The Diabetes Care Rewards program offers a business case for health plans to promote engagement through use of contingent incentives, thus improving health outcomes and lowering costs.
Objectives: To (1) examine the impact of the Diabetes Care Rewards (DCR) program on adherence to care standards and (2) evaluate the economic impact of adherence to care standards.
Study Design: A retrospective observational cohort study design with propensity matching. Additional covariates adjustment was used to minimize residual imbalance.
Methods: Utilization and cost data were compared between individuals enrolled vs individuals eligible for but not enrolled in the DCR program using a standard mean difference. Individuals were employees or their dependents from self-insured companies throughout the United States. Outcomes included adherence to the care standards, service utilization, and costs.
Results: A total of 3318 propensity-matched participants were included. Primary analysis revealed that enrolled members increased adherence to semiannual glycated hemoglobin, annual lipid, and annual urine albumin-creatinine ratio testing. Additionally, enrolled members experienced less utilization of high-acuity services and increased rates of physician visits. In a secondary analysis, the enrolled group was associated with greater pharmaceutical costs.
Conclusions: A behavioral science– and incentive-based diabetes management program was associated with greater rates of adherence to recommended diabetes monitoring care standards, increased routine clinic visits, decreased hospital admissions, and decreased inpatient days. Results indicate the benefits of adherence to evidence-based standards for diabetes care.
Am J Manag Care. 2021;27(3):96-102. https://doi.org/10.37765/ajmc.2021.88597
This study evaluated a commercially available diabetes care management program for members of self-insured health plans across the United States.
Diabetes has a major adverse impact on productivity, disability, and health care costs.1-4 The American Diabetes Association (ADA) recently updated estimates of the economic burden of diagnosed diabetes, reporting a total estimated cost in 2017 of $327 billion, including $237 billion in direct medical costs and $90 billion in reduced productivity.5 For employers, the indirect costs of diagnosed diabetes include increased absenteeism ($3.3 billion) and reduced productivity while at work ($26.9 billion). Much of the medical cost of diabetes is related to comorbidities and complications arising from inadequate management of the disease. To help mitigate the adverse consequences of poor control of diabetes, the ADA provides evidence-based standards of care for persons with diabetes.6 Unfortunately, despite availability of these care standards and expanded therapeutic options, most people with diabetes demonstrate gaps in clinical care, low adherence to glucose monitoring, and inadequate management of cardiovascular risk factors.7-9 The proportion of patients meeting standards for diabetes care, such as glycated hemoglobin (A1C), blood pressure (BP), and low-density lipoprotein cholesterol levels, has not improved significantly between 2005 and 2016.10 Only 1 in 6 individuals with diabetes in the United States is achieving concomitant goals for A1C, BP, and cholesterol, as well as avoiding tobacco use.11
Individuals with diabetes who are actively engaged in the management of their condition have fewer and less serious adverse health outcomes and avoid unnecessary hospitalizations compared with those who are not as engaged.12,13 Preventing unnecessary hospitalizations can significantly reduce overall health care spending.14,15 One promising approach harnesses the evidence base from behavioral science and behavioral medicine research.16,17 The use of structured incentives to engage individuals in health behaviors is an evidence-based behavioral science approach. Successful incentive-based interventions require careful attention to design and implementation.18 Incentives must be contingent, timely, and sufficiently large to both engage individuals and sustain the target health behaviors over time. This study examines the impact of the Diabetes Care Rewards (DCR) program, a behavioral science– and incentive-based care management program designed to increase patient engagement in the management of their diabetes by rewarding timely completion of evidence-based standards of care.19 The DCR program was developed to address the increasing burden of unmanaged diabetes on employee health and productivity and to reduce costs to self-insured employers and their covered employees and dependents. The program was first implemented in 2007 for a financial services company with offices located across the United States. The program utilizes proven principles of behavioral science, is provider-centric, and incorporates ADA standards for diabetes education and self-management support using a structured diabetes health action plan (DHAP) care guide. It was hypothesized that the program would increase the proportion of persons with diabetes completing evidence-based diabetes screening evaluations in addition to reducing total medical costs and overall hospitalization rates.
Design and Setting
We conducted a retrospective observational cohort study to compare utilization and cost data between individuals enrolled vs individuals eligible, but not enrolled, in the DCR program. Individuals were employees or their dependents (ie, members) from 26 self-insured companies throughout the United States. Self-insured companies cover the medical and pharmaceutical expenses of their employees and their beneficiaries, rather than outsourcing these costs to a third-party health insurance provider. The 26 companies were clients of Abacus Health Solutions that participated in the co-pay waiver incentive of the DCR program.
The enrolled cohort was composed of members of any age or gender enrolled in the DCR program. Eligibility for enrollment required a diagnosis of diabetes, using claims-based International Classification of Diseases, Ninth Revision (ICD-9) or Tenth Revision (ICD-10) diagnosis codes, and presence of a 24-month block of continuous enrollment with at least 1 medical or pharmacy claim during that period. This 24-month block included 12 months of claims data prior to DCR program enrollment and 12 months of claims data following enrollment. The unenrolled group was composed of individuals with a diagnosis of diabetes, using claims-based ICD-9 or ICD-10 diagnosis codes or by identifying individuals taking 1 or more glucose-lowering medications, who were eligible for the DCR program but who did not enroll (see eAppendix Table 1 [eAppendix available at ajmc.com]). Members in the control group were also required to have 24 months of continuous claims data. Both groups were engaged using multiple efforts in the forms of mailings to the home for employees and covered dependents, and employees in both groups also received employer postal mail, employer email, and exposure to workplace posters and materials promoting program engagement.
The DCR program provides incentives in the form of pharmacy co-pay waivers to cover out-of-pocket expenses for diabetes medications and supplies. Pharmacy co-pay waivers were contingent on members having an active relationship with their primary care physician or endocrinologist and documenting completion of recommended diabetes care processes: (1) semiannual testing of A1C, (2) annual lipid panel, (3) annual urine albumin-creatinine ratio (ACR), (4) annual eye exam, and (5) annual foot exam. The DCR program utilizes a patented system grounded in behavioral science for contingent activation and deactivation of incentives. Through integration with each member’s pharmacy benefit, co-payments for diabetes medications and supplies were waived at the point of sale if the member was actively adherent to all program criteria; otherwise, co-payments were required. Pharmacies were provided a data feed of information regarding each participant’s adherence status, which was updated daily.
To receive the co-pay waiver incentive, members were required to complete an annual 30- to 45-minute telephonic assessment by a trained certified diabetes nurse educator or clinical pharmacist to develop a written DHAP that is shared with and signed off by their diabetes treatment provider. Using motivational interviewing, these trained clinical staff completed the DHAP as a structured interview using guidelines for diabetes self-management education and support as promoted by the ADA.20 The DHAP addresses core aspects of diabetes self-management, identifies barriers and challenges to health-related goals, and provides training on stimulus-control techniques. Telephonic follow-up was scheduled for 3 months but adjusted by the nurse or pharmacist according to participant needs.
The exposure was defined as enrolled vs not enrolled. This exposure was self-selected by the member and possibly confounded by variables such as demographics and comorbidities. Covariates included in the propensity score matching model were calendar time, age, region, sex, comorbidities (preprogram insulin use status and Elixhauser Comorbidity Index score21), and preprogram total medical costs capped at $200,000 parameterized as quartiles.15 All costs were adjusted to 2018 US$ using the Consumer Price Index inflation calculator for health care services from the US Department of Labor.
To adjust for these baseline (initial 12-month, pre–index date) possible confounders, control members were matched by propensity scores with match tolerance set at 0.10 (0, exact match; 1.0, any control would match any intervention member) using a nearest neighbor approach, and analysis was conducted on the resultant matched sample. Enrolled members were assigned to 2-year time blocks based on the calendar year at the onset of their 24-month claims interval. Unenrolled members were allocated to the earliest 2-year time block for which they had 24 months of continuous claims data and underwent 1:1 propensity score matching with the enrolled members. Those who remained unmatched and unenrolled after propensity score matching and had 24 months of continuous claims data in the next 2-year time block were carried forward into the next 2-year time block. The process of carrying forward those unmatched and unenrolled members was applied to the 2012-2013, 2014-2015, and 2016-2017-time blocks (Figure and eAppendix Figure). Those who were unmatched and unenrolled after propensity score matching in the 2016-2017 time block were excluded from the study (n = 8890); however, no members were excluded due to nonoverlapping propensity scores between the 2 groups.
For the intervention group, the index date was the date of enrollment into the DCR program. The outcomes were ascertained during the 12 months following the index date and baseline information was ascertained from the 12 months preceding the index date. For the control group, each member had the same index date for data inclusion as their matched intervention group member. The outcomes of interest and for which data were available were (1) receipt of medical care (semi-annual A1C, annual lipid, and annual urine microalbumin test), (2) service utilization (hospital admission [yes vs no; counts], hospital days [counts, truncated to 60 days], and outpatient office visits [yes vs no; counts]), and (3) costs (pharmacy and medical [pharmacy and medical measured as per member per month, capped at $200,000 for medical costs]). Capping medical costs is a conventional approach used by health plans to attenuate the impact of catastrophic claim costs and has been used in recent studies of diabetes.15
Characteristics between the intervention and control groups were compared and differences between the groups were assessed using a standardized mean difference (SMD). Analyses were further adjusted for covariates that had an |SMD| greater than 0.10 to adjust for residual imbalances in the matched sample.22,23 In the analysis for each of the primary outcomes—namely, medical care, services and costs—we employed outcome models in the matched sample (ie, adjusted for some covariates included in the propensity score) to allow for further adjustment of possible residual confounding by total preenrollment costs and region.23,24 Pharmacy and medical costs for the year were divided by 12 to obtain the per-member per-month costs. To allow the model to be more flexible, we included total preprogram cost as a continuous variable with quadratic splines and centered this variable on its mean.25 We presented the model results for the receipt of medical care and service outcomes adjusted for preenrollment total costs and region, as well as those for cost outcomes adjusted for the preprogram total costs, region, and the estimated propensity score. For comparison, we also presented the matched, unadjusted analyses. Results that further adjusted for other imbalances in preprogram variables were comparable (see eAppendix). Binary outcomes were modeled using a log-binomial or logistic regression, count outcomes were modeled using negative binomial regression or a zero-inflated Poisson or negative binomial regression, and costs were modeled using gamma regression. When log-binomial models did not converge, log-Poisson models were used to estimate risk ratios.26 Analyses were performed using SPSS version 25 (IBM) and R version 4.0.5 with RStudio version 1.2.5033 (R Project for Statistical Computing).
A total of 14,245 members with diabetes were eligible for enrollment into the DCR program between August 1, 2010, through December 27, 2017. Of this initial pool, 13,948 had 24 months of continuous claims data. The Figure shows the STROBE diagram of the construction of the cohort. After the process of 1:1 propensity score matching, 3318 members were matched: 1659 in the enrolled cohort and 1659 in the unenrolled cohort. Descriptive statistics regarding the preintervention cohort are provided in Table 1. Members were mostly male, middle-aged, from the Northeast, and not prescribed insulin. When comparing baseline characteristics in the matched sample using an SMD, there were differences by region (SMD, 0.17) and preintervention total cost (SMD, 0.36). The enrolled group was more likely to be from the Midwest (7.4% vs 5.8%) and West (9.2% vs 5.2%) compared with the unenrolled group. The enrolled group had slightly higher preenrollment total costs (mean [SD] = $16,137.46 [$26,496.79] vs $12,830.95 [$28,087.70]).
Table 2 displays the estimated effects of enrollment in the DCR program on each of the medical care and service outcomes in the matched sample, both unadjusted and adjusted for preenrollment total costs. After adjustment for preenrollment total costs (with a spline) and region, enrolled members had an increased likelihood of adherence to the semiannual A1C test (risk ratio [RiR], 1.55; 95% CI, 1.45-1.67), annual lipid test (RiR, 1.20; 95% CI, 1.14-1.26), and annual urine microalbumin test (RiR, 1.54; 95% CI, 1.42-1.66). Enrolled members had a lower estimated risk of hospital admissions (RiR, 0.72; 95% CI, 0.59-0.89), lower rate of hospital admits per calendar year (rate ratio [RaR], 0.69; 95% CI, 0.54-0.89), and lower rate of inpatient days per calendar year (RaR, 0.76; 95% CI, 0.50-1.16). Enrolled members also had a 16% increase in the rate of physician visits per calendar year (RaR, 1.16; 95% CI, 1.06-1.27). Results for the medical care and service outcomes were comparable after further adjustment in the outcome model for preenrollment total costs, age, and Elixhauser Comorbidity Index score (see eAppendix Table 1).
The unadjusted and adjusted estimated effects of enrollment in the DCR program on cost outcomes are presented in Table 3. Adjusting for the propensity score, region, and preenrollment total cost (spline), the enrolled group demonstrated greater pharmaceutical costs (per-member per-month cost difference of $62.12; 95% CI, $38.77-$85.47). Results for the cost outcomes were comparable after adjustment for age and Elixhauser Comorbidity Index score (see eAppendix Tables 2 and 3).
In this real-world study, individuals with diabetes who were enrolled in an incentive-based care management program demonstrated increased adherence to recommended diabetes care standards, increased provider outpatient visits, decreased hospital admissions, and decreased inpatient hospital days. The program was associated with greater pharmaceutical expenditures. These findings confirm the hypothesis that a behavioral science– and incentive-based approach to diabetes, incorporating contingent and meaningful rewards to promote engagement with diabetes care teams, can improve adherence to diabetes care while significantly reducing the use of high-acuity health services.27
The DCR program was designed and evaluated specifically with health plans and employers—especially those that are self-insured, for which both productivity and cost benefits are priorities—and provides a health plan–based incentive for initial and continued engagement in completing ADA care standards. Members are provided an enhanced plan benefit (ie, co-pay reductions) for being actively engaged in the program, with this incentive contingent upon the member’s adherence to the standards of diabetes care, which could be completed only through collaboration between the member and their health care provider(s). With respect to improving recommended diabetes care standards, participation in the DCR was associated with greater adherence to A1C and urine ACR testing compared with lipid panel testing. This finding likely reflects the additional requirement for 12-hour fasting and venous blood draw for a lipid exam as opposed to obtaining an A1C level, which does not require fasting and can be obtained by point-of-care testing.
Regarding provider engagement, those members enrolled in the DCR program experienced a 16% increase in the number of primary care provider visits, suggesting more active and effective management of their diabetes, resulting in less need for hospital-based services. This is reflected by the reduction in hospital admissions and total inpatient days. These results are consistent with recently published findings by Zhang et al that showed a positive association between the frequency of contact with a patient’s primary care provider and both a decrease in A1C level and lower rates of 10-year occurrence of cardiovascular events.28
Being enrolled in the DCR program was associated with greater medication use and pharmaceutical expenses. Interestingly, the estimated cost difference indicated a higher pharmacy cost among those enrolled in both unadjusted and adjusted analyses. We anticipated an increase in pharmacy costs, as the DCR program was aimed at increasing engagement in care. Although pharmacy costs were higher for enrolled members than for unenrolled members, this study was unable to capture pharmacy rebates from manufacturers and pharmacy benefit managers to self-insurers. The cost-benefit analysis of this program weighs pharmaceutical expenses against the costs of hospitalizations and other high-acuity services. Rebates decrease the final costs of pharmacy expenditures and, if available, could result in a favorable cost-benefit ratio of the DCR program.
Strengths and Limitations
Strengths of this study include matching enrolled members to eligible unenrolled members using propensity score matching.22,23 Matching on patient characteristics such as age, sex, Elixhauser Comorbidity Index score, geographic region, and insulin use offers some protection against comparing heterogeneous cohorts by ensuring that both study cohorts contained members with certain similar baseline characteristics. Additional strengths include a novel but practical and feasible incentive structure that scaled across multiple diverse self-insured employers.
As an observational study that utilized administrative claims data, this analysis is subject to inherent limitations, including selection bias. Members who enrolled are likely different from those who did not enroll, which is unavoidable in retrospective studies of voluntary programs. To address this limitation, members were propensity score matched in the analysis, which is a valid approach to control for measured confounders, including insulin dependence, to compare members whose disease state has or has not advanced to the point of requiring insulin therapy. When adjusting for covariates in an outcome model to control for residual imbalance between the groups, the outcome model needs to be correctly specified. To address this concern, we flexibly modeled covariate functional forms and conducted sensitivity analyses for covariates included in the outcome model. Because this study was reliant on administrative claims data alone, comparisons between clinically reported and laboratory values of enrolled vs unenrolled members could not be made. Additionally, identifying those with diabetes could be done only by using ICD-9 and ICD-10 codes (see eAppendix) or by identifying individuals taking 1 or more glucose-lowering medications.
By creating behavioral activation with a contingent financial incentive and promoting ongoing participant-provider interactions, the DCR program reported here reduced rates of gaps in care and high-cost health care utilization from the perspective of health plans and self-insured employers. Consequently, a business case exists for health plans to promote patient engagement through use of contingent incentives, with the goal of leading to better health outcomes and lower rates of hospitalizations.
Author Affiliations: Abacus Health Solutions (DKA, EWA, JPW, QZ, MJF), Cranston, RI; Digital Behavioral Health and Informatics Research Program, Department of Psychiatry, Brigham & Women’s Hospital (DKA), Boston, MA; Harvard Medical School (DKA, SNM), Boston, MA; College of Pharmacy, University of Rhode Island (DKA, ALB, KKS, RJR), Kingston, RI; Joslin Diabetes Center (SNM), Boston, MA.
Source of Funding: Funding was provided by Abacus Health Solutions, LLC. Dr Mehta was supported by National Institutes of Health grant No. P30DK036836.
Author Disclosures: Dr Ahern, Dr Aberger, and Mr Wroblewski are employees of and minority shareholders in Abacus Health Solutions, LLC; the intervention described in this manuscript is a commercially available product of Abacus Health Solutions. Mr Zheng is an employee of Abacus Health Solutions. Dr Buchanan performs statistical consulting for Abacus Health Solutions and has received grants for unrelated research. Dr Follick is an employee of and majority shareholder in Abacus Health Solutions. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (DKA, EWA, JPW, SNM, KKS, MJF); acquisition of data (DKA, EWA, JPW, QZ, MJF); analysis and interpretation of data (DKA, EWA, JPW, QZ, SNM, ALB, KKS, RJR, MJF); drafting of the manuscript (DKA, EWA, SNM, ALB, KKS, RJR, MJF); critical revision of the manuscript for important intellectual content (DKA, EWA, SNM, RJR, MJF); statistical analysis (DKA, ALB); provision of patients or study materials (MJF); administrative, technical, or logistic support (DKA, JPW, QZ, MJF); and supervision (DKA, MJF).
Address Correspondence to: David K. Ahern, PhD, Abacus Health Solutions, 1210 Pontiac Ave, Cranston, RI 02920. Email: firstname.lastname@example.org.
1. CDC. National Diabetes Statistics Report, 2017. American Diabetes Association. 2017. Accessed June 30, 2019. https://dev.diabetes.org/sites/default/files/2019-06/cdc-statistics-report-2017.pdf
2. American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care. 2018;41(5):917-928. doi:10.2337/dci18-0007
3. Garvey WT, Mechanick JI, Brett EM, et al. American Association of Clinical Endocrinologists and American College of Endocrinology comprehensive clinical practice guidelines for medical care of patients with obesity. Endocr Pract. 2016;22(suppl 3):1-203. doi:10.4158/EP161365.GL
4. Trogdon JG, Hylands T. Nationally representative medical costs of diabetes by time since diagnosis. Diabetes Care. 2008;31(12):2307-2311. doi:10.2337/dc08-1375
5. American Diabetes Association. Standards of medical care in diabetes—2017. Diabetes Care. 2017;40(suppl 1).
6. American Diabetes Association. Standards of medical care in diabetes—2018. Diabetes Care. 2018;41(suppl 1).
7. Bailey CJ, Kodack M. Patient adherence to medication requirements for therapy of type 2 diabetes. Int J Clin Pract. 2011;65(3):314-322. doi:10.1111/j.1742-1241.2010.02544.x
8. Shah NR, Hirsch AG, Zacker C, Taylor S, Wood GC, Stewart WF. Factors associated with first-fill adherence rates for diabetic medications: a cohort study. J Gen Intern Med. 2009;24(2):233-237. doi:10.1007/s11606-008-0870-z
9. Stetson B, Schlundt D, Peyrot M, et al. Monitoring in diabetes self-management: issues and recommendations for improvement. Popul Health Manag. 2011;14(4):189-197. doi:10.1089/pop.2010.0030
10. Kazemian P, Shebl FM, McCann N, Walensky RP, Wexler DJ. Evaluation of the cascade of diabetes care in the United States, 2005-2016. JAMA Intern Med. 2019;179(10):1376-1385. doi:10.1001/jamainternmed.2019.2396
11. Ali MK, Bullard KM, Saaddine JB, Cowie CC, Imperatore G, Gregg EW. Achievement of goals in U.S. diabetes care, 1999-2010. N Engl J Med. 2013;368(17):1613-1624. doi:10.1056/NEJMsa1213829
12. Lorig K, Ritter PL, Turner RM, English K, Laurent DD, Greenberg J. Benefits of diabetes self-management for health plan members: a 6-month translation study. J Med Internet Res. 2016;18(6):e164. doi:10.2196/jmir.5568
13. Shortell SM, Poon BY, Ramsay PP, et al. A multilevel analysis of patient engagement and patient-reported outcomes in primary care practices of accountable care organizations. J Gen Intern Med. 2017;32(6):640-647. doi:10.1007/s11606-016-3980-z
14. Thornton Snider J, Seabury S, Lopez J, McKenzie S, Goldman DP. Impact of type 2 diabetes medication cost sharing on patient outcomes and health plan costs. Am J Manag Care. 2016;22(6):433-440.
15. Turner RM, Ma Q, Lorig K, Greenberg J, DeVries AR. Evaluation of a diabetes self-management program: claims analysis on comorbid illnesses, health care utilization, and cost. J Med Internet Res. 2018;20(6):e207. doi:10.2196/jmir.9225
16. Fisher EB, Fitzgibbon ML, Glasgow RE, et al. Behavior matters. Am J Prev Med. 2011;40(5):e15-e30. doi:10.1016/j.amepre.2010.12.031
17. Ahern DK, Braun IM, Cooley ME, Bickmore TW. Oncology informatics: behavioral and psychological sciences. In: Hesse BW, Ahern DK, Beckjord E, eds. Oncology Informatics: Using Health Information Technology to Improve Processes and Outcomes in Cancer. Academic Press; 2016:231-251.
18. Thirumurthy H, Asch DA, Volpp KG. The uncertain effect of financial incentives to improve health behaviors. JAMA. 2019;321(15):1451-1452. doi:10.1001/jama.2019.2560
19. Ahern DK, Buckel L, Aberger EW, Follick MJ. eHealth for employee health and wellness: optimizing plan design and incentive management. In: Pronk NP, ed. ACSM’s Worksite Health Handbook. A Guide to Building Healthy and Productive Companies. 2nd ed. Human Kinetics; 2009:248-258.
20. American Diabetes Association. 4. Comprehensive medical evaluation and assessment of comorbidities: standards of medical care in diabetes—2019. Diabetes Care. 2019;42(suppl 1):S34-S45. doi:10.2337/dc19-S004
21. Moore BJ, White S, Washington R, Coenen N, Elixhauser A. Identifying increased risk of readmission and in-hospital mortality using hospital administrative data: the AHRQ Elixhauser Comorbidity Index. Med Care. 2017;55(7):698-705. doi:10.1097/MLR.0000000000000735
22. Ho DE, Imai K, King G, Stuart EA. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit Anal. 2007;15(3):199-236. doi:10.1093/pan/mpl013
23. Stuart EA. Matching methods for causal inference: a review and a look forward. Stat Sci. 2010;25(1):1-21. doi:10.1214/09-STS313
24. Austin PC. Double propensity-score adjustment: a solution to design bias or bias due to incomplete matching. Stat Methods Med Res. 2017;26(1):201-222. doi:10.1177/0962280214543508
25. Howe CJ, Cole SR, Westreich DJ, Greenland S, Napravnik S, Eron JJ Jr. Splines for trend analysis and continuous confounder control. Epidemiology. 2011;22(6):874-875. doi:10.1097/EDE.0b013e31823029dd
26. Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol. 2005;162(3):199-200. doi:10.1093/aje/kwi188
27. Wei W, Pan C, Xie L, Baser O. Real-world insulin treatment persistence among patients with type 2 diabetes. Endocr Pract. 2014;20(1):52-61. doi:10.4158/EP13159.OR
28. Zhang H, Goldberg SI, Hosomura N, et al. Lifestyle counseling and long-term clinical outcomes in patients with diabetes. Diabetes Care. 2019;42(9):1833-1836. doi:10.2337/dc19-0629