Article Text

Not all healthcare inequities in diabetes are equal: a comparison of two medically underserved cohorts
  1. Ashby F Walker1,2,
  2. Michael J Haller1,3,
  3. Ananta Addala4,5,
  4. Stephanie L Filipp3,
  5. Rayhan Lal6,
  6. Matthew J Gurka7,
  7. Lauren E Figg5,
  8. Melanie Hechavarria3,
  9. Dessi P Zaharieva5,
  10. Keilecia G Malden3,
  11. Korey K Hood4,5,
  12. Sarah C Westen8,
  13. Jessie J Wong4,5,
  14. William T Donahoo1,9,10,
  15. Marina Basina4,6,
  16. Angelina V Bernier3,
  17. Paul Duncan2,
  18. David M Maahs4,5
  1. 1University of Florida Diabetes Institute, Gainesville, Florida, USA
  2. 2Department of Health Services Research, Management and Policy, University of Florida, Gainesville, Florida, USA
  3. 3Department of Pediatrics, University of Florida, Gainesville, Florida, USA
  4. 4Stanford Diabetes Research Center, Stanford, California, USA
  5. 5Division of Endocrinology, Department of Pediatrics, Stanford University School of Medicine, Stanford, California, USA
  6. 6Division of Endocrinology, Department of Medicine, Stanford University School of Medicine, Stanford, California, USA
  7. 7Department of Public Health Sciences, University of Virginia, Charlottesville, Virginia, USA
  8. 8Department of Clinical and Health Psychology, University of Florida, Gainesville, Florida, USA
  9. 9Division of Endocrinology, Diabetes, & Metabolism, College of Medicine, University of Florida, Gainesville, Florida, USA
  10. 10Department of Health Outcomes and Biomedical Informatics, College of Medicine, University of Florida, Gainesville, Florida, USA
  1. Correspondence to Dr Ashby F Walker; afwalker{at}ufl.edu

Abstract

Introduction Diabetes disparities exist based on socioeconomic status, race, and ethnicity. The aim of this study is to compare two cohorts with diabetes from California and Florida to better elucidate how health outcomes are stratified within underserved communities according to state location, race, and ethnicity.

Research design and methods Two cohorts were recruited for comparison from 20 Federally Qualified Health Centers as part of a larger ECHO Diabetes program. Participant-level data included surveys and HbA1c collection. Center-level data included Healthcare Effectiveness Data and Information Set metrics. Demographic characteristics were summarized overall and stratified by state (frequencies, percentages, means (95% CIs)). Generalized linear mixed models were used to compute and compare model-estimated rates and means.

Results Participant-level cohort: 582 adults with diabetes were recruited (33.0% type 1 diabetes (T1D), 67.0% type 2 diabetes (T2D)). Mean age was 51.1 years (95% CI 49.5, 52.6); 80.7% publicly insured or uninsured; 43.7% non-Hispanic white (NHW), 31.6% Hispanic, 7.9% non-Hispanic black (NHB) and 16.8% other. Center-level cohort: 32 796 adults with diabetes were represented (3.4% with T1D, 96.6% with T2D; 72.7% publicly insured or uninsured). Florida had higher rates of uninsured (p<0.0001), lower continuous glucose monitor (CGM) use (18.3% Florida; 35.9% California, p<0.0001), and pump use (10.2% Florida; 26.5% California, p<0.0001), and higher proportions of people with T1D/T2D>9% HbA1c (p<0.001). Risk was stratified within states with NHB participants having higher HbA1c (mean 9.5 (95% CI 8.9, 10.0) compared with NHW with a mean of 8.4 (95% CI 7.8, 9.0), p=0.0058), lower pump use (p=0.0426) and CGM use (p=0.0192). People who prefer to speak English were more likely to use a CGM (p=0.0386).

Conclusions Characteristics of medically underserved communities with diabetes vary by state and by race and ethnicity. Florida’s lack of Medicaid expansion could be a factor in worsened risks for vulnerable communities with diabetes.

  • Diabetes Mellitus, Type 1
  • Diabetes Mellitus, Type 2
  • Healthcare Disparities
  • Health Policy

Data availability statement

Data are available upon reasonable request by directly reaching out to the ECHO Diabetes PIs: Dr Ashby Walker (afwalker@ufl.edu), Dr David Maahs (dmaahs@stanford.edu), or Dr Michael Haller (hallemj@peds.ufl.edu). Codes for analysis are available by contacting Dr Matthew Gurka (mjg3p@virgina.edu).

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • Health disparities in diabetes are well documented based on socioeconomic status, race, and ethnicity within the USA.

WHAT THIS STUDY ADDS

  • Not all medically underserved communities are underserved in the same way, and this study highlights that risks are stratified even within similar cohorts of people with diabetes based on race, ethnicity and geographic location.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Interventions aimed at reducing health disparities in diabetes should be uniquely developed with awareness of specific risks for different communities. Advocacy for state-level policy decisions that best support medically underserved communities should be at the forefront of multisector efforts to reduce health disparities in diabetes.

Introduction

Disparities in health outcomes for people living with type 1 diabetes (T1D) and type 2 diabetes (T2D) in the USA are well documented, with substantially higher risks of disease morbidity and mortality for low socioeconomic status (SES), non-Hispanic black (NHB), American Indian, and Hispanic communities.1–6 Low SES communities living with diabetes face a myriad of barriers in receiving the recommended preventive endocrinology care needed to optimize outcomes.7–10 Due to these barriers, low SES communities with diabetes often rely on primary care providers (PCP) for their diabetes care.9–11 Moreover, technologies that have recognized benefits for improving outcomes like HbA1c and reduction of diabetes-related complications are particularly underused in low SES and NHB communities.1 12–16 The development of targeted interventions to reduce diabetes-related health disparities for low SES populations and communities at risk based on race and ethnicity is vital. Given wide variation in healthcare policies in the USA between states, it is also imperative to understand how state-level healthcare policies may impact medically underserved communities living with diabetes. Thus, disparate outcomes in diabetes are likely further stratified according to state-level policy context.

To address health disparities, the Project Extension for Community Healthcare Outcomes (ECHO) model was founded at the University of New Mexico as a provider empowerment program, whereby subspecialists in a given area (referred to as the ‘hub’) train PCPs to deliver care within primary care settings (referred to as ‘spokes’).17 18 This pioneering approach was founded on the recognition that communities at greatest risk for health disparities are more likely to trust and rely on PCPs in their local areas than on subspecialists. The tele-education model of the Project ECHO was first used in New Mexico for dissemination of hepatitis C management knowledge to PCPs in rural areas, and, since that initial iteration, the model is now being used in 193 countries to address complex challenges in healthcare for many chronic and infectious diseases.19

While there are currently more than 30 diabetes-specific Project ECHO programs globally, ECHO Diabetes, based in California and Florida, is one of the only ECHO programs designed to specifically address insulin-requiring diabetes management while rigorously evaluating the impact of this intervention on patient-level, provider-level, and center-level outcomes.6 20–22 Figure 1 provides an overview of data collection efforts of the full ECHO Diabetes trial. The aims of the ECHO Diabetes program are being accomplished through a variation of a stepped-wedge design and strategic efforts to recruit health centers (‘spokes’) that were in areas high in the Neighborhood Deprivation Index and centers that are Federally Qualified Health Centers (FQHC) or FQHC ‘look-alikes’.20–22 FQHCs provide care to over 30 million people in the USA and offer a critical safety net of comprehensive primary care to people living in medically underserved areas in the USA.23–25 Key terms related to FQHCs and the Health Resources and Services Administration guidelines are defined under the Consolidated Health Center Program in section 1905(l)(2)(B) of the Social Security Act [1].23 Federal requirements in the USA mandate that health centers must meet a uniform set of rigorously evaluated criteria to qualify as an FQHC including being located in a medically underserved area, never turning anyone away based on insurance status, allowing people to pay on a sliding scale based on income, and providing comprehensive primary care services.23

Figure 1

Timeline of data collection for the ECHO Diabetes Stepped-Wedge Trial. Data presented in this manuscript reflect the baseline characteristics of the participant-level and spoke-level cohorts recruited for the ECHO Diabetes Stepped-Wedge Trial. This timeline shows all data points across the trial in its entirety. ECHO, Extension for Community Healthcare Outcomes; EHR, electronic health record; HEDIS, Healthcare Effectiveness Data and Information Set.

In this paper, we describe the baseline characteristics of the cohort recruited in the ECHO Diabetes program evaluation from participating spoke sites across California and Florida. California and Florida are two of the most populous states in the USA, are both richly diverse with large Hispanic and immigrant communities, and both have well-documented gaps in healthcare for medically underserved populations.26 An important area of distinction between these states, however, is that California opted to expand Medicaid eligibility through the Affordable Care Act, whereas Florida did not.26 27 Thus, beginning in 2014 in California, Medicaid enrollment allowed for expanded eligibility levels to include those up to 138% of the federal poverty level.26 27 This resulted in a larger number of people qualifying for Medicaid in California and the expanded eligibility threshold reduced the number of people with no health insurance from working poor households that otherwise could not afford private insurance. This pivotal difference in policy context for the two states, coupled with the unique inclusion of communities from FQHC and community health center settings, provides an unparalleled snapshot of differences between and within each state for medically underserved people living with diabetes. We hypothesized that people with diabetes who had their care supported by the Affordable Care Act in California would have better baseline health characteristics than those who did not in Florida. Though our study was not designed as a longitudinal policy analysis, the opportunity to compare the baseline characteristics of the ECHO Diabetes cohorts contributes to a better understanding of potential differences within communities that are medically underserved and heightened awareness surrounding the need for uniquely tailored interventions.

Research design and methods

Participant-level data

Data collected at the participant level involved recruitment of people with insulin-requiring diabetes through informed consent across two recruitment phases in the summer and winter of 2021 in the states of California and Florida using protocols approved by institutional review boards (IRB). Participants were recruited from their medical home where they receive primary care services (ECHO Diabetes ‘spoke’ sites). 11 spoke sites participated in the state of California and nine in Florida. Participating spokes provided recruitment lists based on the following eligibility criteria: (1) diagnosis of T1D or T2D, (2) using intensive insulin management, (3) aged 18 and older and (4) seen for care at the spoke twice in the past year or once in each of the past two consecutive years. For the purpose of the study, eligible individuals used an insulin pump or at least two daily injections consisting of at least one long-acting and one rapid-acting insulin. The eligibility requirement related to frequency of receiving care was to ensure participants were receiving routine care at the spoke site. The recruitment lists were randomized to ensure non-bias, and recruitment attempts took place in the order of randomization. Strategic efforts were made to use study staff for enrollment that had direct connections to diabetes, to underserved communities through lived experience, and that spoke both English and Spanish.28

Recruitment took place by phone or at the spoke sites. Demographic characteristics were collected directly from the participant or obtained via the electronic health record (EHR), and included: diabetes type, date of birth, sex, race/ethnicity, insurance type, along with language and modality preference for survey administration. Baseline and post-surveys were administered in the preferred language electronically using Research Electronic Data Capture,29 30 or via hard copy sent to the participant with a prepaid envelope for return. The Single-Item Literacy Screener31 was also included.

Participant surveys were administered at baseline, and 6 or 12 months after enrollment, depending on the phase of stepped-wedge enrollment of the spoke where they received care. Surveys spanned a range of diabetes-related topics including insulin administration and technology use, diabetes-related complications, social support networks, the Diabetes Distress Scale,32 and the Trust in Physician Scale.33 Likewise, home HbA1c kits were sent at 6-month intervals after enrollment. If patients did not complete a home HbA1c kit, EHR data were extracted, where available, within the predefined data capture time windows (May to July 2021, November 2021 to January 2022, May to July 2022). The Advanced Research and Diagnostic Laboratory at the University of Minnesota provided HbA1c kits and processed all home HbA1c kit samples collected. The Advanced Research and Diagnostic Laboratory has been continually used in large clinical trials for diabetes research, including the Diabetes Control and Complications Trial.34 Research has also demonstrated that HbA1c values collected using the University of Minnesota kits are highly aligned with and comparable to point-of-care HbA1c values.35 Participants received $20 compensation for each survey they completed.

Health center or spoke-level data

Data were also collected at the spoke level. Recruitment of FQHCs in California and Florida was accomplished by direct outreach efforts with a uniform call to participate flyer and PowerPoint developed to communicate the details of the study. Of all the health centers approached in each state, two opted not to participate in California and one opted not to participate in Florida (both citing staff shortages related to data sharing requirements). A list of participating spokes is presented in online supplemental appendix A. Participating spokes were required to deliver aggregate center-level data annually for calendar years 2020, 2021, and 2022. Data collection included all patients with T1D and T2D. Included measures captured were as follows: overall clinic scope of patients with diabetes by type, use of diabetes technology, average HbA1c, and HbA1c capture, in addition to select Healthcare Effectiveness Data and Information Set (HEDIS) and National Committee for Quality Assurance metrics. Data were reported overall and by insurance type; with the exception of HbA1c capture rates, data were reported separately among T1D and T2D. HEDIS metrics were also reported stratified by race/ethnicity.

Supplemental material

Statistical analysis

All data management and analyses were conducted using Statistical Analysis System (SAS) V.9.4. Demographic characteristics were summarized overall and stratified by state (frequencies and percentages, means (95% CIs)). Generalized linear mixed models were used to compute model-estimated rates and means with 95% CIs using the logit and identity links, respectively. A random effect for spoke was included to account for clustering within spoke sites. This modeling framework was used to estimate rates of insulin pump and continuous glucose monitor (CGM) use, along with modality of insulin delivery overall. These models also provided us the ability to compare outcomes between states in addition to HEDIS metrics by state and year, overall and within insurance and racial/ethnic subgroups. Statistical comparisons were made via these models on demographic characteristics between those who did and did not use an insulin pump, and those who did and did not use CGM regularly (α=0.05). Additional comparisons were made via these models on demographic characteristics between states and years for HEDIS metrics. Mean baseline HbA1c was also compared across key demographic characteristics. A sensitivity analysis was conducted to ensure there was no statistically significant difference in the central tendency of the HbA1c kits versus HbA1c collected from the Electronic Medical Record (EMR).

Results

Overall demographic characteristics of cohorts

For participant-level data, 582 people living with diabetes (aged 18 and older) using Multiple Daily Injections (MDI) were recruited across California and Florida from participating spoke sites (table 1 and online supplemental appendix B). Overall, the recruitment rate at the participant level for California was 52.7% and 21.5% for Florida.28 In the sensitivity analysis, participant level data indicated that the demographics of the cohort mirrored the larger population of people with diabetes at participating spokes.28 There were 196 (33.0%) with T1D and 386 (67.0%) with T2D, with a mean age of 51.1 years (95% CI 49.5, 52.6) and a mean HbA1c of 8.7% (95% CI 8.3, 9.0). There were relatively equal gender ratios, with 47.8% female, 46.1% male, and 6.1% unknown sex. In terms of race and ethnicity, 43.7% were non-Hispanic white (NHW), 7.9% were NHB, 31.6% were Hispanic, and 16.8% reported other or unknown for race and ethnicity, and the majority of the cohort listed English as their language preference (84.8%) versus Spanish (15.2%). Among the participants, 19.3% were insured on commercial plans, 20.6% on Medicare, 33.1% on Medicaid, 3.3% via dual-eligible Medicare-Medicaid, and 23.7% were uninsured/self-pay. Technology use was low overall in the cohort, with 83.2% reporting they did not use an insulin pump and 74.1% reporting they did not regularly use CGM. Table 1 provides a summary of demographic characteristics for the overall cohort and for California and Florida. Table 2 shows how technology use and HbA1c varied by demographic characteristics overall.

Supplemental material

Table 1

Participant-level ECHO Diabetes demographics by state

Table 2

Overall participant-level technology use and HbA1c by demographics

For spoke-level data, a cohort of 32 796 people (aged 18 and older) living with T1D and T2D were represented in the aggregate data provided by the health centers (3.4% had T1D and 96.6% had T2D—see table 3). In terms of insurance coverage, the overall panels of adults seen at the participating health centers included 27.4% with commercial plans, 34.0% with Medicaid, 10.8% with Medicare, 4.6% were dual eligible, and 23.3% were uninsured. For the adults living with T1D, the rate of HbA1c>9.0% at baseline among those who were NHW was 29.7%, among NHB was 51.4%, among Hispanic was 36.3%, and among other race/ethnicities was 31.5% (table 4 and online supplemental appendix B). For adults living with T2D, the rate of those who had HbA1c>9% at baseline among those who were NHW was 22.4%, among NHB was 30.1%, among Hispanic was 24.1%, and among other race/ethnicities was 22.9% (table 4). When it came to reporting technology use, only seven of the 20 participating spokes could provide these data for CGM and only six for insulin pumps, as this information was not captured in their EHRs at the beginning of the stepped-wedge trial. Of the health centers that could provide diabetes technology data, at baseline, only 2.0% of the overall spoke cohort were using CGM and <1.0% were using insulin pumps.

Table 3

Center-level ECHO Diabetes aggregate patient demographics

Table 4

Center-level comparison between states at baseline: HbA1c HEDIS data

Differences between states

There were statistically significant differences between the cohorts from California and Florida seen in the participant-level cohort (table 1). There were significantly higher numbers of NHB participants in Florida than in California (23.9% vs 2.4%, p=0.0004), respectively, and higher numbers of uninsured/self-pay participants in Florida than in California (32.9% vs 16.7%, p<0.0001), respectively. While overall levels of technology use were low for both states, Florida lagged behind California with only 10.2% reporting insulin pump use in Florida versus 26.5% in California (p<0.0001, table 1), and 18.3% reporting using a CGM regularly in Florida versus 35.9% in California (p=0.0101, table 1).

Differences in baseline health characteristics were also mirrored in the larger aggregate spoke-level data (table 4). At baseline, Florida had a significantly higher rate of adults with T1D and T2D with a baseline HbA1c>9%. In California, 24.4% of adults aged 18–75 with T1D had a baseline HbA1c>9.0%, compared with 39.3% in Florida (p<0.0001). Likewise, at baseline, 19.1% of people with T2D had HbA1c>9.0% in California, compared with 28.8% in Florida (p>0.0001). When aggregated to the state level, Florida had an overall higher rate of uninsured adults compared with California (27.7% vs 16.4%, respectively, table 4).

Differences within states

There were notable differences between groups within the participant-level cohort as well as the spoke-level cohort. First, in the participant-level data, NHB participants had the highest baseline HbA1c levels with a mean of 9.5% (95% CI 8.9%, 10.0%) compared with NHW with a mean of 8.4% (95% CI 7.8%, 9.0%), Hispanic participants with a mean of 8.9% (95% CI 8.4%, 9.3%), and other/unknown race with a mean of 8.4% (95% CI 7.8%, 9.0%) (p=0.0058, table 2). In spoke-level data, the rate of NHB with baseline HbA1c>9% was notably higher than other racial and ethnic groups (table 4). Second, technology use was higher for NHW participants (table 2). Among those who self-reported use of an insulin pump, 66.3% were NHW, 7.6% were NHB, 16.6% were Hispanic, and 9.5% were other race/ethnicities (p=0.0426, table 2). The same was true for CGM use, where 53.3% were NHW, 10.9% were NHB, 19.8% were Hispanic, and 16.0% were other race/ethnicities (p=0.0192). Language preference was also associated with significant differences in technology when it came to CGM use, with only 8.0% of those using CGM preferring Spanish and 92.0% preferring to speak English (p=0.0386). The strong majority of those using an insulin pump were individuals with T1D (89.1%, p<0.0001), and the use of CGM was statistically significantly higher for individuals with T1D (55%, p<0.0001).

Conclusions

Baseline characteristics of the ECHO Diabetes cohort demonstrate that there are important distinctions in medically underserved communities based on state location (figure 2). Moreover, baseline characteristics of the ECHO Diabetes cohort show that there are stratified health risks, even within medically underserved communities in the USA, based on race, ethnicity, insurance type, and perhaps state-dependent healthcare policy decisions. All of the ECHO Diabetes participants received routine primary care services at an FQHC or FQHC look-alike facility and were thus defined by the Health Resources and Services Administration as being medically underserved.36 They also define medically underserved areas as geographic catchment areas with populations that have inadequate access to healthcare with lack of PCPs, high infant mortality rates, high poverty rates, and/or a high elderly population.23 36 Specific population groups that are called out as being at particular risk include rural, elderly, low literacy, blue collar, low SES and ‘groups who face economic, cultural, or linguistic barriers to health care, and limited access to services populations’.36 Despite these similarities, baseline characteristics were statistically significantly different according to state location as well as race and ethnicity.

Figure 2

Comparison of medically underserved cohorts from California and Florida. This figure compares the baseline characteristics of the ECHO Diabetes patient-level (n=582) and center-level (n=32 796) cohorts from California and Florida including insurance status, insulin pump use, continuous glucose monitor (CGM) use, and the proportion of people with HbA1c>9%. ECHO, Extension for Community Healthcare Outcomes; T1D, type 1 diabetes; T2D, type 2 diabetes.

While race is never specifically identified as part of the Health Resources and Services Administration definition of a medically underserved area, other entities like the National Institutes of Health categories of those that are historically under-represented do include race. Further, national indexes that document associations between geographic location and health risks, like the Neighborhood Deprivation Index,22 classify the racial composition of census tracks as part of their criteria—identifying unique risks for areas where NHB and Hispanic communities live. These broader definitions acknowledge the impact of structural racism5 on health and pervasive inequities that inform health outcomes in the USA.

It is important to note that our study was not designed to evaluate longitudinal impacts of state-level Medicaid policy related to the Affordable Care Act and the observed associations are ecologic. However, the diversity of the cohorts included in this comparison with regards to race, ethnicity, and SES are vastly under-represented in most clinical trials and in large national data repositories for T1D.1 Moreover, FQHC populations are similarly under-represented in studies that solely recruit from endocrinology settings. Thus, our ability to compare these cohorts based on state location presents a rare opportunity to better understand differences within as well as between the groups. The implications of what our ECHO Diabetes cohorts demonstrate on differences in patient populations by state speak to the critical importance of policy-level change in the USA for people living with diabetes. Both states made active, intentional, policy decisions about Medicaid expansion as provided for in the Affordable Care Act. In both instances, the decisions were significantly political and aligned with partisan and ideological points of view. The downstream impact of these partisan-driven policy changes is seen in significantly greater numbers of those who are uninsured in Florida.

Navigating the complex challenges of diabetes is difficult for those with robust insurance and far more daunting for the many who face diabetes management without any meaningful insurance support.3 37 38 There is a critical need for national policies that ensure access to diabetes technologies like insulin pumps and CGM for people living with diabetes, especially for medically underserved communities, as automated insulin delivery has become standard of care for insulin-requiring diabetes.37 Under the 340B federal drug program, people receiving care at FQHCs receive significant discounts on ‘life preserving’ medications including insulin.39 Currently, CGM and insulin pumps are not included in the 340B program and adding them to the discount formulary could improve access for a large population of people living with diabetes who face elevated risks due to economic deprivation.

There are many potential confounders that could impact the data presented in this comparison of cohorts from California and Florida. The Southeastern United States region generally has elevated health risks for stroke and diabetes due to high concentrations of poverty as well as a long-standing history of social welfare policy that has lagged behind the rest of the USA.40 41 Other study limitations include a relatively small number of participants from only two states and under-representation from Asian Pacific Islanders and American Indians. Also, the response rate for participant-level recruitment was low for Florida. The lower recruitment rate for Florida was likely due to a differential requirement by the IRB at each institution: California operated under a waiver of signed informed consent for participants (ie, participants were able to give verbal consent), whereas Florida’s IRB required signed informed consent and that presented additional barriers.28 If anything, this would have potentially biased the Florida cohort to be over-represented in terms of those with higher levels of health and technology literacy as it required the use of e-consent via smartphone or computer for those recruited by phone. Even with this additional barrier, the participant-level cohort mirrored the demographic characteristics of the larger spoke-level cohort and both Florida cohorts lagged behind California’s in key diabetes outcomes. Data triangulation with the larger spoke-level cohort, which included comprehensive data for all people with diabetes at participating spokes, helped minimize bias that may be present in any study that relies exclusively on participant-level data.

Despite our study limitations, characteristics of these cohorts point to the dire need for interventions that are specifically tailored for NHB and Spanish-speaking communities living with insulin-requiring diabetes. Our cohort provides a unique opportunity to compare features of risk associated with race and ethnicity within a context where all participants are medically underserved, even those who are NHW. While some studies have identified differences in HbA1c between NHB and NHW individuals with similar glycemia,42 this does not explain the full difference as demonstrated in other research that includes consideration of systemic inequities for NHB communities.4–6 15 What we see is that race and ethnicity compound health risks for those that are medically underserved with insulin-requiring diabetes. This is referred to in social science paradigms as intersectionality43 and also double jeopardy44—where overlapping categories of SES, race, and ethnicity create elevated risks for health outcomes. Addressing health disparities in diabetes will require broad-scale recognition that not all medically underserved communities are underserved in the same way, and that structural racism must be incorporated into how ‘medically underserved’ is defined and policy solutions found to providing equitable healthcare delivery.

Data availability statement

Data are available upon reasonable request by directly reaching out to the ECHO Diabetes PIs: Dr Ashby Walker (afwalker@ufl.edu), Dr David Maahs (dmaahs@stanford.edu), or Dr Michael Haller (hallemj@peds.ufl.edu). Codes for analysis are available by contacting Dr Matthew Gurka (mjg3p@virgina.edu).

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was reviewed and approved by the Institutional Review Board at the University of Florida (IRB201903243) and the Institutional Review Board at Stanford University (54198). Participants gave informed consent to participate in the study before taking part.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors AFW, DMM, MJG, MJH, and SLF designed the study, oversaw the conduct of the study, and wrote the manuscript. MJG and SLF had access to data and conducted the analysis. AA, RL, LEF, MH, DPZ, KGM, KKH, SCW, JJW, WTD, MB, AVB, and PD contributed input to the study design, reviewed the findings, and provided critical revisions to the manuscript. AFW and DMM are the guarantors of this work and, as such, had full access to all the data and take full responsibility for the integrity of the data and the accuracy of the analysis.

  • Funding Funding for this study was provided by the Leona M and Harry B Helmsley Charitable Trust (2005-03934).

  • Map disclaimer The inclusion of any map (including the depiction of any boundaries therein), or of any geographic or locational reference, does not imply the expression of any opinion whatsoever on the part of BMJ concerning the legal status of any country, territory, jurisdiction or area or of its authorities. Any such expression remains solely that of the relevant source and is not endorsed by BMJ. Maps are provided without any warranty of any kind, either express or implied.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.