Evaluation of the Cascade of Diabetes Care in the United States, 2005-2016

Key Points

Question  Were there any changes in population-level achievement of diabetes treatment targets in the United States from 2005 to 2016?

Findings  In this population-based study of 2488 individuals, approximately 1 in 4 adults with diagnosed diabetes achieved the composite goal in the United States. No significant improvement in any of the individual targets was observed between 2005 and 2016, and gaps in achieving diabetes care targets, particularly among young (18-44 years), female, and nonwhite adults, persisted during the study period.

Meaning  It appears that advances in diabetes care over the past decade have not translated into meaningful improvement in population-level treatment outcomes.

Importance  Treatment advances in diabetes can meaningfully improve outcomes only if they effectively reach the populations at risk.

Objectives  To evaluate whether the cascade of US diabetes care, defined as diabetes diagnosis, linkage to care, and achievement of individual and combined treatment targets, improved from 2005 to 2016 and to investigate potential disparities in US diabetes care.

Design, Setting, and Participants  Nationally representative, serial cross-sectional studies included in the 2005-2016 National Health and Nutrition Examination Survey were evaluated. Data on nonpregnant US adults (age ≥18 years) with diabetes who had reported fasting for 9 or more hours (n = 1742 diagnosed and 746 undiagnosed) were included. Data analysis was performed from August 1, 2018, to May 10, 2019.

Exposures  Time period (2005-2008, 2009-2012, and 2013-2016), age, sex, race/ethnicity, health insurance, and educational level incorporated into logistic regression models predicting odds of target achievement.

Main Outcomes and Measures  Proportion of participants overall and stratified by age, sex, and race/ethnicity who were linked to diabetes care and met glycemic (hemoglobin A_1c <7.0%-8.5%, depending on age and complications), blood pressure (<140/90 mm Hg), cholesterol level (low-density lipoprotein cholesterol <100 mg/dL), and smoking abstinence targets and a composite of all targets.

Results  In 2013-2016, of 1742 US adults with diagnosed diabetes, 94% (95% CI, 92%-96%) were linked to diabetes care; 64% (95% CI, 58%-69%) met hemoglobin A_1c level, 70% (95% CI, 64%-75%) met blood pressure level, and 57% (95% CI, 51%-62%) met cholesterol level targets; 85% were nonsmokers (95% CI, 82%-88%); and 23% (95% CI, 17%-29%) achieved the composite goal. Results were similar in 2005-2008 (composite 23%) and in 2009-2012 (composite 25%). There was no significant improvement in diagnosis or target achievement during the study period. Compared with middle-aged adults (45-64 years) with diagnosed diabetes, older patients (≥65 years) had higher odds (adjusted odds ratio [aOR], 1.70; 95% CI, 1.17-2.48) and younger adults (18-44 years) had lower odds (aOR, 0.53; 95% CI, 0.29-0.97) of meeting the composite target. Women had lower odds of achieving the composite target than men (aOR, 0.60; 95% CI, 0.45-0.80). Non-Hispanic black individuals vs non-Hispanic white individuals had lower odds of achieving the composite target (aOR, 0.57; 95% CI, 0.39-0.83). Having health insurance was the strongest predictor of linkage to diabetes care (aOR, 3.96; 95% CI, 2.34-6.69).

Conclusions and Relevance  It appears that the diabetes care cascade in the United States has not significantly improved between 2005 and 2016. This study’s findings suggest that gaps in diabetes care that were present in 2005, particularly among younger adults (18-44 years), women, and nonwhite individuals, persist.

Diabetes is a major US public health burden, affecting 30.2 million adults, with an expected prevalence of 54 million by 2030.^1,2 Diabetes accounted for $237 billion in direct medical costs in 2017.³

There has been general consensus regarding several major goals of diabetes treatment for over 2 decades: achieving hemoglobin A_1c (HbA_1c) levels, blood pressure (BP), and low-density lipoprotein cholesterol (LDL-C) level treatment targets and promoting smoking cessation. Although numeric treatment targets have shifted slightly over the past decade, the overall approach has not.^4,5 Previous studies reported that population achievement of diabetes treatment goals (the ABCs of diabetes: HbA_1c level [A], BP [B], and LDL-C level [C]) had improved from the late 1990s to 2010 in the United States.⁶ To date, it is not known whether there has been continued progress.

Examining each stage along the diabetes care cascade may shed light on the population-level steps required to improve diabetes care. We define the care cascade as diagnosis, linkage to care, achievement of individual treatment targets, and a composite of all individual targets.^7-9

In a prior study, Ali et al⁹ presented a cascade of diabetes care in the United States in 2007-2012 and reported that diabetes care was suboptimal in that period. We build on their work by (1) using newer data to illustrate the cascade of diabetes care in the United States in 2013-2016, (2) providing a more detailed examination of whether US diabetes care improved from 2005 to 2016, and (3) evaluating potential demographic disparities in the cascade of US diabetes care over this period.

We analyzed the National Health and Nutrition Examination Survey (NHANES) cross-sectional data from 2005 to 2016.¹⁰ To increase statistical reliability of our estimates, we combined NHANES’ 2-year cycles to construct three 4-year intervals: 2005-2008, 2009-2012, and 2013-2016. Data analysis was performed from August 1, 2018, to May 10, 2019. We created a cascade of diabetes care in each period, then investigated whether care variables have improved over time. We further assessed the potential age, sex, and racial/ethnic disparities in US diabetes care. NHANES protocols were approved by the National Center for Health Statistics research ethics review board. With use of publicly available data, this study was considered exempt from review and thus no patient written informed consent was needed.

NHANES used stratified, multistage, probability cluster sampling techniques to obtain samples that are representative of the US, noninstitutionalized civilian population.¹⁰ Data were collected through household interviews and clinical examinations at a mobile examination center (morning and afternoon appointments). The average response rate for the NHANES cycles between 2005 and 2016 was 74% (range, 61%-80%). Study design, survey instruments, and laboratory methodologies are described elsewhere.¹⁰

During the study period (2005-2016), 58 660 participants completed the interview and clinical examination. We included only participants from the morning sessions who were aged 18 years or older, not pregnant, had reported fasting for 9 hours or more, and had results of HbA_1c and fasting plasma glucose (FPG) levels available. Restricting the cohort to participants examined in the morning session was required for assessment of FPG and LDL-C levels. The base case cohort was restricted to adults with diagnosed diabetes; outcomes including individuals with undiagnosed diabetes are reported in eTable 4 and eFigure 6 in the Supplement.

We defined the following outcomes to create a cascade of care for each period. Patients were considered to have diabetes if they responded yes to the question, “other than during pregnancy, have you ever been told by a doctor/health professional that you have diabetes/sugar diabetes?” or if they had either an HbA_1c level of 6.5% or higher (to convert to proportion of total hemoglobin, multiply by 0.01) or FPG level of 126 mg/dL or higher (to convert to millimoles per liter, multiply by 0.0555).⁵ No distinction was made between type 1 and type 2 diabetes.

Diagnosed diabetes was defined by a yes response to the question, “other than during pregnancy, have you ever been told by a doctor/health professional that you have diabetes/sugar diabetes?” Patients were labeled as having undiagnosed diabetes if they responded no to this question but had either an HbA_1c level of 6.5% or higher or FPG level of 126 mg/dL or higher.

Patients with diagnosed diabetes were considered to be linked to diabetes care if they reported that they (1) saw a physician or other health care professional for diabetes in the past year, (2) were taking “diabetic pills” to lower blood glucose levels, (3) were taking insulin, or (4) saw a diabetes specialist in the past 2 years. Those who did not meet any of these criteria were considered to be not linked to diabetes care.

Glycemic control was considered an HbA_1c level less than or equal to the target level, applied to all 3 time periods in the analysis. For each patient, we defined a liberal personalized HbA_1c level target between 7.0% and 8.5%, depending on the patient’s age and presence of complications (cardiovascular disease [CVD] history, nephropathy, and retinopathy) per the 2018 American Diabetes Association guidelines.⁵ For patients younger than 65 years, the target HbA_1c level was 7.0% in the absence of reported complications or 8.0% if diabetes complications were reported. For patients 65 years or older, the target HbA_1c level was 7.5% in patients without complications or 8.5% in those with complications.

We used the mean of all BP measurements (3-4 readings) to estimate each patient’s systolic and diastolic BP. Blood pressure control was defined as a systolic/diastolic BP level less than 140/90 mm Hg per the 2018 American Diabetes Association guidelines.⁵ We also estimated the proportion of patients who met lower BP targets (eFigure 2 in the Supplement).¹¹

Cholesterol control was defined as having an LDL-C level less than 100 mg/dL (to convert to millimoles per liter, multiply by 0.0259). We also estimated the proportion of patients who (1) reported taking statins (ie, general use of statins among US adults with diagnosed diabetes); (2) were eligible for CVD primary prevention (ie, age ≥40 years and no CVD history), reported taking statins, and had an LDL-C level less than 100 mg/dL; and (3) were eligible for CVD secondary prevention (ie, had a CVD history), reported taking statins, and had an LDL-C level less than 70 mg/dL (eFigure 1 in the Supplement).^5,12

Participants were defined as current nonsmokers if they had serum cotinine levels of 10 ng/mL or lower (to convert to nanomoles per liter, multiply by 5.675) or responded no to either, “have you smoked at least 100 cigarettes in life?” or “do you now smoke cigarettes?” Participants were defined as current smokers if they had serum cotinine levels greater than 10 ng/mL or provided a positive response to the question, “do you now smoke cigarettes?”¹³

We further defined combination outcomes using the outcomes combined BC control (BP and LDL-C control); combined ABC control (BC and glycemic control); and combined ABCN control or composite control (ABC and not smoking).

We conducted the statistical analysis using the fasting sample weights provided in the NHANES data to adjust for complex survey designs, missing data, and nonresponse. We tested for differences in cohort characteristics across the 3 periods (2005-2008, 2009-2012, and 2013-2016) using the adjusted Wald F test.

To develop a cascade of diabetes care for each period, we calculated the proportions and 95% CIs of patients with diagnosed diabetes who were linked to diabetes care; met the target HbA_1c, BP, and LDL-C level thresholds; and were nonsmokers. To do so, we used the svyciprop function with likelihood method in the R survey package, version 3.33-2 (R Foundation), which uses a scaled χ² distribution for the log likelihood from a binomial distribution and is known to produce accurate 95% CIs for proportions, especially near 0 and 1.^14-16

To assess whether linkage and achievement of treatment goals improved over time and investigate disparities in diabetes care, we developed a logistic regression model and adjusted for period, demographics (age, sex, and race/ethnicity), as well as educational level and health insurance coverage as a proxy for socioeconomic status. We adjusted for socioeconomic status because it can influence access to care and medication and, thus, achievement of care goals. We also developed a more complex model that further adjusts for diabetes severity and complications since disease severity may influence achievement of diabetes care goals (eTable 1 in the Supplement). Moreover, we calculated the proportion of patients in distinct clinically relevant ranges of HbA_1c level, BP, LDL-C level, and body mass index in each period (eFigure 2 in the Supplement).

To determine which variables had a significant effect on linkage and treatment target achievement, we calculated a 2-tailed, unpaired P value for each categorical variable, using the Wald test for the hypothesis that all coefficients associated with the term were 0. We used the adjusted odds ratio (aOR) and 95% CI to assess the association between the individual levels of each categorical variable and achieving the diabetes care goals in each model. A significance level of .05 was used in all statistical analyses.

Additional analyses assessing the cascade of care for different populations are reported. These include (1) all adults with diagnosed and undiagnosed diabetes (eTable 2 and eFigure 3 in the Supplement), (2) only adults with diagnosed diabetes and linked to diabetes care (eTable 3 and eFigure 5 in the Supplement), and (3) only adults with undiagnosed diabetes (eTable 4 and eFigure 6 in the Supplement).

Over the 3 study periods, a total of 1742 NHANES participants had diagnosed diabetes; an additional 746 participants had undiagnosed diabetes (eTable 4 and eFigure 6 in the Supplement). Missing data in the sample ranged from 0% (demographic variables) to 7.2% (LDL-C level in 2009-2012).

There was no statistically significant difference across the 3 periods regarding most demographic and clinical variables, income, time since diabetes diagnosis, CVD history, or glucose-lowering medication use in the base case cohort. Educational level, insurance coverage, and urinary albumin-creatinine ratio improved in the last period compared with the first period (Table 1). For example, the mean (SE) proportion of participants with an educational level of college graduate or above was 12.8% (2.2%) in 2005-2008, 19.7% (2.2%) in 2009-2012, and 21.1% (2.8%) in 2013-2016.

In 2013-2016, 94% (95% CI, 92%-96%) of US adults with diagnosed diabetes were linked to diabetes care. In the same period, 64% (95% CI, 58%-69%) of all adults with diagnosed diabetes met the individualized HbA_1c target levels, 70% (95% CI, 64%-75%) achieved BP control, 57% (95% CI, 51%-62%) met the LDL-C target level, and 85% (95% CI, 82%-88%) were nonsmokers. However, only 41% (95% CI, 35%-47%) achieved BC control, 25% (95% CI, 20%-31%) met ABC targets, and 23% (95% CI, 17%-29%) met the combined ABCN target (Figure 1). Results were similar in 2005-2008 (composite control 23%) and 2009-2012 (composite control 25%).

The LDL-C level was consistently the least likely target to be achieved across all periods, a finding that was associated with low statin use (eFigure 1 in the Supplement). Time period had no statistically significant association with linkage to diabetes care, meeting individual treatment targets, and the composite (Table 2). The distributions of HbA_1c level, BP, LDL-C level, and BMI, an important diabetes care variable, were stable across the 3 periods (eFigure 2 in the Supplement). The mean (SE) BMI (calculated as weight in kilograms divided by height in meters squared) did not significantly change across the 3 periods (2005-2008: 32.6 [0.4], 2009-2012: 33.0 [0.4], 2013-2016: 32.5 [0.4]; P = .72). The Hosmer-Lemeshow goodness-of-fit test for the null hypothesis that the model fits the data well resulted in P = .08 (linkage to diabetes care), P = .16 (glycemic control), P = .01 (BP control), P = .66 (cholesterol level control), P = .33 (nonsmoking), P = .43 (combined ABC control), and P = .24 (combined ABCN control).

In 2013-2016, of all US adults with diabetes, 74% (95% CI, 70%-78%) were aware of their diagnosis, 70% (95% CI, 65%-74%) were linked to diabetes care, and 20% (95% CI, 16%-25%) met the composite target (eFigure 3 in the Supplement).

Compared with middle-aged (45-64 years) adults with diabetes, older patients (≥65 years) had lower odds of achieving BP control (aOR, 0.37; 95% CI, 0.25-0.55) but higher odds of meeting glycemic (aOR, 3.75; 95% CI, 2.77-5.06), LDL-C level (aOR, 1.70; 95% CI, 1.25-2.30), nonsmoking (aOR, 3.50; 95% CI, 2.38-5.15), and combined ABCN (aOR, 1.70; 95% CI, 1.17-2.48) targets. Younger adults (18-44 years) with diabetes had lower odds of being linked to diabetes care (aOR, 0.34; 95% CI, 0.18-0.67) or achieving ABCN control (aOR, 0.53; 95% CI, 0.29-0.97) compared with middle-aged patients (Table 2). For example, in 2013-2016, 30% of older adults with diabetes achieved the combined ABCN targets, whereas only 20% of middle-aged adults and 12% of young adults with diabetes achieved the combined ABCN goals (Figure 2).

Women had lower odds of achieving the combined ABCN control (aOR, 0.60; 95% CI, 0.45-0.80) than men, mainly owing to lower odds of LDL-C level control (aOR, 0.66; 95% CI, 0.53-0.82). Female sex was also associated with lower odds of linkage to care (aOR, 0.53; 95% CI, 0.30-0.94) (Table 2; eFigure 4 in the Supplement).

Reporting non-Hispanic black race was associated with significantly lower odds of achieving BP (aOR, 0.42; 95% CI, 0.30-0.60), LDL-C level (aOR, 0.62; 95% CI, 0.45-0.85), and combined ABCN (aOR, 0.57; 95% CI, 0.39-0.83) targets compared with reporting non-Hispanic white race (Table 2). Reporting Hispanic ethnicity vs non-Hispanic white race was associated with lower odds of linkage to care (aOR, 0.43; 95% CI, 0.24-0.77), LDL-C level (aOR, 0.60; 95% CI, 0.43-0.84), and combined ABC control (aOR, 0.60; 95% CI, 0.42-0.85) targets (Table 2). For example, in 2013-2016, 25% of US adults with diabetes who reported their race as non-Hispanic white achieved the composite, whereas only 14% of those reporting non-Hispanic black race and 18% of those reporting Hispanic ethnicity achieved the composite (Figure 3).

Adults with diabetes in the United States who had health insurance coverage had higher odds of being linked to diabetes care (aOR, 3.96; 95% CI, 2.34-6.69) and meeting glycemic targets (aOR, 1.61; 95% CI, 1.10-2.36) (Table 2). Most disparities in care persisted in the more complex multivariable logistic regression model (eTable 1 in the Supplement). When analyses were restricted to patients who were linked to diabetes care, age (P = .001) and sex (P = .001) disparities in achieving combined ABCN control persisted, but racial/ethnic disparities did not reach statistical significance (P = .07) (eTable 3 in the Supplement).

The cascade of diabetes care revealed persistent gaps and disparities between 2005 and 2016. Similar to results of a prior study,⁹ our investigation noted that, in the most recent period (2013-2016), less than 1 in 4 US adults with diagnosed diabetes met the composite glycemic, blood pressure, cholesterol, and nonsmoking target (ABCN). In addition, we found that none of the US diabetes care variables improved from 2005 to 2016.

Although, compared with hypertension and hyperglycemia, hyperlipidemia may be relatively easier to treat, the LDL-C level target was the least likely of the 3 to be met. Despite a modest increase in statin use over the 3 study periods, only 6 in 10 US adults with diagnosed diabetes were taking a statin in 2013-2016. Lack of access to care may be an important modifiable factor. Cholesterol level control and achievement of the composite goals were better among adults who were linked to diabetes care compared with those who were not linked or diagnosed (eFigure 5 and eFigure 6 in the Supplement).

In general, having health insurance was the strongest indicator of diagnosis, linkage, and achievement of the composite for all patients combined (eTable 2 in the Supplement), although it did not reach statistical significance for meeting the composite when only adults with diagnosed diabetes were included. Health insurance coverage increased in the United States after 2014 because of the Affordable Care Act, which could have influenced the cascade of care in the third period (2013-2016). Prior observational studies have shown improved diabetes diagnosis and earlier treatment under the Affordable Care Act,^17-19 but we did not observe national changes over time in the proportion of target achievement despite the introduction of the Affordable Care Act.

Certain populations, such as younger age (18-44 years), female sex, and nonwhite adults with diabetes, demonstrated persistent disparities in meeting diabetes care goals over the study period. Given that diabetes morbidity and mortality are mainly due to complications that accumulate over time,²⁰ interventions to promote early control of risk factors for young adults with diabetes should be further expanded.^21,22 Also, access to care and sex disparities in treatment of cardiac risk factors may have contributed to the sex disparities.^23,24 Moreover, while there may be some underlying physiologic differences in the prevalence of hypertension and insulin deficiency that correlate with race/ethnicity,^1,25 to date, access to and effectiveness of health care remain major factors that have not been adequately addressed on a population level despite numerous innovative interventions targeting these populations.^26-30

Despite notable advances in diabetes drug development (eg, US Food and Drug Administration approval of dipeptidyl peptidase-4 inhibitors and glucagon-like peptide-1 agonists to improve glycemic control early in the study period),^31-33 multiple, well-publicized guidelines for BP and cholesterol management, and many attempts to create new care models for diabetes,^34-36 the diabetes care cascade has not significantly improved since the early 2000s. NHANES data indicate that these developments are not effectively reaching the population at large, which could be in part associated with the high drug costs. In addition, barriers to access of care remain an important consideration: racial/ethnic disparities did not reach statistical significance in the cohort linked to care.

Continued attention to the diabetes care cascade may help to monitor the progress and motivate patients and health care professionals to work toward achieving treatment targets and improving population-based outcomes. For example, in Minnesota, continuous monitoring of health care quality and clinic-level performance through annual reports has contributed to sustained improvements in meeting care targets. Although in 2004 only 11.9% of patients with known diabetes in Minnesota achieved optimal diabetes care, defined similarly in that study and our analysis, this rate increased to 44.7% in 2017—more than twice the estimated national value.^37,38

Our findings are limited by the study design. First, we used cross-sectional survey and examination data from noninstitutionalized civilian adults to approximate the diabetes care cascade for the US resident population. Also, since we used cross-sectional data, the results are snapshots of diabetes care during the survey periods. Second, on par with previous research, we used the most recent guidelines to define diabetes and risk factor control and applied those definitions retroactively, albeit consistently, across all survey periods.^6,9 Nevertheless, supplemental analyses illustrating the distribution of HbA_1c levels, BP, and LDL-C levels over time, as well as the proportion of patients who achieved lipid level control for primary and secondary prevention of CVD, showed little change over time. In addition, although the American Diabetes Association began recommending the use of HbA_1c levels for diabetes diagnosis in 2010,³⁹ we used both HbA_1c and FPG criteria retroactively across all periods. Third, we did not assess clinical outcomes in this study. Fourth, we created logistic regression models to predict various outcomes and used aORs to assess the association between different variables and achievement of diabetes care targets. The aOR is considered a close approximation to the relative risk when the prevalence of the risk is low but tends to exaggerate the magnitude of the outcome otherwise. Because the magnitude of the effect sizes in this study is not large, the qualitative judgments and conclusions should be valid.⁴⁰ In addition, although we did not differentiate between type 1 and type 2 diabetes, treatment targets are similar for both.

Despite major advances in drug discovery and movement to develop innovative diabetes care delivery models over the past 2 decades, the diabetes care cascade did not appear to improve for US adults with diabetes between 2005 and 2016. More-frequent diabetes screening, expanded access to care and health insurance, and interventions to improve patients’ adherence to medication and to reduce clinical inertia must remain strategies to improve diabetes outcomes in the United States.^41-44 In addition, the results of this study suggest that recent advances in diabetes treatment have not effectively reached the populations at risk and may indicate an immediate need for better approaches to diabetes care delivery, including a continued focus on reaching underserved populations with persistent disparities in care.

Accepted for Publication: May 13, 2019.

Corresponding Author: Pooyan Kazemian, PhD, Medical Practice Evaluation Center, Massachusetts General Hospital, 100 Cambridge St, Ste 1600, Boston, MA 02114 (pooyan.kazemian@mgh.harvard.edu).

Published Online: August 12, 2019. doi:10.1001/jamainternmed.2019.2396

Author Contributions: Dr Kazemian had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kazemian, McCann, Walensky, Wexler.

Acquisition, analysis, or interpretation of data: Kazemian, Shebl.

Drafting of the manuscript: Kazemian, Walensky, Wexler.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Kazemian, Shebl, Walensky.

Administrative, technical, or material support: Kazemian, McCann, Walensky.

Supervision: Kazemian, Walensky, Wexler.

Conflict of Interest Disclosures: Dr Wexler reported that her husband has equity in Apolo1bio. No other disclosures were reported.

Funding/Support: Research reported in this publication was supported by the Boston Area Diabetes Endocrinology Research Center P30DK057521 grant and the Steve and Deborah Gorlin MGH (Massachusetts General Hospital) Research Scholar Award.

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health, MGH, or Harvard University.