labor losses, and capital investment reductions. Identifying the economic burden of diabetes and its distribution across regions and countries of the world enables policymakers to (1) enact appropriate policies to curb the increase in diabetes-related mortality and morbidity, (2) allocate resources appropriately, and (3) build health care systems that can cope effectively with expected increases in diabetes prevalence. The economic burden of diabetes covers both direct and indirect costs. Direct costs include all expenditures for the treatment of diabetes, including medications, hospitalization, and treatment of complications6. Indirect costs are defined as productivity or production losses associated with morbidity and premature mortality7. Much of the current research on the economic burden of diabetes has focused on individual countries such as the US, China, and Iran 8–10. The annual economic burden of diabetes in the United States, for example, was estimated at US$327 billion in 2017, including productivity losses, and has recently grown to over US$400 billion by 202210. In low-income countries, by contrast, the average cost per person with diabetes can be as low as a few hundred dollars per year. A 2017 global study estimated the cost of illness of diabetes at ~US$1.31 trillion (1.8% of global GDP) in 2015, with North America bearing the highest costs11. Out-of-pocket expenses also comprise the majority of diabetes treatment costs in many developing countries, exacerbating the economic burden on patients12. Several systematic reviews have analyzed regional and global costs, which confirmed substantial economic losses7,13,14. In summary, the lack of information on the future global economic burden of diabetes impedes the assessment of the effectiveness of diabetes prevention and treatment, as well as policy development. In terms of estimation methods, most studies assessing the economic burden of diabetes have simply summed direct and indirect costs (the cost-of-illness approach) without taking into account the impact of economic adjustment mechanisms8–10. On the one hand, the cost-of-illness approach ignores the fact that in the real-world jobs do not remain vacant indefinitely as companies recruit new people or replace lost labor with physical capital (machines or robots). On the other hand, the approach fails to account for changes in population dynamics driven by morbidity- and mortality-related factors, as well as changes in physical capital driven by the impact of diabetes on treatment costs and savings15. The omission of these two aspects often leads to an overestimation of the economic burden from human capital loss. The macroeconomic model of health gains constructed by Chen et al. effectively addresses these issues and is used to assess the global economic burden of cancer, COPD, and road injury16–18. Based on this, this study used a health-augmented macroeconomic model to estimate the macroeconomic burden of diabetes in 190 countries and territories around the world from 2021–2050, to provide data to support the promotion of global diabetes management. Results We calculated the global macroeconomic burden of diabetes as the sum of the GDP difference between the status quo and the counterfactual scenario of eliminating diabetes for 2021–2050. Table 1 summarizes the economic burden for the 126 countries/areas with complete data and the 64 countries with missing data. The former accounts for a total of 86.0% of the global population and 95.3% of GDP. The US has the largest economic burden of diabetes, followed by China and India. In terms of share of GDP, Kuwait, the US, and South Korea accounted for the largest share (0.34 % each). Guyana, the US, and Singapore had the highest economic burden per capita (Supplementary Table 2 and Fig. 3). The total macroeconomic burden and economic burden as a share of GDP for each country are displayed in Figs. 1 and 2.Fig. 1Macroeconomic burden due to diabetes from 2021 to 2050.This map shows the total macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050 (in billions of 2017 international dollars (INT$)), based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger absolute economic losses; gray indicates missing data (“data missing”). Countries with higher absolute burdens are mainly concentrated in large-population and/or large-economy settings (for example, the United States, China, and India), whereas the absolute burden is relatively smaller in most low-income countries. This figure illustrates the global spatial heterogeneity and concentration of diabetes-related economic losses.Fig. 2Economic burden of diabetes as a proportion of total GDP in 2021–2050.This map shows the macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050, expressed as the percentage (%) of total GDP over the same period, based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger relative losses (as a share of GDP); gray indicates missing data (“data missing”). Countries and territories with higher relative burdens are typically concentrated in some small island settings and several high-income economies, whereas many countries in Africa and South Asia show lower relative burdens. This figure illustrates the uneven distribution of diabetes-related economic losses across world regions.Table 1Total macroeconomic burden, economic burden as a proportion of total GDP, and per capita economic burden attributable to diabetes mortality and morbidity in 2020–50 by World Bank region, by World Bank income group, and globallyEconomic loss, billions of 2017 INT$ (uncertainty interval*)Proportion of total GDP in 2021–50, × 10−3 % (uncertainty interval*)Per capita loss, 2017 INT$ (uncertainty interval*)By World Bank regionNorth America1862 (1853–1871)313 (311–314)4648 (4626–4669)East Asia and Pacific1572 (1482–1679)95 (90–102)660 (622–705)Europe and central Asia829 (812–848)95 (93–98)897 (878–917)Middle East and north Africa319 (304–337)139 (132–147)587 (559–621)South Asia274 (245–310)48 (43–55)127 (114–144)Latin America and Caribbean260 (252–269)105 (102–109)372 (361–386)Sub-Saharan Africa60 (51–71)42 (36–50)36 (31–43)By World Bank country income groupHigh income3217 (3183–3253)200 (198–203)2549 (2521–2577)Upper-middle income1427 (1338–1533)82 (77–88)502 (470–539)Lower-middle income512 (461–574)57 (51–63)138 (125–155)Low income19 (16–23)41 (34–49)21 (17–25)Global (190 countries and territories)5177 (5000–5384)120 (116–125)591 (571–615)By intermediate Scenario10%82 (66–102)2 (1–2)9 (7–12)20%190 (159–228)4 (3–5)21 (18–26)GDP=gross domestic product.*Uncertainty intervals in parentheses are calculated based on the lower and upper bounds of 95% uncertainty intervals for Global Burden of Disease Study 2021 mortality and morbidity data. Macroeconomic burden due to diabetes from 2021 to 2050. This map shows the total macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050 (in billions of 2017 international dollars (INT$)), based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger absolute economic losses; gray indicates missing data (“data missing”). Countries with higher absolute burdens are mainly concentrated in large-population and/or large-economy settings (for example, the United States, China, and India), whereas the absolute burden is relatively smaller in most low-income countries. This figure illustrates the global spatial heterogeneity and concentration of diabetes-related economic losses. Economic burden of diabetes as a proportion of total GDP in 2021–2050. This map shows the macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050, expressed as the percentage (%) of total GDP over the same period, based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger relative losses (as a share of GDP); gray indicates missing data (“data missing”). Countries and territories with higher relative burdens are typically concentrated in some small island settings and several high-income economies, whereas many countries in Africa and South Asia show lower relative burdens. This figure illustrates the uneven distribution of diabetes-related economic losses across world regions. Table 1Total macroeconomic burden, economic burden as a proportion of total GDP, and per capita economic burden attributable to diabetes mortality and morbidity in 2020–50 by World Bank region, by World Bank income group, and globallyEconomic loss, billions of 2017 INT$ (uncertainty interval*)Proportion of total GDP in 2021–50, × 10−3 % (uncertainty interval*)Per capita loss, 2017 INT$ (uncertainty interval*)By World Bank regionNorth America1862 (1853–1871)313 (311–314)4648 (4626–4669)East Asia and Pacific1572 (1482–1679)95 (90–102)660 (622–705)Europe and central Asia829 (812–848)95 (93–98)897 (878–917)Middle East and north Africa319 (304–337)139 (132–147)587 (559–621)South Asia274 (245–310)48 (43–55)127 (114–144)Latin America and Caribbean260 (252–269)105 (102–109)372 (361–386)Sub-Saharan Africa60 (51–71)42 (36–50)36 (31–43)By World Bank country income groupHigh income3217 (3183–3253)200 (198–203)2549 (2521–2577)Upper-middle income1427 (1338–1533)82 (77–88)502 (470–539)Lower-middle income512 (461–574)57 (51–63)138 (125–155)Low income19 (16–23)41 (34–49)21 (17–25)Global (190 countries and territories)5177 (5000–5384)120 (116–125)591 (571–615)By intermediate Scenario10%82 (66–102)2 (1–2)9 (7–12)20%190 (159–228)4 (3–5)21 (18–26)GDP=gross domestic product.*Uncertainty intervals in parentheses are calculated based on the lower and upper bounds of 95% uncertainty intervals for Global Burden of Disease Study 2021 mortality and morbidity data. Total macroeconomic burden, economic burden as a proportion of total GDP, and per capita economic burden attributable to diabetes mortality and morbidity in 2020–50 by World Bank region, by World Bank income group, and globally GDP=gross domestic product. *Uncertainty intervals in parentheses are calculated based on the lower and upper bounds of 95% uncertainty intervals for Global Burden of Disease Study 2021 mortality and morbidity data. Table 2 presents the results of the World Bank study for each region and country income group. We estimated the global macroeconomic burden of diabetes to be INT$ 5.177 trillion (5.000–5.384) in 2021–2050, with a discount rate of 3% in the main analysis Fig. 3. Without discounting, this estimate would be INT$ 9.550 trillion (confidence interval 9.227–9.927); with a discount rate of 2% it would be INT$ 6.306 trillion (6.091–6.557); and with a discount rate of 5% it would be INT$ 3.562 trillion (3.438–3.706) (Supplementary 5). Our main results suggest that the burden of diabetes is equivalent to 0.120% (confidence interval 0.116–0.125) of total global cumulative GDP, or to a per-capita burden of INT$591 (571–615). In the intermediate scenario analysis, we found that a 20% reduction in diabetes burden by 2050 would yield global savings of INT$190 billion (159–228). Under the counterfactual scenario, reducing diabetes-related incidence and mortality to the lowest observed levels globally was estimated to yield a global macroeconomic gain of INT$1.11 trillion over 2021–2050 (Supplementary Table 13).Fig. 3Per capital burden due to diabetes from 2021 to 2050.This map shows the per-capita macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050 (in 2017 international dollars (INT$) per capita), based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger per-capita economic losses; gray indicates missing data (“data missing”). Countries and territories with higher per-capita burdens are more common in some high-income economies and several small-population settings, whereas per-capita burdens are relatively lower in most low-income countries. This figure highlights pronounced regional differences in diabetes-related economic losses on a per-capita basis.Table 2Comparison of macroeconomic burden and lifetime health burden of diabetes, by World Bank region and country income groupEconomic burden in 2021–50, billions of 2017 INT$ (global %)DALYs in 2021, millions (global %)Total GDP in 2021, billions of 2017 INT$ (global %)Population in 2021, millions (global %)By World Bank regionNorth America1862 (36%)5.5 (7%)22945 (17.2%)375 (4.8%)East Asia and the Pacific1572 (30.4%)22.8 (28.9%)43588 (32.6%)2346 (30.1%)Europe and central Asia829 (16%)9.4 (11.9%)32889 (24.6%)921 (11.8%)Middle East and North Africa319 (6.2%)5.1 (6.5%)7757 (5.8%)456 (5.8%)South Asia274 (5.3%)18.6 (23.6%)12100 (9.1%)1893 (24.3%)Latin America and the Caribbean260 (5%)9.7 (12.3%)9942 (7.4%)641 (8.2%)Sub-Saharan Africa60 (1.2%)7.7 (9.8%)4369 (3.3%)1163 (14.9%)By World Bank country income groupHigh income3217 (62.2%)14.9 (19%)61915 (92.8%)1233 (31.7%)Upper-middle income1427 (27.6%)29.4 (37.5%)48683 (73%)2771 (71.4%)Lower-middle income512 (9.9%)30 (38.3%)21679 (32.5%)3137 (80.8%)Low income19 (0.4%)4 (5.1%)1154 (1.7%)626 (16.1%)Global (204 countries and territories)5177 (100%)78.8 (100%)133589 (100%)7795 (100%)DALY disability-adjusted life-year, GDP gross domestic product. Per capital burden due to diabetes from 2021 to 2050. This map shows the per-capita macroeconomic burden of diabetes across 190 countries and territories from 2021 to 2050 (in 2017 international dollars (INT$) per capita), based on a health-augmented macroeconomic model. The burden estimates reflect losses in effective labor supply due to diabetes-related mortality and morbidity, as well as the diversion of resources to treatment away from savings and investment. Countries and territories are shaded by burden level, with darker colors indicating larger per-capita economic losses; gray indicates missing data (“data missing”). Countries and territories with higher per-capita burdens are more common in some high-income economies and several small-population settings, whereas per-capita burdens are relatively lower in most low-income countries. This figure highlights pronounced regional differences in diabetes-related economic losses on a per-capita basis. Table 2Comparison of macroeconomic burden and lifetime health burden of diabetes, by World Bank region and country income groupEconomic burden in 2021–50, billions of 2017 INT$ (global %)DALYs in 2021, millions (global %)Total GDP in 2021, billions of 2017 INT$ (global %)Population in 2021, millions (global %)By World Bank regionNorth America1862 (36%)5.5 (7%)22945 (17.2%)375 (4.8%)East Asia and the Pacific1572 (30.4%)22.8 (28.9%)43588 (32.6%)2346 (30.1%)Europe and central Asia829 (16%)9.4 (11.9%)32889 (24.6%)921 (11.8%)Middle East and North Africa319 (6.2%)5.1 (6.5%)7757 (5.8%)456 (5.8%)South Asia274 (5.3%)18.6 (23.6%)12100 (9.1%)1893 (24.3%)Latin America and the Caribbean260 (5%)9.7 (12.3%)9942 (7.4%)641 (8.2%)Sub-Saharan Africa60 (1.2%)7.7 (9.8%)4369 (3.3%)1163 (14.9%)By World Bank country income groupHigh income3217 (62.2%)14.9 (19%)61915 (92.8%)1233 (31.7%)Upper-middle income1427 (27.6%)29.4 (37.5%)48683 (73%)2771 (71.4%)Lower-middle income512 (9.9%)30 (38.3%)21679 (32.5%)3137 (80.8%)Low income19 (0.4%)4 (5.1%)1154 (1.7%)626 (16.1%)Global (204 countries and territories)5177 (100%)78.8 (100%)133589 (100%)7795 (100%)DALY disability-adjusted life-year, GDP gross domestic product. Comparison of macroeconomic burden and lifetime health burden of diabetes, by World Bank region and country income group DALY disability-adjusted life-year, GDP gross domestic product. Among the World Bank regions, the macroeconomic burden of diabetes was highest in North America, which also had the highest per capita losses, followed by East Asia and the Pacific (Table 2). North America also had the largest economic losses as a proportion of output, equivalent to 0.313% of tax revenue (uncertainty interval 0.311–0.314). As expected, the economic burden of diabetes was positively correlated with country income groups, as well as with total gross domestic product (GDP) and the proportion of economic losses per capita: high-income countries bore the largest burden, with total economic losses of INT$ 3.217 trillion (3.183–3.253) and losses per capita of INT$ 2549 (2511–2577). In contrast, diabetes was estimated to cost low-income countries INT$19 billion (16–23) in total and INT$21 (17–25) per capita. The Supplementary 5 provides estimate under alternative discount rates by World Bank region and World Bank income group. Diabetes resulted in 78.8 million disability-adjusted life-years (DALYs) worldwide in 2021. The distribution of the economic burden was not proportional to population size and DALY. For example, North America accounted for only 7.0% of the global DALY but 36.0% of the global economic burden, while South Asia accounted for 23.6% of the global DALY and only 5.3% of the economic burden. Notably, although the economic burden of diabetes was relatively low in low- and middle-income countries (37.8 % of the global GDP), the burden of disease as measured by the DALY was very high (81.0% of the global DALY; Table 2). Discussion This study uses macroeconomic modeling to comprehensively estimate the macroeconomic burden of diabetes in 190 countries worldwide. This study extends existing macroeconomic burden frameworks by incorporating productivity losses that vary with individuals’ educational attainment and work experience. Our study suggests that diabetes costs a total of INT $5.177 trillion between 2021 and 2050, exceeding Japan's 2021 GDP (the world’s fourth-largest economy in constant 2017 international dollars). The importance of this study lies in supporting a shift in how policymakers evaluate chronic disease prevention and control. Traditionally, such efforts have often been viewed as a financial burden; however, by redefining health as an investment in human capital with quantifiable returns, we transform the economic impact of diabetes into macro indicators that decision-makers can directly utilize (including total loss, percentage of cumulative GDP, and per capita burden). This suggests the potential for economic growth and fiscal space that effective prevention and control may unlock. Therefore, diabetes management should not be framed solely as expenditure, but rather as a strategic investment with long-term economic returns. This shift in perspective facilitates greater attention to diabetes prevention at the policy level, fostering effective integration and sustained resource investment. Low-income and middle-income countries account for 81% of the global diabetes DALY burden and 85% of the global population, but only 37.8% of the global economic burden of diabetes. This discrepancy reflects substantial cross-country heterogeneity not only in diabetes prevalence, but also in health system capacity, treatment coverage, and cost structures. Large-scale global analyses have documented marked variation in diabetes prevalence across countries and regions, shaped by differences in demographic profiles, urbanization, lifestyle risk factors, and socioeconomic conditions19. At the same time, diabetes detection, treatment, and control vary widely across health systems: treatment coverage exceeds 50% in many high-income countries but remains below 10% in several low-resource settings, particularly in parts of sub-Saharan Africa and South Asia20. These disparities in health system performance and access to effective care have important economic implications. Countries with more developed health systems and broader treatment coverage tend to incur higher per-capita economic losses from diabetes, driven by higher medical expenditures and greater opportunity costs associated with lost productivity in higher-income labor markets. By contrast, in countries with limited treatment access, the measured macroeconomic burden may appear smaller despite a substantial underlying disease burden, reflecting lower treatment intensity, lower unit costs, and different labor-market structures. Together, these patterns underscore that cross-country differences in the economic burden of diabetes are shaped not only by disease prevalence, but also by systematic differences in health system organization, treatment coverage, and the cost profile of diabetes care20,21. National-level differences are also pronounced. For example, Mexico’s per capita GDP loss (approximately INT$567) is about four times that of India (INT$133), even though India has a higher number of diabetes cases and a greater overall disease burden. This disparity can largely be attributed to the stark contrast in health expenditures between the two countries: India’s per capita health expenditure is only INT$236, whereas Mexico’s is as high as 1190. Similarly, Pakistan—another middle-income country with a high diabetes prevalence—has a per capita health expenditure of only INT$167, and a correspondingly low per capita economic loss. By contrast, small high-income countries such as Kuwait (with per capita health expenditures of INT$2,908) and Singapore (INT$6,353) exhibit extremely high per capita GDP losses. These patterns suggest that in countries with stronger economies and higher baseline healthcare costs, the per capita economic burden of diabetes tends to be heavier. This is not only due to broader treatment coverage and higher service prices, but may also be compounded by the added complexities of chronic disease management associated with an aging population21,22. Conversely, countries with lower per capita health expenditures — even if they face a more severe disease burden — experience smaller per capita macroeconomic losses. This may be explained by factors such as limited treatment intensity, lower labor opportunity costs, or a younger population age structure in these countries20. Such disparities underscore the highly heterogeneous economic impact of diabetes, reflecting systematic differences across countries in population age structure, the level of diabetes management, and the cost structures of healthcare systems. Given the substantial impact of diabetes on global health and economies, we underscore the need to invest in effective public health interventions to mitigate future burden. A coherent strategy should be grounded in supportive macro-level policies, establishing a continuum of actions spanning prevention, treatment, and long-term management. At the level of prevention, we propose building an integrated care system centered on primary health care, leveraging cross-sectoral collaboration to promote healthy lifestyles and early intervention so as to reduce diabetes risk at its source23. International agencies such as the WHO highlight that multisectoral collaboration is critical for diabetes prevention and control; in parallel, the latest guidelines from the International Diabetes Federation identify weight management, physical activity, and healthy nutrition education as core elements of prevention strategies24,25. At the level of treatment, we argue that strengthening primary health care systems should be a priority, enabling them to provide continuous, standardized diabetes care and to ensure the availability and affordability of essential medicines26,27. This approach is aligned with WHO’s goal of fully integrating diabetes into primary health care and expanding coverage of diagnostic and treatment services28. For long-term management, we recommend the use of technology-enabled approaches (such as digital health tools and telemedicine) to enhance patients’ self-management capacity and the efficiency of chronic disease management. Such technological integration is expected to strengthen patient–provider interactions, facilitate monitoring of disease progression and support individualized interventions, thereby effectively preventing complications29. As technologies advance, emerging tools including artificial intelligence, digital health, and telemedicine can improve the efficiency of diabetes prevention, treatment, and management, while also expanding equitable access to services. These technologies can help narrow gaps in health care arising from urban-rural residence, gender and socioeconomic status, thereby advancing health equity in the context of diabetes30. At the macro level, we recommend that governments regard diabetes prevention and control as a critical health investment in macro-policy planning. By implementing primary care-centered integrated care, strengthening cross-sectoral collaboration and adopting technology-driven management frameworks, countries may be better positioned to curb the epidemic of diabetes and its broader socioeconomic consequences. Our study has several limitations. First, we did not consider changes in workforce participation rates of household members who may be required to care for a person with diabetes and, therefore, may have underestimated the economic burden of diabetes. Second, due to missing data, we calculated diabetes-related health expenditures, workforce participation, mortality, morbidity, and other indicators for 64 countries using available data, with interpolation and projection. However, the countries with complete data in this study account for more than 90% of the global population and 95% of global GDP, so any exclusion due to data gaps is unlikely to substantially affect the overall results. Third, we extrapolated diabetes treatment costs to all countries based on U.S. and other high-income country data, assuming that per-case costs are proportional to per capita health expenditures. Although widely used in global health analyses and avoiding naive cross-country transfer of unit cost estimates, this approach remains a necessary simplification due to the scarcity of country-specific cost data. Fourthly, our framework focuses on health-induced changes using standard growth inputs and therefore does not capture potential secular labor-market shifts (e.g., accelerated automation or changes in human-capital utilization) beyond those embedded in the input data. This scope limitation should be considered when interpreting long-run projections, and future work could incorporate scenarios with exogenous labor-share declines or technology-driven productivity changes. Additionally, our modeling approach required certain simplifying assumptions. For example, to project morbidity impacts we assumed a proportional relationship between YLD and YLL for diabetes (Supplementary 2, Eq. 7). In reality, the conversion of morbidity into lost productivity is more complex, and a fixed YLD:YLL ratio may oversimplify the impact of non-fatal diabetes complications on economic output. This assumption, along with others (such as holding certain demographic patterns constant), limits the model’s precision. Finally, our estimates inherit the uncertainty and potential bias of GBD inputs (prevalence, mortality, YLL/YLD), which vary across countries and over time and therefore propagate into our macroeconomic results. In addition, our treatment-cost calibration assumes a constant disease-specific proportionality to per-capita health expenditure based largely on US spending shares; even with PPP adjustment, this may overstate absolute costs in lower-cost systems, so we prioritize relative metrics (share of GDP; per-capita loss) and interpret country point estimates cautiously. In summary, the global macroeconomic burden of diabetes is substantial, and our study provides a comparable cross-country perspective. Using a macroeconomic model that accounts for labor and capital adjustments, we translate the impacts of diabetes into long-run output losses and characterize how these losses arise through reduced labor supply and reduced savings and capital accumulation, and how they are distributed across countries. These estimates clarify the economic stakes of diabetes control: high-income countries incur larger absolute losses, whereas low- and middle-income countries incur larger proportional losses, underscoring the need for context-specific strategies. Overall, our results support viewing diabetes prevention and treatment not only as a health priority but also as an economic investment. Policy responses should be tailored to local contexts—high-income countries can mitigate large absolute losses by scaling up prevention and care, whereas low-income regions may require continued international support to strengthen basic prevention and care infrastructure. Methods Model description We estimated the macroeconomic burden of diabetes across countries and regions with the health-augmented macroeconomic model, which has been described in detail Supplementary 2. The definition of diabetes follows the Global Burden of Disease (GBD) Study’s diabetes category3. In the health-augmented macroeconomic model, diabetes affects through the following two channels. On the one hand, both diabetes-related morbidity and mortality reduce the effective labor supply. The morbidity of diabetes reduces productivity and thus increases absenteeism, and mortality due to diabetes reduces the size of the working-age population. On the other hand, any cost associated with diabetes treatment reduces total savings and investment, thus hindering physical capital accumulation. Families affected by diabetes need to divert resources from savings to cover out-of-pocket expenses or increase premiums to pay for treatment through insurance companies (Fig. 4). Notably, a shift in consumption from other goods and services to health care does not constitute an economic loss, but merely a reallocation of resources within the economy. This is because a shift from consumption to health expenditure does not reduce the economy's productive capacity. In contrast, shifting funds from physical capital investment to health-care consumption reduces the prospects for future capital growth.Fig. 4Summary of Pathways from Diabetes to Economic Growth.Conceptual diagram summarizing the channels through which diabetes affects economic growth in the model. Diabetes increases mortality and morbidity, reducing effective labor supply (through a smaller working-age population and/or reduced productivity/labor participation). Diabetes-related medical expenditure diverts resources from savings, thereby reducing physical capital investment/accumulation, further constraining output. Reduced labor supply and reduced physical capital investment jointly lower economic growth (GDP) over time, with interactions between labor and capital captured by the aggregate production function. Summary of Pathways from Diabetes to Economic Growth. Conceptual diagram summarizing the channels through which diabetes affects economic growth in the model. Diabetes increases mortality and morbidity, reducing effective labor supply (through a smaller working-age population and/or reduced productivity/labor participation). Diabetes-related medical expenditure diverts resources from savings, thereby reducing physical capital investment/accumulation, further constraining output. Reduced labor supply and reduced physical capital investment jointly lower economic growth (GDP) over time, with interactions between labor and capital captured by the aggregate production function. To quantify the macroeconomic burden of diabetes over the period 2021–2050, we constructed economic projections for two scenarios: (1) total output Yt in the status quo scenario without additional interventions to reduce mortality and morbidity of diabetes in year t; and (2) total output \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{{{{\rm{Y}}}}_{{{\rm{t}}}}}$$\end{document}Yt¯ in the counterfactual scenario of eliminating diabetes at zero cost in year t. Economic forecasting using the macroeconomic production function can be further divided into two parts: (a) projections of effective labor supply, and (b) projections of physical capital accumulation. The macroeconomic burden of diabetes was defined as the cumulative difference in projected annual GDP between these two scenarios.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {{\rm{Y}}}={\sum }_{{{\rm{t}}}=2021}^{2050}(\bar{{{{\rm{Y}}}}_{{{\rm{t}}}}}-{{{\rm{Y}}}}_{{{\rm{t}}}})$$\end{document}ΔY=∑t=20212050(Yt¯−Yt) Consider an economy in which time t = 1,2, …, ∞ evolves discretely. we considered the following production function for this economy:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{Y}}}}_{{{\rm{t}}}}={{{\rm{A}}}}_{{{\rm{t}}}}{{{\rm{K}}}}_{{{\rm{t}}}}^{{{\rm{\alpha }}}}{{{\rm{H}}}}_{{{\rm{t}}}}^{1-{{\rm{\alpha }}}}$$\end{document}Yt=AtKtαHt1−αwhere Yt is aggregate output; At is the technological level at time t, which we assume evolves exogenously; Kt is the physical capital stock (i.e., machines, factory buildings, etc.); and Ht represents aggregate human capital. The parameter α is the elasticity of final output with respect to physical capital. Solow’s framework, upon which the World Health Organization’s original EPIC macroeconomic model is based31, only considers physical capital and raw labor as factors of output production32. However, the aggregate production function recognizes that output is not only produced with these factors but also with effective labor, including health, which is a crucial determinant. Physical capital evolves according to3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{K}}}}_{{{\rm{t}}}+1}=(1-{{\rm{\delta }}}){{{\rm{K}}}}_{{{\rm{t}}}}+{{{\rm{Y}}}}_{{{\rm{t}}}}-{{{\rm{C}}}}_{{{\rm{t}}}}-{{{\rm{TC}}}}_{{{\rm{t}}}}=(1-{{\rm{\delta }}}){{{\rm{K}}}}_{{{\rm{t}}}}+{{{\rm{\gamma }}}}_{{{\rm{t}}}}{{{\rm{Y}}}}_{{{\rm{t}}}},$$\end{document}Kt+1=(1−δ)Kt+Yt−Ct−TCt=(1−δ)Kt+γtYt,where δ refers to the depreciation rate, γ refers to the saving rate, TC refers to the costs of ongoing treatment of diabetes, and C refers to the amount of consumption. Note that aggregate output Yt is used for three purposes: (1) to pay treatment costs (hospitalization, medication, etc.), (2) to consume the amount, and (3) to save. Individuals of age group a are endowed with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{h}}}}_{{{\rm{t}}}}^{({{\rm{\alpha }}})}$$\end{document}ht(α) units of human capital and supply \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{l}}}}_{{{\rm{t}}}}^{({{\rm{\alpha }}})}$$\end{document}lt(α) units of labor from age 15 up to their retirement at age R, i.e., for a ∈ [15, R]. Children younger than 15 and retirees older than R do not work. R varies by country and could correspond to a high age (e.g., some people older than 80 could also be working). In the theoretical derivations, R indicates the upper bound of the summation. In our simulations, we used labor projections data from the International Labor Organization, where positive values for the labor force exist for cohorts older than 65. Aggregate human capital in the production function (1) is then defined as the sum over the age-specific effective labor supply of each age group:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{H}}}}_{{{\rm{t}}}}={\sum }_{{{\rm{a}}}=15}^{{{\rm{R}}}}{{{\rm{h}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}{{{\rm{l}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}{{{\rm{N}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)},$$\end{document}Ht=∑a=15RhtaltaNta,where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{N}}}}_{{{\rm{t}}}}^{{{\rm{a}}}}$$\end{document}Nta denotes the number of individuals in age group a. Note that aggregate human capital increases with the number of working-age individuals who live in the economy (i.e., with a higher \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{N}}}}_{{{\rm{t}}}}={\sum }_{{{\rm{a}}}=15}^{{{\rm{R}}}}{{{\rm{N}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}$$\end{document}Nt=∑a=15RNt(a)), with individual human capital endowment (i.e., with a higher \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{h}}}}_{{{\rm{t}}}}^{{{\rm{a}}}}$$\end{document}hta for at least one a), and with labor supply (i.e., with a higher \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{l}}}}_{{{\rm{t}}}}^{{{\rm{a}}}}$$\end{document}lta for at least one a). This human capital function itself allows for substituting capital (machines or robots) for workers. In addition, if the disease predominantly affects one age group, then workers from other age groups can substitute. We followed Mincer33 and constructed the average human capital of the cohort aged a according to an exponential function of education and work experience:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{h}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}=\exp \left[{{{\rm{\eta }}}}_{1}\left({{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}\right)+{{{\rm{\eta }}}}_{2}\left({{\rm{a}}}-{{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}-5\right)+{{{\rm{\eta }}}}_{3}{\left({{\rm{a}}}-{{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}-5\right)}^{2}\right]$$\end{document}hta=expη1yeta+η2a−yeta−5+η3a−yeta−52where η1 is the semi-elasticity of human capital with respect to average years of education as given by \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}$$\end{document}yeta, and η2 and η3 are the semi-elasticities of human capital with respect to experience of the workforce \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left({{\rm{a}}}-{{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}-5\right)$$\end{document}a−yeta−5 and experience of the workforce squared \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\left({{\rm{a}}}-{{\rm{y}}}{{{\rm{e}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}-5\right)}^{2}$$\end{document}a−yeta−52, respectively. Here, we assumed a school entry age of 5 years throughout. Diabetes impedes physical capital accumulation because savings finance part of the treatment costs. Physical capital accumulation in the counterfactual scenario can be written as6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{{{\rm{K}}}}}_{{{\rm{t}}}+1}={\bar{{{\rm{\gamma }}}}}_{{{\rm{t}}}}{\bar{{{\rm{Y}}}}}_{{{\rm{t}}}}+\left(1-{{\rm{\delta }}}\right){\bar{{{\rm{K}}}}}_{{{\rm{t}}}},$$\end{document}K¯t+1=γ¯tY¯t+1−δK¯t,7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{{{\rm{\gamma }}}}}_{{{\rm{t}}}}{\bar{{{\rm{Y}}}}}_{{{\rm{t}}}}={\bar{{{\rm{I}}}}}_{{{\rm{t}}}}={\bar{{{\rm{Y}}}}}_{{{\rm{t}}}}-{\bar{{{\rm{C}}}}}_{{{\rm{t}}}}={{{\rm{\gamma }}}}_{{{\rm{t}}}}{\bar{{{\rm{Y}}}}}_{{{\rm{t}}}}+{{\rm{\chi }}}{{\rm{T}}}{{{\rm{C}}}}_{{{\rm{t}}}}.$$\end{document}γ¯tY¯t=I¯t=Y¯t−C¯t=γtY¯t+χTCt.where χ is the fraction of the treatment cost that is diverted to savings. Diabetes influences the accumulation of human capital by reducing the population and labor participation rate. the evolution of labor supply in the status quo scenario is given by8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{L}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}={{{\rm{l}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}{{{\rm{N}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}{{\rm{with}}} \, {{{\rm{N}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}=\left[1-{{{\rm{\sigma }}}}_{{{\rm{t}}}-1}^{\left({{\rm{a}}}-1\right)}\right]{{{\rm{N}}}}_{{{\rm{t}}}-1}^{({{\rm{a}}}-1)},$$\end{document}Lt(a)=lt(a)Nt(a)withNt(a)=1−σt−1a−1Nt−1(a−1),where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}$$\end{document}σta is the overall mortality rate of age group a at time t. Mortality and morbidity reduce effective labor supply. The reduction of the population size \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{N}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}$$\end{document}Nt(a) captures the mortality effect. Let \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{{{\rm{r}}},{{\rm{t}}}}^{\left({{\rm{a}}}\right)}$$\end{document}σr,ta denote the mortality rate of people in age group a due to diabetes and let \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\sigma }}}}_{-{{\rm{r}}},{{\rm{t}}}}^{\left({{\rm{a}}}\right)}$$\end{document}σ−r,ta be the overall mortality rate due to causes other than diabetes, then we have9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left(1-{{{\rm{\sigma }}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}\right)=\left(1-{{{\rm{\sigma }}}}_{{{\rm{r}}},{{\rm{t}}}}^{\left({{\rm{a}}}\right)}\right)\left(1-{{{\rm{\sigma }}}}_{-{{\rm{r}}},{{\rm{t}}}}^{\left({{\rm{a}}}\right)}\right).$$\end{document}1−σta=1−σr,ta1−σ−r,ta. In the counterfactual scenario, in which we denote variables with an overbar, the size of the cohort aged a at time t (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\,\bar{{{\rm{N}}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}$$\end{document}N¯ta) evolves according to10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{{{\rm{N}}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}=\left[1-{{{\rm{\sigma }}}}_{-{{\rm{r}}},{{\rm{t}}}-1}^{\left({{\rm{a}}}-1\right)}\right]{\,\bar{{{\rm{N}}}}}_{{{\rm{t}}}-1}^{\left({{\rm{a}}}-1\right)}.$$\end{document}N¯ta=1−σ−r,t−1a−1N¯t−1a−1. Then the loss of labor due to mortality accumulates over the years according to11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\,\overline{{{\rm{N}}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}={{{\rm{N}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}/{\prod }_{{{\rm{\tau }}}=0}^{\min \left\{{{\rm{t}}},{{\rm{a}}}\right\}-1}[1-{{{\rm{\sigma }}}}_{{{\rm{r}}},{{\rm{t}}}-1-{{\rm{\tau }}}}^{\left({{\rm{a}}}-1-{{\rm{\tau }}}\right)}].$$\end{document}N¯ta=Nta/∏τ=0mint,a−1[1−σr,t−1−τa−1−τ]. The reduction of the labor participation rate \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{l}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}$$\end{document}lt(a) captures the morbidity effect because people with an illness typically reduce their labor supply, either by reducing working hours or by leaving the workforce. The labor participation rate in the counterfactual scenario \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{{{\rm{l}}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}$$\end{document}l¯t(a) can be calculated as12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{{{\rm{l}}}}}_{{{\rm{t}}}}^{({{\rm{a}}})}\approx {{{\rm{l}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}/{\prod }_{{{\rm{\tau }}}=0}^{\min \left\{{{\rm{t}}},{{\rm{a}}}\right\}-1}[1-{{{\rm{p}}}}^{{{\rm{\tau }}}}{{{\rm{\sigma }}}}_{{{\rm{r}}},{{\rm{t}}}-1-{{\rm{\tau }}}}^{\left({{\rm{a}}}-1-{{\rm{\tau }}}\right)}{{{\rm{\xi }}}}^{\left({{\rm{a}}}-1-{{\rm{\tau }}}\right)}],$$\end{document}l¯t(a)≈lta/∏τ=0mint,a−1[1−pτσr,t−1−τa−1−τξa−1−τ],where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\xi }}}}^{\left({{\rm{a}}}\right)}$$\end{document}ξa measures the size of the morbidity effect relative to the relevant mortality rate and where\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\,{{{\rm{p}}}}^{{{\rm{\tau }}}}$$\end{document}pτ is the probability that a patient died from diabetes by time t. Because the impact of morbidity is hard to estimate directly, we first defined13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\xi }}}}^{\left({{\rm{a}}}\right)}=\frac{{{\rm{loss}}}\; {{\rm{of}}}\; {{\rm{labor}}}\; {{\rm{due}}}\; {{\rm{to}}}\; {{\rm{morbidity}}}\; {{\rm{in}}}\; {{\rm{age}}}\; {{\rm{group}}}\; {{\rm{a}}}}{{{\rm{loss}}}\; {{\rm{of}}}\; {{\rm{labor}}}\; {{\rm{due}}}\; {{\rm{to}}}\; {{\rm{mortality}}}\; {{\rm{in}}}\; {{\rm{age}}}\; {{\rm{group}}}\; {{\rm{a}}}}.$$\end{document}ξa=lossoflaborduetomorbidityinagegroupalossoflaborduetomortalityinagegroupa. Next, we assumed that the following holds in any given year for each age group a:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{\xi }}}}^{\left({{\rm{a}}}\right)}=\frac{{{{\rm{YLD}}}}^{\left({{\rm{a}}}\right)}}{{{{\rm{YLL}}}}^{\left({{\rm{a}}}\right)}},$$\end{document}ξa=YLDaYLLa,where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{YLD}}}}^{\left({{\rm{a}}}\right)}$$\end{document}YLDa represents the years lived with diabetes and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{{\rm{YLL}}}}^{\left({{\rm{a}}}\right)}$$\end{document}YLLa represents the years of life lost due to diabetes. In sum, by reducing the prevalence of diabetes, the counterfactual scenario is associated with an increase in labor supply as compared with the status quo scenario. We approximated the change in labor supply (at time t for age group a) by15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta {{{\rm{L}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}\approx {{{\rm{l}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}{{{\rm{N}}}}_{{{\rm{t}}}}^{\left({{\rm{a}}}\right)}{\sum }_{{{\rm{\tau }}}=0}^{\min \left\{{{\rm{t}}},{{\rm{a}}}\right\}-1}{{{\rm{\sigma }}}}_{{{\rm{r}}},{{\rm{t}}}-1-{{\rm{\tau }}}}^{\left({{\rm{a}}}-1-{{\rm{\tau }}}\right)}\left[1+{{{\rm{p}}}}^{{{\rm{\tau }}}}{{{\rm{\xi }}}}^{\left({{\rm{a}}}-1-{{\rm{\tau }}}\right)}\right].$$\end{document}ΔLta≈ltaNta∑τ=0mint,a−1σr,t−1−τa−1−τ1+pτξa−1−τ. Data sources This study used publicly accessible data (Supplementary 3) that did not meet the regulatory definition of human participant research and were exempt from Institutional Review Board review under the Common Rule. Data on diabetes incidence, prevalence, mortality, years of life lost due to premature mortality [YLL], and years lost due to disability [YLD] were from GBD 2021 by the Institute for Health Metrics and Evaluation3. Data on educational attainment, labor force participation, population, and physical capital were from the Barro-Lee Education Database34, the International Labor Organization35, the United Nations Department of Economic and Social Affairs (DESA)36, and the Penn World Table projections37, respectively. Diabetes treatment costs were from Dieleman et al38. GDP projections for the status quo scenario and savings rate data were from the World Bank World Development Indicators database and the World Economic Outlook database39–41. More details on data sources and parameter values are provided in the Supplementary 3. To make estimates comparable across countries, economic data and estimations were converted to 2017 INT$. We estimated the economic burden of diabetes for 190 countries and territories globally, with results presented separately by World Bank income groups (low-income, lower-middle-income, upper-middle-income, and high-income) and by World Bank regional classifications (East Asia & Pacific, Europe & Central Asia, Latin America & Caribbean, Middle East & North Africa, North America, South Asia, and Sub-Saharan Africa). Of the 190 countries and territories, 126 had all the data inputs we needed, and their economic burden of diabetes was calculated directly through the health-augmented macroeconomic model; the remaining 64 countries and territories lacked the necessary data, and we referred to previous studies that used linear projections to infer their economic burden of diabetes (Supplementary 4). In addition, we present two intermediate scenarios in this study to simulate the impact of a gradual reduction in the burden of falls under different intervention scenarios (10% and 20% reductions by 2050, respectively). We further conducted a supplementary feasible counterfactual analysis by reducing diabetes-related incidence and mortality in each country to the lowest observed global levels, holding all other model inputs constant. Sensitivity analysis Confidence intervals for the economic burden of diabetes were provided by varying mortality and morbidity rates according to the lower and upper bounds (95% uncertainty intervals) of the GBD mortality and morbidity data. In addition, we provided estimates with a discount rate of 3% in the main analysis and estimates with discount rates of 0%, 2%, and 5% in the Supplementary 5. Analyses were conducted using R 4.3.0 software. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file