Ensuring Equitable Care in Diabetes Management Among Patients of Health Resources & Services Administration–Funded Health Centers in the United States

We used data from the 2014 Health Center Patient Survey (HCPS), which was sponsored by the HRSA Bureau of Primary Health Care. HCPS was a cross-sectional, nationally representative in-person survey of patients served by HCs. Interview questions focused on sociodemographic characteristics, health conditions and behaviors, and access to and utilization of health care services. The HCPS oversampled patients who identified as non-Hispanic Asian, non-Hispanic AI/AN, or non-Hispanic Native Pacific Islander and those ≥65 years of age to improve the representation of these groups. A three-stage sampling design to select HC organizations, sites, and patients was used to obtain the sampling frame. A total of 169 HC organizations, 520 sites within those HCs, and a random sample of patients who entered the sites and those with at least one prior visit to the HC in the past year were eligible to participate in the survey. A total of 7,002 patient interviews were completed. The response rate was 66% of the total patients initially identified and 91% among patients confirmed to be eligible.

We also included data from the 2014 Uniform Data System (UDS), an administrative data set reported by all HC organizations that includes information on HC patient demographics, size and staffing, types of services delivered, clinical quality measures, and revenues. UDS captures aggregate information at the HC organization level rather than individual delivery sites within the organization. UDS data were merged with the HC where each respondent was interviewed (24). For our study, we restricted the sample to adult HC patients who were ≥18 years of age (n = 5,592), diagnosed with nongestational diabetes by a health care professional (n = 1,199), and identified the HC where they were interviewed as their usual source of care (n = 1,129). We also restricted our analysis to 858 patients with type 2 diabetes and excluded 271 respondents with type 1 diabetes because of potential differences in management of such patients. We lacked data on whether patients had other forms of diabetes; therefore, such patients may have been included in our sample. We also excluded 22 respondents who were non-Hispanic Native Hawaiian or Pacific Islander and other race/ethnicity because of heterogeneity (e.g., rates of type 2 diabetes, care-seeking patterns, and provider interactions) of these groups and our inability to study them separately due to their sparse sample size (25). Thus, our final analytical sample was 836 patients.

Diabetes management indicators were selected from the ADA’s standards of care (10). We examined receipt of at least two A1C tests annually (vs. fewer or none), foot examination (i.e., seen or talked with a foot doctor in the past year vs. not), eye examination (i.e., seen or talked with an eye doctor in the past year vs. not), and timely cholesterol check (i.e., patient’s last cholesterol check by a provider was less than 1 year vs. ≥1 year ago). We also examined care management interventions such as receipt of a referral to a specialist (vs. no referral), individualized treatment plan (vs. no individualized plan), and a telephone call; an appointment with the nurse; or a home visit from the HC to teach diabetes self-care (vs. no self-care services) (10).

We also examined three negative diabetes-related outcomes, including being told by a provider at the last visit that blood glucose levels were too high (vs. too low or just right), having a diabetes-related hospitalization or ED visit within the past year (vs. none), and being very confident in ability to manage diabetes (vs. somewhat/not too/not at all confident). We considered confidence in diabetes management independent of diabetes management indicators because confidence could be affected by primary care interventions such as health education and provider counseling. We also examined indicators of high service use, including five or more outpatient visits to the HC in the past year (vs. fewer), any all-cause ED visits in the past year (vs. none), and all-cause hospitalization after an ED visit (vs. not hospitalized or did not visit an ED). We chose to examine five or more primary care visits because most patients with well-managed diabetes may require up to four visits in quarterly intervals (26). A detailed description of dependent variable construction and survey questions are presented in Supplementary Table S1.

Our primary independent variable of interest was race/ethnicity, which included non-Hispanic White, Hispanic/Latino, non-Hispanic African American, non-Hispanic Asian, and non-Hispanic AI/AN.

We used Andersen’s model of health care utilization to control for other determinants of utilization organized by predisposing, enabling, and need factors (27). Predisposing factors included self-reported age (18–44 vs. 45–64 or ≥65 years), sex, education (high school graduate or equivalent or more education vs. less than high school education), employment status (employed vs. unemployed or not in the labor force), area of residence (urban vs. rural), and U.S. Census region of residence (Northeast vs. South, Midwest, and West). Enabling factors included health insurance status (Medicaid vs. uninsured or other), federal poverty guideline (≥100 vs. 101–199 or ≥200%), and length of time the patient had been a patient at the HC as a measure of continuity of care (<1 vs. 1–3 or ≥3 years).

Need factors included self-reported weight status (underweight or normal weight vs. overweight/obesity). Patients were considered to have underweight or normal weight if their BMI was <25 kg/m², overweight if their BMI was 25–29.9 kg/m², or obesity if their BMI was ≥30 kg/m². Other need determinants included self-rated health status (fair or poor vs. excellent, very good, or good); needing help with activities of daily living; diagnosis of hypertension, asthma, or cardiovascular disease; length of time the patient had been diagnosed with diabetes (within the past 2 years vs. longer); and insulin use. HC characteristics were included as additional control variables. These included the number of HC patients in the last year as a measure of HC size and revenue per capita to quantify success in generating revenues.

We created generalized multilevel structural equation logistic regression models to assess the relationship of race/ethnicity and the dependent variables while controlling for other confounding factors. We included respondents with complete responses to all of the variables studied and excluded those with any missing data. We weighted the analyses to account for the complex survey methodology of the HCPS. We further rescaled the weights to reduce the influence on the results of HCs that had a large number of patients. We discuss results with probability values of 0.05 or less in the text. Analyses were conducted using Stata, v. 16.0, statistical software.

Sample characteristics overall and by race/ethnicity are displayed in Table 1. Patients in our sample were mostly middle-aged (45–64 years [54%]), were high school graduates (65%), and lived in the Southern region of the U.S. (35%). They were also predominately Medicaid insured (46%), poorer (59% at ≤100% of the federal poverty guideline), long-term users of the HC (59% at >3 years). Many had a co-occurring hypertension diagnosis (81%) and reported fair or poor health (65%). There were statistical differences in predisposing and need characteristics between racial/ethnic groups. For example, more non-Hispanic Whites had a high school education or more, fewer Asians reported having obesity weight status, more African Americans and Whites had hypertension, more non-Hispanic Whites had cardiovascular disease, and more non-Hispanic Whites and African Americans used insulin than other groups.

Data on diabetes management showed high rates of A1C testing (76%), eye doctor visits or contacts (73%), cholesterol checks (87%), and having an individual treatment plan (68%) in the last year (Table 2). But, the rate of foot doctor visits was lower (38%). Data on diabetes outcomes showed that more than half (54%) had blood glucose levels that were too high, although about half reported having high confidence in diabetes self-management (49%) and fewer (13%) reported diabetes-related ED visits or hospitalizations in the past year. General outcomes included high rates of five or more HC visits (44%), one or more all-cause ED visits (54%), and all-cause hospitalization after an ED visit (45%).

Table 3 presents the association of race/ethnicity and all dependent variables from regression models. We found racial/ethnic differences among several diabetes management variables. Non-Hispanic African American (odds ratio [OR] 0.25), Hispanic/Latino (OR 0.26), and non-Hispanic Asian (OR 0.11) patients with diabetes all had lower odds of having a cholesterol check <1 year ago compared with non-Hispanic Whites. Non-Hispanic AI/AN had higher odds (OR 2.55) of foot doctor visits than non-Hispanic White patients. Hispanic/Latinos had lower odds (OR 0.45) of receiving an individual diabetes management plan. There were no significant racial/ethnic differences in diabetes-related outcomes. However, non-Hispanic African American patients had lower odds (OR 0.43) of having five or more office visits than non-Hispanic White patients. Full regression model results are displayed in Supplementary Tables S2–S4.

Our analyses showed racial/ethnic parity in several indicators and some disparities for specific groups in several other indicators for HC patients with type 2 diabetes. Our findings of racial/ethnic parity in most of the diabetes management indicators (A1C, eye doctor visits, and specialist referrals) counters the existing literature that shows disparities, with fewer A1C tests and foot doctor visits for Hispanics/Latinos (13–15,23,28). Not all patients may require an exam by a podiatrist, although disparities in rates of exams are noteworthy. Our finding of more foot doctor visits among non-Hispanic AI/ANs is consistent with the literature (14,23,29). This finding may be because AI/ANs suffer disproportionately from diabetes-related limb amputations and are less likely than Whites to check their feet for sores or irritation (6,29). The variations in underlying severity or levels of self-care may require further effort to manage some populations with type 2 diabetes. Our finding that Hispanic/Latino patients had a lower likelihood of receiving an individualized treatment plan was supported by a study of the older adult population in California (30). Our finding of disparities in cholesterol checks for nearly all racial/ethnic groups was consistent with the literature (31,32).

Although our findings of parity in diabetes management are positive indicators of equitable treatment of all HC patients, variability among racial/ethnic groups in type 2 diabetes–related adverse outcomes suggests that parity in care management procedures such as eye doctor visits and specialist referrals may not be enough. For example, non-Hispanic African American and Hispanic populations are more likely to develop diabetic retinopathy, and non-Hispanic Native Americans have the highest prevalence rate of diabetic retinopathy (6). Our findings suggest that eye examinations may be low among populations that are at particular risk of diabetic retinopathy. Other research has indicated the importance of primary care provider coordination with eye care providers and the provision of health education to ameliorate barriers to eye care (33,34).

Our findings of parity in type 2 diabetes–related outcomes (high blood glucose levels, ED visits, and hospitalizations) were also positive given the existing evidence that indicates higher rates of these outcomes among non-Hispanic African American populations because of greater severity of disease and lower socioeconomic status (16,35). Parity in self-care confidence in our data may reflect the cultural competence of HC providers and the provision of diabetes health education and literacy assistance to HC patients (36). HCs have chiefly aimed to provide diabetes education and tools for self-management through their clinics (37,38).

HCs have been addressing procedural clinical care strategies for several decades, and these practices are one possible explanation for the parity findings. Health Disparities Collaboratives in many HCs aim to reduce health disparities and improve chronic care and are reported to improve clinical processes in the short term and clinical processes and outcomes in the long term (22).

Although we expected to find disparities in high levels of service use (i.e., multiple HC visits), the direction of differences observed was unanticipated. We found a lower likelihood of having five or more HC visits for non-Hispanic African Americans. Other studies, however, found that non-Hispanic African Americans had more outpatient visits than non-Hispanic Whites and that non-Hispanic African Americans and Hispanics were more likely to have all-cause ED visits and hospitalizations (16,23,35). In one study, the higher number of outpatient visits for African Americans and Hispanic/Latinos was attributable to greater severity of diabetes (39).

The HCPS was a cross-sectional survey and did not allow for causal inference. Survey responses were self-reported and subject to recall and acquiescence bias. For example, patients who remembered their diabetes diagnosis may have had greater disease severity and were more likely to remember receiving diabetes care management from the HC or to report lower confidence in managing their diabetes. We were unable to study all racial/ethnic groups such as non-Hispanic Native Hawaiians and Pacific Islanders because of their small sample sizes. In addition, the number of patients with type 1 diabetes was very small for reliable estimates, and we were therefore unable to assess diabetes management or outcomes for these patients by race/ethnicity. We also lacked data on severity of diabetes to assess whether our findings of parity were desirable given the disproportionate burden of type 2 diabetes by race/ethnicity.

Collectively, our findings indicate racial/ethnic parity in most and disparity in some indicators of care for some HC patients with type 2 diabetes. Further research is needed to assess whether provision of diabetes care management for each racial/ethnic group is commensurate with patients’ diabetes severity.

Addressing disparities identified in the time since last cholesterol check, the provision of individualized treatment plans, and the number of HC visits in this study requires additional action. More research is needed to examine reasons for such disparities. For example, lower rates of recent cholesterol checks may have been the result of lower rates of laboratory orders by providers or the result of patient factors such as limited access to a laboratory. For the former, approaches such as strengthening providers’ compliance with guidelines may be required; for the latter, providing transportation or other means of improving access may be required.

More recent initiatives and awards by HRSA have focused on improving the quality of diabetes care and diabetes management among HCs (40,41). HRSA, through its Health Disparities Reducer category, provides quality improvement awards to HCs that have demonstrated improvement across racial/ethnic groups for a number of outcomes, including diabetes management, hypertension management, and low birth weight (41). Including additional reporting of quality of care by race/ethnicity can further highlight disparities and promote actions to address them. In addition, HCs have moved toward greater use of telehealth and community health workers for type 2 diabetes self-management patient support (42,43). These efforts are likely to improve overall type 2 diabetes outcomes but may have a differential effect on racial/ethnic groups. For example, socioeconomic differences in access to devices with cameras and Internet access may exacerbate disparities, whereas the involvement of community health workers may ameliorate disparities. Similarly, awards and initiatives should address effectively reducing racial/ethnic disparities in both the severity and outcomes of type 2 diabetes.

The mission of HRSA-funded HCs includes “improving the health of the nation’s underserved communities and vulnerable populations by assuring access to comprehensive, culturally competent, quality primary health care services” (44). Within this context, our results indicate the likelihood of progress with room for improvement.

This research was funded by the U.S. Department of Health and Human Services, Health Resources & Services Administration, under contract number HHSH250201300023I. The views expressed in this publication are solely the opinions of the authors and do not necessarily reflect the official policies of the U.S. Department of Health and Human Services or the Health Resources & Services Administration, nor does mention of the department or agency names imply endorsement by the U.S. government.

At the time of this study, B.H. and J.B. were employees of the U.S. Department of Health and Human Services, which funded this study, and A.S. is an employee of the same department. No other potential conflicts of interest relevant to this article were reported.

N.P. and A.S. conceived of the study. N.P. and X.C. directed the statistical analysis and interpreted the data. N.P. drafted the manuscript. C.L. and W.Z. assisted with analysis and interpretation of the data. All authors reviewed and revised the manuscript critically for intellectual content and approved the final manuscript for submission. N.P. is the guarantor of this work and, as such, 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.

Data pertaining to this study were previously presented at the virtual AcademyHealth Annual Research Meeting in June 2021 and the virtual American Public Health Association Annual Meeting & Expo in October 2021.