Abstract

PURPOSE The Enhanced Primary Care Diabetes (EPCD) model is nurse led and leverages interdisciplinary support to improve diabetes quality indicators. The model has been found to be effective overall; however, because narrowing health disparities is a key objective, we aimed to assess differential effectiveness of the model among various racial and ethnic groups.

METHODS This retrospective cohort study compared the time to meeting the D5, a publicly reported quality measure (composite indicator of glycemic and blood pressure control, aspirin use for secondary prevention of cardiovascular disease, statin use, and documented abstinence from tobacco use), after enrollment in the EPCD program by Black, Hispanic/Latine, and Asian patients compared with White patients with diabetes (age 18-75 years) receiving care at 13 primary care practices by multivariable Cox proportional hazards regression. Patients enrolled in the program from January 1, 2020 to December 31, 2020; the study period end date was August 1, 2022.

RESULTS The EPCD program enrolled 1,749 patients (none of whom met the D5 at entry) and 1,061 (60.7%) met the D5 during the study period. Black patients were less likely to meet the D5 compared with White patients (adjusted hazard ratio 0.68; 95% CI, 0.52-0.90; P = .007); there was no difference among Asian and Hispanic patients compared with White patients. Compared with White patients (median 1.1/year; interquartile range [IQR] 0.4, 2.7), Asian patients had fewer nurse touch points (median 0.8/year; IQR 0, 1.4) during the study period, whereas Black patients had more (median 2.2/year; IQR 0.6, 4.0) and Hispanic patients showed no significant difference.

CONCLUSIONS Time to meeting the D5 was longer for Black patients compared with White patients in the EPCD model, despite greater engagement with the care team. Further research is needed to identify factors driving these disparities.

Key words:INTRODUCTION

More than one-half of the 38.1 million US adults with diabetes are people of the global majority (PGM), who are minoritized in the United States and include non-Hispanic Black, Hispanic/Latine, and non-Hispanic Asian individuals.1 The prevalence of diabetes is greater among PGM individuals than among White individuals.1 Black and Hispanic/Latine individuals also have greater rates of acute and chronic diabetes complications2,3 as well as worse control of hyperglycemia, hypertension, and dyslipidemia compared with White individuals with diabetes.4,5 These disparities in diabetes prevalence, management, and outcomes are driven by a broad range of social determinants of health (SDOH) and structural barriers to high-quality care and optimal health. Lower income and education level, lack of insurance and geographic access to health care, and discrimination and racism are among the SDOH associated with worse diabetes outcomes in PGM.6-8

Collaborative practice models such as the Chronic Care Model, endorsed by the American Diabetes Association,9 hold promise to improve outcomes and mitigate disparities in diabetes management and outcomes in PGM populations. A broad variety of models implemented in different primary care settings and serving different patient populations have been described and involve self-management support, interdisciplinary care, protocol development, and use of technology showing improvements in clinical outcomes such as hemoglobin A1c, cholesterol levels, and blood pressure, among others.10 Care team members involved in these models vary greatly and often include a combination of physicians, advanced practice clinicians (ie, nurse practitioners and physician assistants), medical residents, nurses, pharmacists, and social workers. However, few studies have specifically focused on whether these models equitably care for underserved patient populations.

We previously described the implementation11 and effectiveness12,13 of an interdisciplinary nurse-led Enhanced Primary Care Diabetes (EPCD) model across primary care practices of the Mayo Clinic, in Olmsted County, Minnesota. In this model, primary care team registered nurses (RNs) identify adults who have diabetes, are paneled to their care team, and are not meeting the Minnesota Community Measurement (MNCM) comprehensive diabetes care quality indicators (termed the D5),14 to engage them in a longitudinal care pathway aimed at improving patient engagement and access to care, supporting self-management, identification of SDOH with referrals to appropriate care team members and community-based organizations, completion of recommended screening and testing, and treatment intensification/modification to improve care quality and outcomes. This program was shown to improve attainment of the D5, which is a composite indicator of glycemic (hemoglobin A1c < 8%) and blood pressure (<140/90 mm Hg) control, aspirin use for the secondary prevention of cardiovascular disease, statin use, and documented abstinence from tobacco use.14 The objective of this study was to examine the equity of the EPCD program in engaging patients (ie, documentation that the RN has contacted the patient to begin working with them) and helping patients from different racial and ethnic backgrounds meet D5 goals.

METHODSDesign

This was a retrospective cohort study using electronic health record (EHR) data from 13 family medicine and internal medicine practices at Mayo Clinic Rochester. This study was deemed exempt by the Mayo Clinic Institutional Review Board, given that it focused on quality improvement. Results are reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for retrospective cohort studies.15

Setting

Mayo Clinic is an integrated health care delivery system, providing care for local, regional, national, and international patients, with a clinical and academic hub in Rochester, Minnesota. Primary care patients are paneled to a specific clinician; patients generally cannot be paneled to Mayo Clinic Rochester if they also receive primary care elsewhere, ensuring continuity of care and complete capture of clinical encounters and health services. Each practice has access to a clinical pharmacist (on-site or remote) working under a collaborative practice agreement and a social worker (on-site or remote). For specialty diabetes care, referrals can be made either to the endocrinology practice (which serves local and nonlocal patients on a consultative basis) or to a diabetes nurse practitioner who only serves local primary care patients but is part of the endocrinology division. Referrals can also be made to RN-certified diabetes care and education specialists and to registered dieticians, both of whom are part of the endocrinology division (ie, there is no dedicated primary care resource).

Study Population

We focused on all patients aged 18 to 75 years with a diabetes diagnosis who entered the EPCD model from January 1, 2020 to December 31, 2020 and assessed outcomes to August 1, 2022. Patients who declined research authorization; were incarcerated, pregnant, in hospice, or receiving palliative care; permanent nursing home residents; and those who died during the study period were excluded. We categorized the study cohort by race and ethnicity as follows: Asian, Black, Hispanic/Latine, White, other race or ethnicity, unknown or did not disclose.

Enhanced Primary Care Diabetes Model

The EPCD model is a nurse-led primary care–wide initiative with multidisciplinary support, which has been described in detail.11,12 In brief, primary care team RNs are paired with several primary care physicians (PCPs) on their care team. They use an EHR report to identify patients on the PCP’s panel who are not meeting D5 indicators. Nurses actively engage in patient care when they are not meeting any 1 of the D5 indicators. A detailed process algorithm is used to guide RN actions based on the patient’s specific situation, including criteria for scheduling tests and PCP visits, referrals for smoking cessation services, engaging a clinical pharmacist for treatment recommendations, and contacting a social worker and/or relevant community-based organizations for guidance on addressing SDOH. Patients are contacted by the RN via telephone to assess the current state of diabetes management, treatment adherence, and barriers to self-management and attainment of D5 goals. The RNs also engage with their panel of patients longitudinally to support care coordination as related to diabetes and its complications. The PCPs and RNs meet monthly for a duration (generally 30-60 minutes) determined by the number of paneled patients with diabetes, to review panel progress and discuss challenges.

Independent Variable and Covariates

The primary independent variable was self-reported race/ethnicity obtained at the time of registration. Covariates were determined a priori and obtained from the EHR and included patient age, gender, Rural-Urban Continuum Code (RUCC),16 diabetes type, number of glucose-lowering medications, and insulin use. Diabetes type was established using the most recent International Classification of Diseases 10th Revision diagnosis code before program enrollment. Address (to calculate RUCC) was ascertained from the EHR based on self-report at the time of registration. The number of glucose-lowering medications and insulin use were identified from the EHR active medication list within 120 days before program enrollment.

Outcomes

The primary outcome was the time to meeting the all-or-none D5 composite measure (detailed in the Supplemental Appendix). Per MNCM specifications, missing data for a D5 component were considered as not meeting that component.14 The D5 was extracted from an EHR registry maintained for reporting to MNCM. Care team engagement was a secondary outcome, measured by documentation of evaluation for diabetes management using a specific diabetes episode of care in the EHR. Episodes of care allow for longitudinal tracking of chronic disease processes and were used by RNs to implement the EPCD.

Statistical Analysis

Data are summarized using No. (%) for categorical data and mean (SD) or median (interquartile range [IQR]) for continuous data. We assessed associations among baseline patient characteristics including race/ethnicity and meeting D5 criteria using univariate and multivariable Cox proportional hazards regression. The multivariable model included race/ethnicity, age, gender, RUCC group, diabetes type, number of medications in the past 120 days, insulin use in the past 120 days, and number of D5 criteria met at baseline. We estimated values for meeting D5 criteria by race/ethnicity using the Kaplan-Meier method, which accounts for patients having differing lengths of follow-up. Two-sided P values <.05 were considered statistically significant. All analyses were performed using SAS version 9.4 (SAS Institute Inc) and R version 4.2.2 (R Project for Statistical Computing).

RESULTSStudy Cohort

From January 1, 2020 to December 31, 2020, the EPCD program enrolled 1,749 patients, of whom 728 (41.6%) were women, 1,400 (80%) were non-Hispanic White, 1,700 (97.2%) lived in urban areas, and 1,723 (98.5%) were diagnosed with type 2 diabetes (Table 1). Enrolled patients were taking a median of 13 medications, and 560 (32%) were using insulin. Most patients were meeting 4 of the D5 indicators (67%) at enrollment. The least-met indicator at baseline was HbA1c < 8% (45.9%).

Table 1.

Baseline Patient Characteristics and D5 indicatorsa

A total of 1,061 (60.7%) patients met the D5 by the end of the study period. Values for D5 attainment over the study period by race/ethnicity are shown in Figure 1. There was no difference in median follow-up time among groups (median 1.9 years for all groups; P = .31). Black patients enrolled in the EPCD program were significantly less likely to meet the D5 composite measure compared with White patients in the univariate analysis (hazard ratio [HR] 0.60; 95% CI, 0.45-0.79; P < .001) and multivariable analysis (HR 0.68; 95% CI, 0.52-0.90; P = .007) (Table 2). There were no significant differences in the time to meeting the D5 among the other racial and ethnic groups compared with White patients.

Figure 1.

Rates of D5 Attainment by Patients of Different Racial/Ethnic Groups

Table 2.

Factors Associated With Meeting the D5 Care Quality Indicator

The overall median number of nurse touch points per year was 1.2 (IQR 0.4, 2.9). Compared with White patients (median 1.1 per year; IQR 0.4, 2.7), Asian patients had fewer nurse touch points (median 0.8 per year; IQR 0, 1.4; P = .024), whereas Black patients (median 2.2 per year; IQR 0.6, 4.0; P < .001) and patients of unknown race/ethnicity (median 2.2 per year; IQR 0.6, 4.0; P = .002) had more. Nurse touch points were not significantly different for Hispanic/Latine and White patients.

In the multivariable analysis, patients who were older (HR 1.09; 95% CI, 1.03-1.16; P = .005), treated with more total medications (HR 1.02; 95% CI, 1.01-1.03; P < .001), and not treated with insulin (HR 0.85; 95% CI, 0.74-0.98; P = .023) were significantly more likely to meet the D5 composite goal (Table 2). Patients meeting 2 (HR 0.24; 95% CI, 0.16-0.35; P < .001) or 3 (HR 0.39; 95% CI, 0.34-0.46; P < .001) of the D5 indicators at baseline were less likely to attain the D5 compared with those meeting 4 of the D5. The only factor to change significance in the univariate analysis compared with the multivariable analysis was patients with type 2 diabetes being more likely to meet the D5 compared with those with type 1 diabetes (HR 2.79; CI, 1.25-6.21; P = .012).

DISCUSSION

Despite the overall demonstrated effectiveness of the EPCD model in improving diabetes care and outcomes among primary care patients with diabetes,12 the present study identified racial disparities in the likelihood of benefiting from EPCD enrollment. Specifically, Black patients with diabetes were 32% less likely to achieve the D5 (an indicator of high-quality diabetes care) compared with White patients, despite evidence that they were more actively enrolled in the program (ie, had greater documentation that RNs were working on their care).

This study builds on an extensive body of evidence showing pervasive racial disparities in diabetes epidemiology, treatment, and health outcomes. Black individuals in the United States are at increased risk of developing diabetes compared with White individuals,1 are less likely to receive diabetes preventive care17 and self-monitor blood glucose,18 have worse glycemic control,19 experience greater rates of long-term microvascular16 and macrovascular complications,20,21 have increased risk of hospitalization as a result of severe hypoglycemia and hyperglycemic crises,22,23 and have increased mortality.3 Our findings underscore the importance of examining clinical initiatives not only for overall effectiveness in meeting their goals, but also for their equity (a core component of care quality as defined by the National Academy of Medicine).24 Whereas quality measures used for public reporting and value-based reimbursement focus extensively on care effectiveness (ie, overall meeting of the D5), none assess equity.25

We found that Black patients engaged in the EPCD model were significantly less likely to attain the D5 despite being engaged with the care team RNs (ie, actively receiving the intervention) more often than White patients. This was surprising, given that our hypothesis was that D5 attainment would be mediated by active engagement with an RN for diabetes management and care coordination.9 Importantly, whereas we were able to track whether contact had been made, we do not know the depth, breadth, and quality of the interaction. It is possible that EPCD encounters were not tailored for socioeconomic and cultural differences across racial and ethnic subgroups, leading to disparities in outcomes, particularly as many individuals who self-identify as Black in the Mayo Clinic Rochester patient population are refugees from Somalia and other African nations, and the overall population served by Mayo Clinic Rochester is predominantly non-Hispanic White.26 Further research will need to qualitatively examine the quality of nurse, care team, and patient interactions among different racial and ethnic groups to probe for potential reasons for the observed disparities in D5 attainment despite program participation. With SDOH being a potential primary driver of disparities in diabetes outcomes for PGM, community engagement and digital health initiatives are among the many proposed future practice-based research directions to address barriers to equitable care.27,28 It will be important to support implementation of these as part of the EPCD model and evaluate the effect on health outcomes and disparities.

In addition to racial differences in EPCD effectiveness, we identified other factors associated with lower rates of D5 attainment. These include younger age, being treated with fewer glucose-lowering drugs before EPCD enrollment, requiring insulin therapy, and having fewer of the D5 components met at the time of EPCD entry (ie, having larger care gaps). These associations are not surprising and are consistent with prior studies. Younger patients with diabetes—both type 1 and type 2 diabetes—face unique socioeconomic barriers to optimal diabetes care that might be poorly accommodated and met by the health care system.29 Patients requiring insulin therapy are more complex and often require frequent dose adjustments when not meeting glycemic targets; this level of support exceeds what can be offered by the EPCD’s one-time pharmacist e-consult (ie, asynchronous chart review) for medication therapy adjustment.13 Patients with type 1 diabetes and those with type 2 diabetes on multidose insulin regimens would also benefit from referral to endocrinology or longitudinal management by a clinical pharmacist. The EPCD did not include certified diabetes care and education specialists or registered dieticians because those health professionals—whose engagement is vital for optimal diabetes care—were not available in primary care; their absence might have contributed to the inability to optimally support more complex and greater-needs patients with diabetes.

This study has several limitations. Though it was conducted at 13 different practice sites, all are within the same integrated health care system in the US upper Midwest. Therefore, our findings might not fully generalize to more urban areas or to other sociodemographic groups whose SDOH needs and barriers to care might be different than those observed in our study. Given that this was an observational study, our focus was on identifying high-risk patient subgroups and deficiencies in care delivery to be addressed via quality improvement efforts rather than establishing causality. In addition, our evaluation period spanned the COVID-19 pandemic, which likely affected patients’ diabetes management; nurse staffing and bandwidth to focus on chronic disease management; and institutional focus on chronic illness care. This might explain, in part, why the total number of RN contacts within the EPCD was low overall. Moreover, we expect that not all communication encounters might have been captured in our analysis because RNs might not have consistently categorized their encounters as being EPCD related. With a large overall sample size and variable potential nurse actions within the EPCD model workflow, we were unable to categorize nurse touch points to assess differences. Given that the sample sizes for PGM are fewer than those for White patients, CIs are larger and more uncertain for these groups. Most patients who identified as Black were Somali immigrants, which might limit generalizability to the larger population. Lastly, SDOH might be a key driver of the study findings, and we were unable to reliably assess these.

In conclusion, Black patients with diabetes were less likely to benefit from the EPCD model (ie, less likely to achieve the D5 goals) despite more engagement with care team RNs than White patients. Qualitative research is needed to identify driving factors for these disparities and design outreach that better supports the needs of PGM.

Footnotes

Conflicts of interest: In the last 36 months, R.G.M. has received unrelated research support from the National Institute of Diabetes and Digestive and Kidney Diseases, National Institute on Aging, Patient-Centered Outcomes Research Institute, National Center for Advancing Translational Sciences, and the American Diabetes Association. She also serves as a consultant to Emmi/UpToDate Patient Engagement Solutions (Wolters Kluwer) and Yale New Haven Health’s Center for Outcomes Research and Evaluation. The other authors report none.

Read or post commentaries in response to this article.

Funding support: L.R.S. received internal funding from the Mayo Midwest Pharmacy Research Committee for research time and data retrieval services.

Supplemental materials

Received for publication April 30, 2024.Revision received March 3, 2025.Accepted for publication March 10, 2025.© 2025 Annals of Family Medicine, Inc.