Trial of a New Diabetes Education Model: Closing the Gap in Health Disparity for People with Diabetes

Abstract

Background:

Quality measures relating to diabetes care in America have not improved between 2005 and 2016, and have plateaued even in areas that outperform national statistics. New approaches to diabetes care and education are needed and are especially important in reaching populations with significant barriers to optimized care.

Methods:

A pilot quality improvement study was created to optimize diabetes education in a clinic setting with a patient population with significant healthcare barriers. Certified Diabetes Care and Education Specialists (CDCES) were deployed in a team-based model with flexible scheduling and same-day education visits, outside of the traditional framework of diabetes education, specifically targeting practices with underperforming diabetes quality measures, in a clinic setting significantly impacted by social determinants of health.

Results:

A team-based and flexible diabetes education model decreased hemoglobin A1C for individuals participating in the project (and having a second A1C measured) by an average of −2.3%, improved Minnesota Diabetes Quality Measures (D5) for clinicians participating in the project by 5.8%, optimized use of CDCES, and reduced a high visit fail rate for diabetes education.

Conclusions:

Diabetes education provided in a team-based and flexible model may better meet patient needs and improve diabetes care metrics, in settings with a patient population with significant barriers.

Keywords

CDCES diabetes education care model quality improvement diabetes quality measures health care disparities team based

Introduction

Type 2 diabetes currently affects an estimated 10.5% of the U.S. population, and is a significant cause of life-altering complications.¹ Although it is well established that optimization of glycemic management, smoking cessation, and medication-based interventions to control blood pressure and cholesterol can significantly decrease the likelihood of developing diabetic complications,^2-5 a recent review of National Health and Nutrition Examination Survey (NHANES) data suggests that there has been limited progress in improving the quality of diabetes care between 2005 and 2016, with less than 1 in 4 Americans meeting goals for A1C management, blood pressure management, tobacco use, and cholesterol management. Quality of care measure disparities are especially significant for individuals who are younger, female, non-white, or uninsured.⁶

The State of Minnesota has done somewhat better in achieving diabetes quality measure goals. Continued monitoring and public reporting of D5 measures (A1C <8, blood pressure <140/90, statin use, no tobacco use, aspirin if vascular disease) has helped Minnesota significantly exceed national statistics in diabetes quality of care: “Although in 2004 only 11.9% of patients with known diabetes in Minnesota achieved optimal diabetes care . . . this rate increased to 44.7% in 2017—more than twice the estimated national value”.⁶ Despite this, quality measures for diabetes care in Minnesota have largely plateaued, and populations with multiple barriers to care continue to lag behind the general population.⁷

Park Nicollet’s Brookdale Clinic (PNBC), in Brooklyn Center, Minnesota, serves a population with multiple barriers to healthcare. Brooklyn Center is an inner-ring suburb of Minneapolis, in an area of significant demographic diversity. Based on 2019 US Census data, 55% of residents are non-Caucasian, and approximately 24% foreign-born. Thirty-four percent report a language other than English being spoken in the home, 17% report income in the poverty-range, and 11% of individuals under the age of 65 report a lack of health insurance.⁸ Barriers to optimized healthcare, including financial, socioeconomic, language, cultural, citizenship status, and health literacy barriers, have historically made “quality based” care difficult in this community. At the same time, PNBC sees a population disproportionately impacted by type 2 diabetes because of a combination of ethnicity and demographics.⁹

The traditional diabetes education model has been challenging in this environment, with provision of diabetes education significantly impacted by transportation, cultural, language, and health literacy barriers. These barriers have threatened the sustainability of our current diabetes education model due to high visit-fail rates of up to 44.4% (average fail rate 27.6% in 2018). The existing diabetes education model utilized scripted education messaging and visit format largely based on models developed for more homogeneous patient populations with fewer barriers to care. It was recognized that our current model was not adequately meeting this populations’ needs, nor adequately addressing barriers, to improve glycemic management and optimize health.^10,11 We sought to test a new model allowing better integration, and aligning more closely with team-based diabetes care, including better integration into an existing medical home model, closer coordination of care with individual clinicians, and improved ability to use telehealth outreach and diabetes technology to optimize care.

The vision of this project was to create a more agile diabetes education model which allowed real-time support for patients, meeting them “where they are” with their diabetes management, to reduce barriers and improving the safety and efficacy of diabetes management through improved patient engagement and team based care.

Specific aims: Several specific and measurable aims were identified as potential targets for improvement, relative to the established model of care, based on “SMART” (Specific, Measurable, Achievable, Realistic, Time-bound) criteria:

Increase the number of patients touched by this model by 10%, relative to baseline traditional model

Refer 10% of patients who have barriers to care to appropriate resources

Improve the A1C of patients impacted by this intervention by 1%

Increase the number of patients impacted by this project currently achieving the D5 measures (A1C <8, blood pressure <140/90, statin use, no tobacco use, aspirin if vascular disease ) by 5%

Improve ability to leverage diabetes technologies by using continuous glucose monitoring (CGM) in 50% of participants

Methods

Context: PNBC is a primary care-based community clinic located in an inner ring suburb of Minneapolis. Services provided include primary care (Family Medicine 7 clinicians, Internal Medicine 7 clinicians, Pediatrics 6 clinicians), urgent care, optometry and midwife services.

PNBC has had on-site diabetes education services since 2005. Bringing this on-site improved access to diabetes education, especially for individuals with limited transportation options, and significantly improved the number of patients meeting all D5 measures. However, with time, these gains have plateaued, and the sustainability of the diabetes education program at PNBC has become problematic due to the significant visit fail rate. The cause of the high visit fail rate is felt to be multifactorial and related to a combination of barriers: cultural; financial; scheduling; transportation; and inability to take time off work.

Diabetes education at PNBC has used a model which pairs a CDCES RN and RDN in providing centrally scheduled, traditionally structured, diabetes education in a 1-2 h shared time slot. Individuals with the ability to attend and engage in traditional diabetes education experience well-documented benefits both in management self-efficacy and also in glycemic optimization as measured by A1C improvement.¹⁰ The PNBC experience, however, is that the benefits of traditionally structured diabetes education are only accessible to individuals who either don’t have, or are able to navigate, barriers to traditionally structured diabetes education. The goal of this quality improvement initiative was to provide the core components of diabetes self-management education and support as well as medical nutrition therapy in a way that minimized the barriers associated with our traditional scheduling and delivery model for diabetes education.¹⁰

Interventions: No change was made to the amount of scheduled time allocated to Diabetes Education; a CDCES-trained RN and RDN saw patients 7 hours per day, 2 days per week. The organization-wide central scheduling model was replaced with local scheduling by the CDCESs, to allow visits, and scheduling of appointments, at the same time of primary care clinician visits. Grant funding through the Park Nicollet Foundation allowed patients to be seen for the duration of this project without co-pay or billing, removing potential financial barriers.

Because of concerns regarding patient volume and CDCES capacity in the setting of a multiple clinicians, the CDCESs initially focused on the patient cohort of 3 clinicians newer to the practice, whose D5 measures lagged behind more established clinicians in the practice. For these clinicians, CDCESs provided team-based diabetes education assistance in titration of medications, as well as registry based engagement of individuals not meeting D5 measures. As the project progressed, 2 additional clinician practices were included in the pilot. CDCESs reviewed the schedules of project clinicians prior to the day of the visit, to identify individuals who might benefit from further optimization of diabetes management, and were available on a just-in-time basis, engaging patients while they were in clinic, to establish a relationship with them.

The CDCESs employed a mobile cart with supplies and educational materials and worked out of an exam room during visits, instead of the dedicated CDCES rooms, located in a distant hallway. Additionally, the CDCES team coordinated with the primary care RN care coordinators (two), a clinical pharmacist (available on a part-time basis), and a clinical social worker. Interactions with clinicians involved in the project, as well as other primary care clinicians, were encouraged and frequent, and were facilitated by the use of secure real-time messaging.

The flexibility and open-endedness of the scheduling model allowed for frequent diabetes education touch points, either in person, via secure EMR based e-mail, or by phone, for education and medication titration. Working in conjunction with clinicians, and with protocol-based insulin titration guidelines, the CDCESs were able to titrate insulin-based therapies more frequently than would be typical in a primary care setting, where titration often lags because of the infrequent cadence of visits. The CDCESs were able to work directly with clinicians, often in real-time, to engage both clinician and patient in shared decision making regarding advancement of therapies and other medication changes.

Additionally, remote data upload and CGM-based monitoring were explored for individuals as appropriate. For some, this allowed an additional mode of remote data acquisition to facilitate titration of diabetes therapies.

Study of the interventions: Monthly review of the intervention and data by the project team allowed modification of process based on “PDCA” (Plan, Do, Check, Act) cycles. Patient volumes, as well as quality measures (based on registry data) were reviewed at these monthly meetings.

Measures: Change in A1C for project participants were tracked using established EMR-based registry. This project also tracked D5 data for clinician and clinic patient panels, which includes percent of patient population meeting the combined goal of A1C <8.0, blood pressure <140/90, use of statin medication, non-use of tobacco products, and aspirin use, if established vascular disease. This information is a publicly reported measure in the State of Minnesota. These measures were tracked for pilot clinicians and for the clinic as a whole over the course of the project.

Analysis: Data were gathered with spreadsheet-based tracking, as well as EMR-based registry data maintained by our organization to track diabetes care quality. Data comparison was performed based on these data sources.

Ethical considerations: This work was not subject to formal ethics review. Conflicts of interest are declared, but none influenced the conduct of this project. All standard HIPAA guidelines were followed, and patients were not billed for services during the duration of this project, as it was grant funded.

Results

Results: A total of 157 appointments, excluding phone calls, were conducted: 57 unique patients were seen by CDCESs over the course of 6 months for diabetes education visits, and an additional 44 consults were done ad hoc with the primary care clinician. The demographic profile of these 57 individuals is shown in Table 1.

Table 1.

Population Characteristics. Patient Characteristics (n = 57).

Characteristic
Age (years)	Avg.	52.6 (SD = 12.6, range 27-91)
Sex		Male = 29 (51%)
Sex		Female = 28 (49%)
Race/ethnicity	Caucasian (non-Hispanic)	10 (17%)
	AA or African	27 (47%)
	Hispanic/Latino	11 (20%)
	Asian	9 (16%)
Baseline A1C	Avg.	9.8% (SD = 2.7, range 6.5-19%)
Baseline CGM use	None
CGM use during project	Professional CGM	5 (9%)
	Personal CGM	8 (14%)
	Total CGM	13 (23%)
Interpreter use	English language barrier	14 (25%)

Twenty-nine of 57 patients had updated A1C lab results at the time of follow-up; outcome data for these 29 participants is shown in Table 2. Of these 29 individuals, 67% had a greater than 1% improvement in A1C. 48% had a greater than 2% improvement, and 30% had a greater than 3% improvement. Only 7% of updated A1C results did not improve.

Table 2.

Percentage of Individuals in Practice Meeting All 5 “Minnesota D5” Diabetes Quality Measures. (A1C <8.0, BP <140/90, Statin Therapy if Vascular Disease, Aspirin Use if Vascular Disease, No Tobacco Use).³

	Clinician (QI participant “n”)	QI participants at A1C goal post-intervention	QI participants mean reduction in A1C post-intervention	Total clinician panel at D5 goal pre²	Total clinician panel at D5 goal post	Improvement in total clinician panel D5 “% at goal”
Initial clinician Cohort¹	Clinician A (QI participants = 6)	33.3% (2/6)	−1.7 (range −4.0 to 0.1)	38.7% (n = 24/62)	40.6% (n = 28/69)	1.9%
	Clinician B (QI Participants = 5)	60.0% (3/5)	−2.2 (range −4.7 to −0.3)	33.3% (n = 7/21)	40.7% (n = 11/27)	7.4%
	Clinician C (QI participants = 13)	61.5% (8/13)	−2.3 (range −8.3 to 0.6)	27.3% (n = 24/88)	39.0% (n = 41/105)	11.7%
Initial 3 clinicians	(QI participants = 24)	54.2% (13/24)	−2.1 (range −8.3 to 0.1)	32.2% (n = 55/171)	39.8% (n = 80/201)	7.6%
Clinicians added at Month 3	Clinician D (QI participants = 1)	0% (0/1)	+1.0	43.0% (n = 34/79)	46.4% (n = 39/84)	3.4%
	Clinician E (QI participants = 4)	75.0% (3/4)	−4.2 (range −9.0 to −2.0)	43.8% (n = 35/80)	48.2% (n = 40/83)	4.4%
Composite 5 clinicians	(QI participants = 29)	55.2% (16/29)	−2.3 (range −9.0 to 1.0)	37.4% (32.1-42.9)^a (n = 120/321)	43.2% (38.1-48.4)^a (n = 159/368)	5.8%
Adult primary care clinic 6 month trend	Composite 14 Clinicians, (n = 2221)			47.8% (n = 1025/2144)	49.6% (n = 1101/2221)	1.8%

Exact confidence interval.

QI Participant number for individual clinicians represents individuals seen by CDCES and having availability of a 2nd hemoglobin A1c (for project cohort this includes 29/57 individuals, or 51% of individuals followed in the study).

%D5 “At Goal” represents the number of patients in that clinicians total pool of individuals with diabetes who achieve %D5 “At Goal”.

% Change in Minnesota D5 Quality Measures over the 6 month period of the project. The Minnesota D5 quality measures are publically reported measures of the quality of diabetes care, measured for individual clinicians, but reported to the State of Minnesota annually as a clinic-based composite measure. The Minnesota D5 Quality Measure represents the percentage of individuals meeting the composite measure of A1C <8.0, BP <140/90, statin therapy if vascular disease, aspirin use if vascular disease, and no tobacco use.

A1C Improvement: The average A1C for the 29 patients with updated A1C results improved by 2.3%.

Initially, work was directed toward 3 clinicians in the primary care practice who had the lowest percent of A1Cs at goal for their total population with diabetes (newest in the practice, and with less well established patient cohorts). In 3 months, their collective A1C% at goal (<8.0% for their total panel with diabetes) had improved by 7.6%; this percentage included new patients added to practice who were not at goal. One provider increased his percent of patients with A1C at goal by 11.3% over the first 3 months of the project.

After the initial 3 months it was determined that the educators had capacity to add 2 additional providers to the project. Over the next 3 months these 2 additional clinicians saw their diabetes population A1C% at goal rate improve by 6.8%.

Minnesota D5 Measures improvement: D5 is reported as an “all or none” quality measure. All 5 of the D5 quality metrics must be a goal for a patient to be scored as having reached the D5 composite target. The D5 scores for each clinic and health care organization in the Minnesota are publically report on the Minnesota Community Measures website.¹² D5 improvement from 7/4/2019 to 12/31/2019 for the initial 3 providers (total number of individuals in practices with diabetes = 201) was 32.2% at goal in July, to 39.8% at goal at the end of December, an increase of 7.6% in patients meeting all 5 of their diabetes goals (Minnesota D5). One clinician increased his percentage of patients (total number of individuals in practice with diabetes = 105) meeting all 5 goals by 11.7% over this time period.

With the addition of 2 providers (aggregate of individuals with diabetes for all 5 clinicians = 368) D5 improved by 5.6% (37.6% in July 2019 to 43.2% in December 2019), despite the addition of 120 patients to the practices whose D5 measures were not meeting D5 goals.

For the adult primary care group as a whole (14 clinicians), percent attaining all 5 diabetes quality measures (n = 2,221) increased by 1.8% from 47.7% to 49.6% over the 6 months of the project.

Improved access to diabetes education: Diabetes Education visit fail rate decreased from 30.1% in the second half of 2018, to 19.9% in the 2nd half of 2019 (10.2% improvement).

Over the course of this pilot, 23% of patients seen used CGM and had CGM data reviewed with educators (9% professional and 14% personal CGM). Cost and the requirement to wear the sensor were the 2 major limitations to CGM use.

Although more subjective, because the RN and RDN CDCESs had a consistent physical proximity to primary care providers and RN Care Coordinators, they were consulted frequently regarding the management of mutual patients. The presence and accessibility of CDCESs in the primary care setting fostered collaborative decision making and helped to move diabetes therapy along.

During the timeframe of this project, the CDCESs tracked these more informal “curbside consults”. Reasons clinic staff consulted CDCESs ranged from questions regarding healthy eating resources, access to medications for those unable to afford medications, insulin dosing, and options for non-insulin therapy. Many of these issues were best addressed in a collaborative setting with team members (care coordinators, social work, and the clinician staff), and this care model facilitated these collaborative interactions very well.

Discussion

Nationally, gains made in improving the quality of our diabetes care as measured by metrics for A1C at goal, blood pressure at goal, statin use, and nonsmoking, have plateaued, and did not improve between 2005 and 2016. The A1C measure has actually worsened during that period of time. Gaps in care, especially for individuals who are younger, female, non-white, or uninsured, existed in 2005, and have persisted.⁶

The State of Minnesota has done somewhat better than the national average in this regard but improvement in composite measures have plateaued and, despite significant advances in diabetes therapy, further progress has stalled.^6,7

This project sought to explore alternative models of diabetes education delivery beyond the traditional diabetes education model. This model was team-based, more interactive and longitudinally oriented, and allowed CDCESs autonomy to engage patients outside the traditional diabetes education model. The goal was to test a model that would provide individualization to overcome significant care barriers related to social determinants of health. While limited in scope, the hope was to create a model which was more agile in addressing barriers to optimized diabetes care in a clinic with significant demographic challenges to optimization of diabetes care. The model evaluated in this pilot allowed CDCESs to better address barriers to care and improve cycle time in advancing and titrating therapies, working in conjunction with clinicians and medical home team members.

Interpretation: Moving from the traditional diabetes education model to a more agile and team-based model improved diabetes quality measures for a subset of PNBC clinicians who struggled with diabetes quality measures and also improved the failure-to-attend rate, thus optimizing the benefits of onsite diabetes education. In the setting of diverse demographics and significant barriers to optimized diabetes care, this model improved quality measures for the individuals impacted, as well as clinicians involved and the clinic as a whole, largely by addressing barriers which had limited ability to participate in traditional diabetes education. Increased longitudinal connection with CDCESs also allowed increased opportunity for medication optimization and titration by improving titration “cycle time” compared to the typical 3 month cadence of clinician visits for individuals not meeting diabetes goals.

Finally, a team-based approach, with improved proximity of CDCESs to clinicians, working out of the same hallway, encouraged collaboration, interaction, and “curbside” access in the sharing of ideas, possibly accounting for “halo effect” extending to clinicians outside the parameters of the project.

While this was a limited pilot study, we hope that learnings from this project can be used to create a new or hybrid diabetes education model which better serves diverse populations, and allows us to move past the current inertia in improving diabetes quality measures.

Limitations: This was a quality improvement project, meant to be descriptive, and was not designed or conducted with the intent to calculate statistical significance. Because this project was conducted in an open-ended and uncontrolled environment in a real world primary care practice, a significant number of individuals seen by CDCESs did not have the availability of a second A1C value. This likely biased the degree of improvement in A1C seen, although it is likely that individuals without a second A1C benefited also, based on the project cohort being seen on average 2.9 times by educators. Potential bias from confounders like time trends was not evaluated. This project was limited in scope, involving 5 clinicians and the patient cohorts associated with those practices. Diabetes education availability was limited to 2 days per week, based on staffing parameters prior to the project. Diabetes education costs were grant-funded for the duration of the project which allowed provision of diabetes education without cost to the patient, and coverage of the salaries of the CDCESs. Financial viability, therefore, was not evaluated as part of this project. It is hoped that as health systems and the American medical system evolve towards a pay-for-performance model of reimbursement, the financial viability of the model will be justified by improvement in diabetes care quality, with a vision towards reducing the total cost of care, and the total burden of complications of diabetes, by optimizing upfront care to reduce complications in vulnerable populations.

Conclusion

Conclusions: Moving from the traditional scripted diabetes education model, to a more team-based and agile model, allowed optimized provision of diabetes education for individuals with significant barriers to traditional diabetes education, and improved quality measures, in a setting in which traditional diabetes education models have not had optimal success.

Footnotes

Acknowledgements

The authors wish to acknowledge the significant contributions of clinicians of the Family Medicine Department, the Care Coordination staff, and clinic management at Park Nicollet’s Brookdale Office, for their contributions to this project, and dedication to improving the care of the community they serve through progressive action. They also thank the Park Nicollet Foundation for generous support of this project, and for their ongoing commitment to addressing community needs and overcoming barriers in improving the health of the community.

Abbreviations

A1C, hemoglobin A1c; CDCES, Certified Diabetes Care and Education Specialists; CGM, continuous glucose monitoring; D5, diabetes 5 parameters, a Minnesota diabetes quality measure; EMR, electronic medical record; HIPAA, health information portability and accountability act; NHANES, National Health and Nutrition Examination Survey; PNBC, Park Nicollet’s Brookdale Clinic; RDN, registered dietitian nutritionist; RN, registered nurse; T2D, type 2 diabetes.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: RMB has received research support, consulted, or has been on a scientific advisory board for Abbott Diabetes Care, Ascensia, CeQur Corporation, DexCom, Hygieia, Insulet, Johnson & Johnson, Lilly, Medtronic, Novo Nordisk, Onduo, Roche, Sanofi and United Healthcare. His technology research is funded in part by NIH/NIDDK. RMB’s employer, non-profit HealthPartners Institute, contracts for his services and no personal income goes to RMB. TWM has received research and speaking support from Abbott Diabetes Care, Dexcom, Medtronic, Insulet, Lilly, and Novo Nordisk. TWM’s employer, non-profit HealthPartners Institute, contracts for his services and no personal income goes to TWM. ALC receive research support and/or provide consultation for Medtronic, Abbott, Sanofi, Dexcom, Insulet, Eli Lilly, Novo Nordisk, and UnitedHealthcare. ALC’s employer, non-profit HealthPartners Institute, contracts for his services and no personal income goes to ALC. JSL, EAJ, JAC, JJH, and JAD do not report conflicts of interest.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this project was provided by the Park Nicollet Foundation, a nonprofit whose mission is to “Bring enhanced care to Park Nicollet clinics, specialty centers and Methodist Hospital and partner with schools and area nonprofits to address unmet community needs and help people overcome barriers to care”.

ORCID iDs

Thomas W. Martens

Richard M. Bergenstal

Anders L. Carlson

References

CDC.gov [internet] Centers for Disease Control and Prevention. National Diabetes Statistics Report, 2020. Atlanta, GA: Centers for Disease Control and Prevention, U.S. Dept of Health and Human Services. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf. Accessed April 6, 2020.

Gaede

Lund-Andersen

Parving

, et al. Effect of a multifactorial intervention on mortality in type 2 diabetes. N Engl J Med. 2008;358(6):580-591.

Buse

Ginsberg

Bakris

, et al. American Heart Association; American Diabetes Association. Primary prevention of cardiovascular diseases in people with diabetes mellitus: A scientific statement from the American Heart Association and the American Diabetes Association. Diabetes Care. 2007;30(1):162-172.

Nathan DM; DCCT/EDIC Research Group. The diabetes control and complications trial/epidemiology of diabetes interventions and complications study at 30 years: overview. Diabetes Care. 2014;37(1):9-16.

UK Prospective Diabetes Study (UKPDS) Group. Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet. 1998;352(9131):837-853.

Kazemian

Shebl

McCann

, et al. Evaluation of the cascade of diabetes care in the United States, 2005-2016. JAMA Intern Med. 2019;179(10):1376-1385.

MN Community Measurement [internet]. MN health scores, from MN community measures. Annual reports. https://mncm.org/reports/#mnhealthscores. Accessed February 14, 2021.

Census.gov [internet]. United States Census Bureau. https://www.census.gov/quickfacts/brooklyncentercityminnesota. Accessed April 3, 2020.

Spanakis

Golden

SH.

Race/ethnic difference in diabetes and diabetic complications. Curr Diab Rep. 2013;13(6):814-823.

10.

Powers

Bardsley

Cypress

, et al. Diabetes self-management education and support in adults with type 2 diabetes: a consensus report of the American Diabetes Association, the Association of Diabetes Care & Education Specialists, the Academy of Nutrition and Dietetics, the American Academy of Family Physicians, the American Academy of PAs, the American Association of Nurse Practitioners, and the American Pharmacists Association. Diabetes Care. 2020;43(7):1636-1649.

11.

Fan

Sidani

Factors influencing preferences of adults with type 2 diabetes for diabetes self-management education interventions. Can J Diabetes. 2018;42(6):645-651.

12.

MNHealthScores [internet]. Minnesota community measures website. http://www.mnhealthscores.org/diabetes-13184