Trends and Inequalities in Diabetes-Related Complications Among U.S. Adults, 2000–2020

Diabetes is a common and debilitating chronic condition; in 2021 approximately 38.1 million adults aged 18 years or older had diabetes—or 14.7% of all U.S. adults (both diagnosed and undiagnosed diabetes) (1). Diabetes is frequently associated with macrovascular (e.g., heart failure [HF], myocardial infarction [MI], and stroke) and microvascular (e.g., end-stage renal disease [ESRD], diabetic retinopathy) complications, lower-extremity amputation, and acute events such as diabetic ketoacidosis (DKA) and hyperosmolar hyperglycemic state (HHS) (2). As patients with diabetes live longer, the burden of these complications and consequent hospitalization rates are expected to increase (2). For example, between 2000 and 2018, the proportion of hospitalizations of adults with diabetes in the U.S. increased from 17.1% to 27.3% (3). Notably, these hospitalizations were more prevalent among certain demographic groups, suggesting differential distribution of diabetes-related complications and/or inequalities in health care access and utilization.

Previous analyses of national survey data indicated that risk of diabetes-related complications and their secular trends vary by age, race and ethnicity, and sex (1,4,5). As an update to previous studies, we examined national surveys, administrative data, and registry data for the period 2000–2020 to assess trends in diabetes-related complications among adults with diagnosed diabetes overall and by age, race and ethnicity, and sex. For each complication we also assessed trends in inequalities across each demographic subgroup.

For hospitalization trends we analyzed data from the 2000–2020 National (Nationwide) Inpatient Sample (NIS). NIS is a nationally representative sample of hospital inpatient stays across 48 states participating in the Healthcare Cost and Utilization Project (HCUP) of the Agency for Healthcare Research and Quality. The data with the systematic sampling design used for NIS approximates 20% of discharges, approximately seven million hospital stays from all hospitals, which can be weighted to represent >35 million hospitalizations annually in the U.S. (6). NIS provides ICD-9, Clinical Modification (ICD-9-CM), codes before 1 October 2015 and ICD-10-CM codes thereafter. Due to this transition, we used ICD-9-CM codes for 1 January 2000 through 30 September 2015 and ICD-10-CM codes starting from 1 January 2016. Based on HCUP recommendations, data for 2015 included only the first three quarters of the year and the denominator used to calculate the rates was weighted by three-fourths (7). NIS also collected data on patient demographics, hospital characteristics, payment sources, and patient disposition.

Hospitalizations with diabetes were defined according to a diabetes ICD-9-CM or ICD-10-CM code listed as primary or secondary diagnosis on the patient’s record (ICD-9-CM, 250; ICD-10-CM, E10, E11, or E13). Diabetes-related complications included HF, MI, stroke, ESRD, nontraumatic lower-extremity amputation (NLEA), and hyperglycemic crisis (see Supplementary Table 2 for ICD-9-CM and ICD-10-CM codes) and were selected because of high likelihood of requiring hospital admission. Any of the six conditions listed as the primary diagnosis among inpatients with diabetes were considered to be diabetes complications. Incident ESRD was defined according to the number of people aged ≥18 years initiating treatment for ESRD (i.e., dialysis or kidney transplantation) who had diabetes listed as the primary cause of ESRD. ESRD data came from the 2000–2020 United States Renal Data System, which is a clinical and claims-based registry from the Centers for Medicare & Medicaid Services (8). Annual data (2000–2020) from the National Health Interview Survey (NHIS) were used to estimate the number of U.S. adults (≥18 years of age) with diagnosed diabetes. NHIS is a nationally representative, cross-sectional household survey of the civilian noninstitutionalized U.S. population that has been conducted since 1957 by the Centers for Disease Control and Prevention National Center for Health Statistics. Diagnosed diabetes was defined according to a “yes” answer to the question, “[(If female), Other than during pregnancy] Have you ever been told by a doctor or health professional that you have diabetes or sugar diabetes?”

Institutional review board approval was not required for this study, as all data were deidentified.

We report the crude weighted number of hospitalizations with each complication at the start (2000) and end (2020) of the study period, overall and with stratification by age-group (18–44, 45–64, 65–74, and ≥75 years), race and ethnicity (Hispanic, non-Hispanic [NH] Black, NH White, and NH other), and sex (female and male). Due to the heterogeneity of the adults NH other category and the challenges with drawing meaningful results for the groups of that category, we only present data for the categories Hispanic, NH Black, and NH White. Annual rates of complications were calculated as the weighted number of hospitalizations with the specified complication divided by the weighted number of adults with diagnosed diabetes. Rates were age standardized in the above groupings based on the age distribution of the 2000 U.S. standard population. Analyses were conducted with SAS-callable SUDAAN (version 11.0.1; Research Triangle Institute) to account for the complex sampling designs in NIS and NHIS. Taylor series linearization was used to estimate the variance of the ratio of the numerator and denominator.

Measures of absolute inequality included the between-group variance (BGV), which summarizes squared deviations from the population average for nominal categories (e.g., race and ethnicity and sex), and slope index of inequality (SII) for ordinal categories (e.g., age), which measures the average health status between most and least advantaged groups while accounting for all other subgroups with a linear regression model (9). Results were exported to HD*Calc, a tool used to calculate various indices of inequalities, for calculation of the annual BGV and SII for each complication to measure inequalities (10). A value of 0 for BGV and SII represents no inequalities. BGV can only be nonnegative; higher values represent greater sex and racial and ethnic inequalities. The SII can be negative or positive. A negative SII for age represents a higher concentration of complications among younger age-groups, while a positive value represents greater concentration of complications among older age-groups. Use of these summary measures to monitor inequalities over time may provide an advantage over pairwise comparisons in accounting for changes in population size of each subgroup over time (9) and because a single value is provided, allowing for ease of communication to stakeholders. The advantage of BGV for assessing sex and racial and ethnic inequalities is that the population mean is used for the reference group rather than a specific subgroup, thereby avoiding the stigmatization or idealization of specific subgroups (10).

For each complication, we examined trends in rates and inequalities over time using log-linear models from the Joinpoint trend analysis software program. We set a maximum of three joinpoints (i.e., four trend segments) based on the total number of data points, and the best model fit was determined with permutation tests to identify the trend segments (11). In Joinpoint, we estimated the annual percent change (APC) for each trend segment and average annual percent change (AAPC) for the overall trend and calculated 95% CIs. Race and ethnicity information in NIS was incomplete prior to 2012 (>20% missing); thus, we only calculated trends by race and ethnicity from 2012 to 2020. Log-linear models are used for joinpoint regression, and we were unable to estimate trends in age inequalities for NLEA due to negative and positive SII values for some years (11). Statistical significance for trends was assessed at a two-sided P value <0.05.

The coronavirus disease 2019 pandemic may have influenced the 2020 complication rates given the disruption in health care services, rising infections, and adverse mental health effects (12). Thus, we conducted sensitivity analyses with exclusion of 2020 data to assess whether that influenced the trends in complications.

Patient characteristics by diabetes complication in 2000 and 2020 are shown in Supplementary Table 1. There were changes in the distributions of age, sex, primary payer type, disposition, and U.S. region between 2000 and 2020, as well as race and ethnicity between 2012 and 2020, for all diabetes-related complications (all P values <0.05). Specifically, the proportion of uninsured patients for HF, MI, stroke, and NLEA hospitalizations increased from 2000 to 2020, while the proportion of patients with Medicaid increased for those same complications and hyperglycemic crisis. The proportion of privately insured patients for hospitalizations decreased across all diabetes-related complications, and the proportion of patients with Medicare for hospitalizations decreased for all complications except for hyperglycemic crisis.

The overall trends for each complication are shown in Fig. 1, and the trends by age, race and ethnicity, and sex are displayed in Fig. 2 and Supplementary Figs. 1–6. The rates and APC for the trends of each complication by age, race and ethnicity, and sex are shown in Table 1 for HF, MI, and stroke and Table 2 for ESRD, NLEA, and hyperglycemic crisis. Trends in age, race and ethnicity, and sex inequalities for each complication are displayed in Table 3. Sensitivity analyses results with exclusion of 2020 data are shown in Supplementary Tables 3 and 4. The direction and magnitude of trends were generally similar with some shifts in the joinpoint years. Significant differences in magnitude and direction have been noted below.

Overall, hospitalization rates for HF declined during 2000–2012 (APC −4.7%, P < 0.001) and increased thereafter (APC 3.9%, P < 0.001) (Table 1 and Fig. 1). Hospitalizations were lowest for the 18–44 years age-group; however, rates increased steadily during 2013–2020 (APC 6.5%, P = 0.007) (Table 1, Fig. 2, and Supplementary Fig. 1A). A similar increase was observed for the 45–64 years age-group during 2012–2020 (APC 6.6%, P < 0.001). The 18–44 and 45–64 years age-groups contributed a larger proportion of the hospitalizations in 2020 (combined 33.4%) than in 2000 (28.6%) (Supplementary Table 1). The declines in hospitalizations for those ages ≥75 years during 2017–2018, coupled with an increase for those ages <75 years during that same time period, resulted in overall declines in age inequalities after 2017, as the SII moved closer to 0 (APC −15.0%, P = 0.009) (Table 3). However, in sensitivity analyses, hospitalizations increased from 2013 to 2019 for those ages ≥75 years (Supplementary Table 3), contributing to increases in age inequalities during that period (Supplementary Table 4).

Hospitalizations for HF increased in all racial and ethnic groups, with NH Black adults having roughly double the rates of NH White and Hispanic adults (Table 1, Fig. 2, and Supplementary Fig. 2B). Racial and ethnic inequalities increased during 2012–2020 (APC 11.8%, P = 0.012) (Table 3). Hospitalizations among males surpassed those among females beginning in 2005 (Table 1, Fig. 2, and Supplementary Fig. 2C), and there were steady increases in HF hospitalizations for both sexes during 2012–2020 (males, APC 3.7%, P < 0.001, and females, APC 4.1%, P < 0.001).

Hospitalizations for MI significantly declined during 2000–2008 (APC −6.0%, P < 0.001) and remained level from 2009 to 2020 (APC 0.7%, P = 0.13) (Table 1 and Fig. 1). In sensitivity analyses, there was a slight, but significant, increase from 2009 to 2019 (APC 1.5%, P < 0.001) (Supplementary Table 3). Middle-aged individuals (45–64 years) with diabetes were the only age-group with a significant increase in MI hospitalization rates (2009–2020 APC 2.5%, P < 0.001) (Table 1, Fig. 2, and Supplementary Fig. 2A) and contributed a higher proportion of the hospitalizations in 2020 (37.2%) than in 2000 (32.3%). Overall, age inequalities declined during 2000–2020 as the SII moved toward 0 (AAPC −6.0%, P < 0.001) (Table 3). MI hospitalization rates were highest for NH White adults followed by NH Black adults, but the trend was flat for all racial and ethnic groups (Table 1, Fig. 2, and Supplementary Fig. 2B). In sensitivity analyses, there were significant increases for NH Black adults from 2012 to 2019 and for Hispanic adults from 2014 to 2019 (Supplementary Table 3). MI hospitalization rates were higher among males than females, and these rates increased during 2009–2020 (APC 1.2%, P = 0.037) (Table 1, Fig. 2, and Supplementary Fig. 2C). No significant trends were observed in racial and ethnic or sex inequalities (Table 3).

Overall, hospitalizations for stroke declined through 2009 (APC −3.9%, P < 0.001) and then increased close to 2000 levels (APC 2.8%, P < 0.001) (Table 1 and Fig. 1). As with HF and MI, an age gradient was also evident for stroke, with stroke rates for the oldest age-group (≥75 years) almost two times the rate of the next-oldest group (65–74 years) (Table 1, Fig. 2, and Supplementary Fig. 3A). Although with lower absolute stroke hospitalization rates, those aged <65 years were the only ones with an increasing trend (Table 1, Fig. 2, and Supplementary Fig. 3A), resulting in overall decreasing age inequalities as the SII became closer to 0 (AAPC −3.6%, P = 0.006) (Table 3). In sensitivity analyses, those 65–74 years old had significant increases from 2014 to 2019 and no significant changes in age inequalities during that period (Supplementary Tables 3 and 4). Hospitalization rates were highest for NH Black adults, approximately double the rates of Hispanic adults, and remained highest as rates increased similarly across racial and ethnic groups during 2012–2020 (Table 1, Fig. 2, and Supplementary Fig. 3B). Racial and ethnic inequalities increased during 2012–2020 (APC 12.0%, P = 0.014) (Table 3). Stroke hospitalization rates were similar for male and female adults, with an increasing trend in the last decade (Table 1, Fig. 2, and Supplementary Fig. 3C) and an overall increase in sex inequalities during 2000–2020 (AAPC 6.0%, P = 0.018) (Table 3) for both groups.

Overall, ESRD incidence rates declined during 2000–2010 (APC −3.2%, P < 0.001) and then slightly increased (APC 1.0%, P = 0.044) (Table 2 and Fig. 1). There was a steady increase in ESRD incidence among those aged 45–64 years in the second decade (2010–2020, APC 1.4%, P < 0.001), representing the single largest case contributor, with >40% of the ESRD cases (Table 2, Fig. 2, Supplementary Fig. 4A, and Supplementary Table 1). ESRD incidence declined after 2016 (APC −5.3%, P = 0.009) for those ≥75 years of age. With these trends there was a significant overall decline in age inequalities during 2000–2020 as the SII moved toward 0 (AAPC −4.6%, P < 0.001). ESRD incidence was substantially higher for NH Black adults, followed by Hispanic adults (Table 2, Fig. 2, and Supplementary Fig. 4B); however, rates only increased among NH White adults (2012–2020 APC 2.3%, P = 0.028), so there was a reduction in racial and ethnic inequalities after 2012 (AAPC −4.2%, P = 0.042) (Table 3). ESRD rates were higher for males than for females. There were similar initial declines for males and females followed by stabilization for females (2010–2020 APC 0.7%, P = 0.204) and a slow increase for males (2009–2020 APC 1.1%, P = 0.030), resulting in a significant increase in sex inequalities during 2000–2020 (AAPC 5.3%, P = 0.005) (Table 2, Supplementary Fig. 4C, and Table 3).

Overall, NLEA hospitalizations significantly declined during 2000–2009 (APC −5.2%, P < 0.001) and then increased to rates surpassing those in 2000 (2010–2020 APC 5.9%, P < 0.001) (Table 2 and Fig. 1). NLEA hospitalization increased in the past decade for all age-groups except for the ≥75 years age-group, increasing at the highest rate, at 7.5%, in the 45–64 years age-group (P < 0.001) (Table 2), cases among whom comprised 50.1% of the cases in 2020 (Supplementary Table 1). The highest rates of NLEA hospitalization during 2012–2019 were among NH Black adults, but the rates were the same for NH Black and NH White adults in 2020 (Table 2, Fig. 2, and Supplementary Fig. 5B). NLEA hospitalizations increased across all three racial and ethnic groups, and there were no significant trends in racial and ethnic inequalities (Table 3). NLEA hospitalization rates were higher among males during 2000–2020, with increasing rates of NLEA hospitalizations among both sexes during 2009–2020. The rate of increase was higher for males, leading to a significant increase in sex inequalities during 2009–2020 (APC 15.3%, P < 0.001) (Table 2, Supplementary Fig. 5C, and Table 3).

Overall, hyperglycemic crisis hospitalization rates increased steadily during 2000–2020 (APC 2.2%, P < 0.001), and the cases had more than doubled from 100,101 in 2000 to 231,995 in 2020 (Table 2, Fig. 1, and Supplementary Table 1). The largest number of events (>50%) was among those aged 18–44 years, among whom the hospitalization rate was at least five times higher than that of other age-groups (Table 2, Supplementary Fig. 6A, and Supplementary Table 1). There was an increased trend in the last decade among all age-groups, but the trend for the 18–44 years age-group increased steadily since 2000 (APC 2.2%, P < 0.001), resulting in increasing age inequalities during 2000–2020 as the SII values deviated further from 0, becoming more negative (APC −1.4%, P < 0.001). Hyperglycemic crisis rates were highest among NH Black adults followed by NH White adults (Table 2, Fig. 2, and Supplementary Fig. 6B). There was only a significantly increasing rate during 2012–2020 for NH Black adults (AAPC 3.6%, P = 0.047), resulting in significant increasing inequalities during 2012–2020 (AAPC 11.7%, P = 0.045) (Table 3). Rates were similar for males and females, but the gap widened in the last decade, with a higher increasing rate among males during 2009–2020 (APC 4.5%, P < 0.001) (Table 2, Fig. 2, and Supplementary Fig. 6C).

Although there were declines in the early 2000s, during the last decade rates of diabetes-related HF, stroke, NLEA, and ESRD have increased, while MI rates remained flat. Hyperglycemic crisis rates steadily increased over most of the past two decades. In addition, age, sex, and racial and ethnic inequalities have worsened for some complications, with the most alarming findings of increasing trends for age-groups 18–44 and 45–64 years, male adults, and NH Black adults. The decreasing age inequalities for hospitalizations for HF, MI, and stroke likely are attributed to the increasing rates among younger age-groups, while the declines in racial and ethnic inequalities in ESRD incidence are due to increases among NH White adults rather than improvements for disproportionately affected groups.

As concluded previously, the increase in complications beginning in 2008 may be attributed to the rising complication rates among young (18–44 years) and middle-aged (45–64 years) adults (4,13). Several factors may explain the rise among these age-groups. The rising cost of diabetes care, such as for insulin and other diabetes medications, particularly affects younger age-groups (<65 years), who are more likely to be under- or uninsured (14). We found increases in the proportion of hospitalizations for patients with no insurance or Medicaid for all complications, while the proportion of those privately insured decreased. This shift may reflect barriers to receiving care, as those who are uninsured may be unable to afford out-of-pocket costs and those with Medicaid may have fewer care options than those privately insured (15). An increase in high-deductible health plan (HDHP) enrollment among those ages <65 years may have also contributed to a reduction in diabetes monitoring, routine care, and medication adherence and an increase in acute health care use (16). Indeed, Fang et al. (17) found that achievement of glycemic and lipid targets has declined in recent years. Rates of obesity, a risk factor for many of these complications, are high for all age-groups (18). However, rates of severe obesity are higher among those ages <60 years and the rates have been increasing over time—a trend not seen in the ≥60 years age-group (18). Furthermore, the Great Recession of 2007–2009 may have partially contributed to the rise in complications in general, as it negatively affected mental health and health behaviors, which could have adversely impacted diabetes management (19).

Differences in access to health care and quality of care may explain the growing racial and ethnic inequalities, especially the persistently higher rates of HF, stroke, ESRD, and hyperglycemic crisis among NH Black adults. Racial and ethnic minority groups are more likely to be under- or uninsured (14), and those enrolled in HDHPs are less likely to have health savings accounts and thus more vulnerable to suboptimal care (20). Moreover, NH Black adults may be less likely to receive the recommended diabetes quality of care such as HbA_1c and lipid testing (21). The lack of access and suboptimal health care could have contributed to the lower prevalence of controlled HbA_1c, blood pressure, and lipid levels among NH Black adults with diabetes and subsequently higher rates in the observed complications (22).

Hypertension, a major risk factor for HF, MI, stroke, and ESRD, may be a contributor to the increases in these complications, as findings of a previous study showed overall declines in the prevalence of controlled blood pressure from 2011 to 2018 among those with diabetes (17). Additionally, steadily increasing ESRD incidence, observed in this study, may have also partially contributed to the rising HF, MI, and stroke rates given that chronic kidney disease is a risk factor for cardiovascular disease (23). Harding et al. (13) found increases in HF rates in 2013–2017. Our data show sustained increases through 2020.

Substance use disorders (SUDs) may play a role in the observed complication trends, as they are on the rise, with overdose deaths increasing fivefold during the last 20 years (24). Younger populations are at higher risk for substance use, especially males (25), and evidence shows that SUD rates are higher among adults with type 2 diabetes (26). SUD among individuals with diabetes leads to poorer glucose control (27) and an increased risk of diabetes-related complications (28). Although we are unable to assess causal associations for these data, the increasing complications in the younger age-groups may be partially explained by the increase in SUD (29).

Although advancements have been made in wound care and revascularization procedures (30), we observed an increase in rates of NLEA during the last decade. Population-based studies on diabetic foot ulcers (DFU) are lacking, but worldwide, DFU, often the precursor to NLEA, are on the rise (31). Young and middle-aged adults appear to be predominately affected. They present with more severe ulcer stage and are more likely to have worse glucose control in comparison with older patients with diabetes and DFU (31,32). Advanced-stage DFU are also more common among patients with lower income and no insurance or who are underinsured (33). Access to health care is critical to DFU prevention and treatment. Foot examinations by health care providers for adults with diabetes have decreased during the last decade, and it is possible that changes in health care access have contributed to decreases in diabetes management and preventive services and subsequent increase in NLEA (34). Rates of peripheral neuropathy, peripheral arterial disease, and cardiovascular disease are higher among men, which, along with poorer self-care practices and less access to care, is likely explain the sex differences (31).

Findings of an earlier study (35) showed an increase in DKA and HHS from 2009 through 2015, and we see that trend continuing. Access to insulin may play an important role in the observed high burden and increase in hyperglycemic crisis hospitalizations among younger adults (36). The combination of rising costs in insulin in the past two decades (37), increases in enrollment of HDHPs (35), and higher rates of uninsurance among younger age-groups (14) may have resulted in medication rationing (36,38). Additionally, cannabis use has increased in the younger age-groups (29) and has been shown to be associated with increased risk for DKA among adults with type 1 diabetes (39). Furthermore, since MI and stroke can be precipitating factors of HHS (40), increases in MI hospitalizations for middle-aged adults and increases in stroke hospitalizations for young and middle-aged adults, observed in this study, may have contributed to the rising hyperglycemic crisis rates for these age-groups.

This study had several limitations. Since NIS provides event-level and not patient-level data, many of the hospitalizations could have been representative of readmissions and not unique patients, resulting in overestimation of hospitalization rates. However, readmissions from 2010 to 2018 for patients with diabetes have been declining (3). Although NIS is a nationally representative data set that represents the majority of U.S. hospitals (including all states except ID and AL), NIS does not include hospitalizations for federal hospitals (e.g., the Department of Veteran Affairs and Indian Health Service). Additionally, we were only able to assess inequalities across Hispanic, NH Black, and NH White adults due to limited sample size. Thus, other racial and ethnic groups in the U.S. are not represented here, and further work should be done to examine inequalities in those groups with analytic methods appropriate to sample size. Furthermore, there was an inability to differentiate between diabetes types in NHIS, which in this study precluded examination of hospitalization rates by diabetes type. Stratification by diabetes duration was also not possible. The shift in using ICD-9-CM and ICD-10-CM codes in October 2015 may have influenced the findings, but most of the observed increasing trends began before this time point, suggesting that changes in coding had minimal impact on the overall patterns. Finally, we use hospitalizations as a proxy for complications, and they may not be equivalent.

In summary, in this study we used nationally representative data to examine diabetes-related complications across two decades and included novel measures of inequalities. Despite the declines in diabetes-related complications in the early 2000s and resurgence in the 2010s, we observed a continued increase of several complications. Although these data do not provide definitive etiologies, they do highlight potential populations of focus for prioritizing and tailoring diabetes secondary prevention efforts. Reducing inequalities and reversing these growing trends in diabetes-related complications are important, as is continued surveillance to monitor progress.

Acknowledgments. The authors give special thanks to Jacob T. Wittman (Division of Diabetes Translation, Centers for Disease Control and Prevention) for support on preparing the figures.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. R.S., K.M.B., I.A.H., M.E.P., F.J.P., C.S.H., and S.R.B. were involved in the conception, design, and conduct of the study and the interpretation of the results. R.S. conducted statistical analysis and wrote the first draft of the manuscript, and all authors edited, reviewed, and approved the final version of the manuscript. R.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 84th Scientific Sessions of the American Diabetes Association, Orlando, FL, 21–24 June 2024. A video presentation can be found in the online version of the article at https://doi.org/10.2337/dci24-0022.

Handling Editors. The journal editor responsible for overseeing the review of the manuscript was Steven E. Kahn.