Workforce Innovation Reduces Medicaid Costs in Chronic Care

ABSTRACT

Effective chronic disease management must extend beyond clinical visits into the daily lives of patients, particularly in low-income communities with a disproportionate burden of illness. This study examines City Health Works’ intervention model, which deploys highly trained nonclinical health coaches as tightly integrated extensions of primary care teams to support patient self-management. In a 12-month evaluation of Medicaid patients with poorly controlled diabetes and hypertension at a NYC Health + Hospitals outpatient site, the intervention achieved significant reductions in health care costs compared with a matched comparison group. These findings suggest that a technology-enabled, community-based workforce model can cost-effectively improve chronic disease management when closely linked to primary care delivery.

Am J Manag Care. 2025;31(Spec. No. 10):SP698-SP708. https://doi.org/10.37765/ajmc.2025.89796

Chronic diseases represent a significant health care challenge in the US, driving the majority of health care expenditures. Diabetes affects more than 11% of the population and 25% of older adults, with associated costs reaching $327 billion in 2017.^1,2 Only 26% of the US population achieved control of diabetes in 2018-2019.³ Similarly, hypertension affects almost half of all adults, but only a quarter have it under control.

People in low-income households experience a higher prevalence of multiple chronic conditions,⁴ which is driven by factors such as social stressors, limited health literacy, reduced health care access, and diminished trust in health care services. Consequently, they face compounded impacts of chronic disease and socioeconomic disadvantage. Although chronic conditions are not curable, complications can be prevented through effective control.

Previous efforts to use community health workers for diabetes self-management support have demonstrated clinical improvement but have rarely evaluated cost reduction.^5,6 Some disease management programs relied on physicians to implement complex, time-intensive protocols, whereas others used claims data to drive remote interventions by nurses. Although some of these approaches improved clinical quality indicators, significant cost savings have remained elusive.^7,8

Intervention

Evidence-based care for chronic disease control requires a dual focus on continuous medical checkups⁹ and active patient self-management. City Health Works developed an intervention modeled on the Wagner Chronic Care Model¹⁰ to stabilize the clinical status of patients with poorly controlled diabetes and/or hypertension.¹¹ This study examined outcomes for patients with diabetes and hypertension, but other research has evaluated City Health Works’ approach for those with asthma¹² and studies are underway for patients with chronic obstructive pulmonary disease.¹³

The health coaching intervention had 4 core components: (1) planned, regular interactions between patients and primary care teams; (2) patient education in self-management skills, supported by community engagement; (3) investment in health care provider training for evidence-based chronic care; and (4) technology integration for clinical information sharing and point-of-care workforce support. The coaching curriculum was adapted from the Association of Diabetes Care & Education Specialists’ 7 self-care behaviors (ADCES7)¹⁴ for delivery by a nonclinical workforce.

City Health Works recruited individuals with a minimum of a high school education from the communities it served and trained them in the competencies needed for successful chronic disease self-management. Health coaching was delivered under the supervision of registered dietitian certified diabetes educators. Training entailed a dedicated up-front learning period followed by on-the-job training, entailing real-time communication with the clinical supervisor and weekly panel review discussions. The 12-month coaching intervention consisted of a 3-month active phase (~ 8-10 in-person sessions) followed by a 9-month maintenance phase (monthly phone check-ins). The intervention components are outlined in the Table.

METHODS

Study Population and Design

The study enrolled adult patients with rising clinical risk, as determined by laboratory results and clinician judgment. Inclusion criteria included patients 18 years or older with a recent hemoglobin A_1c level higher than 9% and/or blood pressure greater than 140/90 mm Hg, with a primary care visit within the prior 3 months. Exclusion criteria included end-stage illness, pregnancy, active cancer care, active substance use disorder, and serious mental illness. Health coaches contacted identified patients by phone for enrollment.

Between March 2016 and June 2017, 100 program participants meeting these criteria were enrolled. Using New York State (NYS) Medicaid data, personal identifiers collected at enrollment were matched to Medicaid claims. Sixty-six program participants were confirmed to be enrolled in Medicaid at the time of program entry and made up the program group for analysis.

To construct a comparison group, we selected all other NYS Medicaid enrollees who had visited any NYC Health + Hospitals (NYCHH) facility during the 16-month enrollment window. For each potential comparison group member, a pseudo-enrollment month was randomly selected from the distribution of enrollment months in the treatment group. Demographic, diagnostic, and utilization measures were calculated from NYS Medicaid claims data for the 12 calendar months prior to the calendar month of (pseudo-)enrollment (hereafter referred to as the baseline period), as well as for the first and second years of follow-up (months 1-12 and 13-24 after the enrollment month).

Baseline variables for matching, weighting, and impact estimation procedures included as follows:

• Demographics: age, race/ethnicity, sex, Supplemental Security Income (SSI) coverage, dual Medicaid/Medicare coverage, months of Medicaid enrollment
• Costs and utilization: total Medicaid fee-for-service and Medicaid managed care plan payments; cost and number of visits/inpatient stays for individual categories of service, including inpatient, emergency department, primary care provider, specialty care, laboratory, pharmacy, home health, and other costs
• Program qualifying diagnoses: diabetes, hypertension, metabolic syndrome (primary or secondary diagnosis on any claim/encounter report)
• Exclusionary diagnoses: serious mental illness, substance use disorder, pregnancy, cancer (primary or secondary diagnosis on any claim/encounter report)

Comparison Group Matching and Weighting

Comparison group refinement occurred in 3 stages: (1)eligibility filtering:applying diagnostic inclusionary/exclusionary criteria to all Medicaid enrollees who had visited NYCHH facilities during the City Health Works enrollment period and were screened for having at least 2 months of Medicaid enrollment in both baseline and the first follow-up year (relative to each pseudo-enrollment date); (2) matching: using 4:1 Mahalanobis distance matching on baseline demographics, diagnoses, and utilization, constrained to match within 0.2 SD of a treatment propensity score; and (3) weighting: applying entropy balance weights to further balance baseline covariates and estimate the average treatment effect among the treated.

Statistical Analysis

Program impacts on Medicaid costs in total and within major service categories were estimated using Poisson analysis of covariance models, adjusting for least absolute shrinkage and selection operator–selected baseline measures.

RESULTS

Baseline Characteristics

The program group’s baseline characteristics reflected a diverse, high-need population, with a mean (SD) age of 59 (11) years (range, 20-80). Of this group, 67% were women, 50% were Hispanic, 21% were non-Hispanic Black, 44% were Medicare beneficiaries, and 36% were receiving SSI. Mean (SD) total Medicaid costs were $24,441 ($38,198).

Baseline utilization patterns showed mean (SD) values of the following: hospital stays, 0.83 (1.89); hospital cost, $9502 ($24,236); primary care visits, 5 (7); primary care cost, $668 ($1028); home health cost, $7250 ($21,079); and pharmacy cost, $6139 ($13,530).

Comparison Group Characteristics

Comparison group matching and weighting resulted in virtually identical baseline characteristics in the program and comparison groups. For details, see eAppendix Table 1 (eAppendix available at ajmc.com).

Intervention Impact

The program achieved significant cost savings. There was a 36% reduction in Medicaid expenditures compared with the control group. The mean cost was $26,091 per program participant vs $40,953 in the comparison group, resulting in a first-year cost savings per patient of $14,863 (P = .49). The 12-month intervention cost was $2125 per patient, yielding a return on investment (ROI) of 7:1.

As the month-over-month trend in baseline costs in the matched and weighted sample makes clear (eAppendix Table 1), costs increased rapidly in both groups during the baseline period, which may help explain the continuing increase in comparison group costs in the follow-up period. Cost impacts in the second year were attenuated toward zero and not statistically significant, suggesting the program may have focused on averting costs and improving care efficiency within a shorter time horizon.

DISCUSSION

This case study suggests that training and supervising a nonclinical workforce to extend the reach of primary care may be associated with reduced Medicaid spending for patients with chronic conditions. The model implemented at City Health Works—integrating disease-specific knowledge, motivational interviewing techniques,¹⁵ and real-time clinical supervision—offers a potential strategy to address workforce shortages and cost inefficiencies in chronic disease management.

Although the results are promising, the effects diminished in the second year, with comparison group costs decreasing far more in the second year (relative to the first year) than treatment group costs. This could reflect the fact that comparison group episodes of acute care were finally being resolved in the second follow-up year, with similar episodes in the treatment group having been resolved in year 1 due to the intervention. It is also possible, however, that if the intervention had been extended into a second year of active treatment, further cost savings might have been realized in the treatment group, resulting in positive year 2 effects.

Our findings highlight the opportunity to intervene earlier among patients experiencing recent clinical deterioration, rather than targeting only the highest-cost superutilizers.¹⁶ The study cohort included patients identified through clinical criteria and primary care clinician judgment as having poor control of their chronic conditions but not yet experiencing catastrophic health events. A similar targeting strategy was associated with success in programs such as CareMore Health¹⁷ and Waymark,¹⁸ suggesting that engaging populations during the time in which clinical status is worsening may maximize clinical and cost outcomes.

Moreover, prior qualitative research conducted by the Urban Institute reinforces the importance of tight integration between community-based health workers and primary care teams.¹¹ City Health Works’ emphasis on electronic health record access, real-time supervisory support, and structured escalation pathways enabled frontline health coaches to identify and address emerging medication-related issues effectively—an operational feature that may be critical to replicating outcomes in other settings.

Several limitations should be considered when interpreting these results. This small, exploratory study lacked a formal a priori power analysis and relied on Medicaid claims data, restricting assessment to utilization and cost measures without direct clinical outcomes. Although matching, weighting, and a difference-in-differences design were used to strengthen internal validity, the possibility of unmeasured confounding remains. Furthermore, the program was implemented within a single organizational setting, which may limit generalizability to other Medicaid populations or care environments. Nonetheless, these findings build on early industry research on ROI¹⁹ and offer preliminary evidence supporting further research on nonclinical workforce models in chronic care management.

The use of a comparison group and a quasi-experimental approach lends suggestive evidence that workforce innovations such as those piloted at City Health Works can cost-effectively complement primary care and help manage high-need, high-cost populations. Future research should evaluate similar models in larger, more diverse populations and incorporate direct measures of clinical outcomes.

CONCLUSIONS

As primary care workforce shortages persist, findings from this study suggest that training a new nonclinical workforce in chronic condition management can cost-effectively strengthen value-based care models. The intervention demonstrated success in controlling costs without increasing workflow burden for primary care providers. This approach may be particularly valuable in low-income and rural communities, where clinician shortages and patient needs are most acute. Although the results are promising, the effects diminished in the second year, underscoring the need for ongoing support. Chronic conditions are inherently difficult to manage long term due to fluctuating health status, life stressors, and aging, all of which can undermine stability, even with strong initial gains. The model offers a scalable strategy for health care organizations seeking material cost savings in full-risk arrangements while improving chronic disease management among vulnerable populations.

Acknowledgments

The views and opinions expressed here are those of the authors and do not necessarily reflect the official policy or position of the New York State Department of Health. Examples of analysis performed within this article are only examples. They should not be utilized in real-world analytic products.

Author Affiliations: Diverge Health, Carmel, IN (MK), Brooklyn, NY (BI), and Chicago, IL (BB); Department of Applied Statistics, Social Science, and Humanities, Steinhardt School of Culture, Education, and Human Development, New York University (TM), New York, NY; Peterson Center on Healthcare (PS), New York, NY; Ronald McDonald House New York (JH-R), Brooklyn, NY; Triple Aim Partners (KF), Chicago, IL; Columbia Business School (LG), New York, NY.

Source of Funding: Department of Population Health at the New York City Health and Hospitals Corporation.

Author Disclosures: Ms Kaur, Dr Ives, and Mr Bhansali are employed by and own shares in Diverge Health, which delivers care to Medicaid patients. Mr Francis is employed as an actuary by Triple Aim Partners and participates in Society of Actuaries continuing education. 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 (MK, BI, PS, JH-R, LG); acquisition of data (TM); analysis and interpretation of data (BI, TM, PS, KF, LG); drafting of the manuscript (MK, BI, TM, JH-R, KF, BB, LG); critical revision of the manuscript for important intellectual content (MK, BI, TM, PS, JH-R, BB, LG); statistical analysis (TM, KF); provision of study materials or patients (MK); and administrative, technical, or logistic support (BB).

Send Correspondence to: Manmeet Kaur, MBA, Diverge Health, 3839 Pelham Rd, Carmel, IN 46074. Email: manmeet1024@gmail.com.

REFERENCES

1. National Diabetes Statistics Report. CDC. Accessed August 12, 2025. https://www.cdc.gov/diabetes/php/data-research/index.html
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. Chen Y, Rolka D, Xie H, Saydah S. Imputed state-level prevalence of achieving goals to prevent complications of diabetes in adults with self-reported diabetes — United States, 2017–2018. MMWR Morb Mortal Wkly Rep. 2020;69(45):1665-1670. doi:10.15585/mmwr.mm6945a1
4. Jayathilaka R, Joachim S, Mallikarachchi V, Perera N, Ranawaka D. Do chronic illnesses and poverty go hand in hand? PLoS One. 2020;15(10):e0241232. doi:10.1371/journal.pone.0241232
5. Pérez-Escamilla R, Damio G, Chhabra J, et al. Impact of a community health workers-led structured program on blood glucose control among Latinos with type 2 diabetes: the DIALBEST trial. Diabetes Care. 2015;38(2):197-205. doi:10.2337/dc14-0327
6. Kim KB, Kim MT, Lee HB, Nguyen T, Bone LR, Levine D. Commu­nity health workers versus nurses as counselors or case managers in a self-help diabetes management program. Am J Public Health. 2016;106(6):1052-1058. doi:10.2105/AJPH.2016.303054
7. Brandt S, Hartmann J, Hehner S. How to design a successful disease-management program. McKinsey & Company. October 1, 2010. Accessed April 15, 2024. https://www.mckinsey.com/industries/healthcare/our-insights/how-to-design-a-successful-disease-management-program
8. Grudniewicz A, Gray CS, Boeckxstaens P, De Maeseneer J, Mold J. Operationalizing the chronic care model with goal-oriented care. Patient. 2023;16(6):569-578. doi:10.1007/s40271-023-00645-8
9. Sonmez D, Weyer G, Adelman D. Primary care continuity, frequency, and regularity associated with Medicare savings. JAMA Netw Open. 2023;6(8):e2329991. doi:10.1001/jamanetworkopen.2023.29991
10. Wagner EH. Chronic disease management: what will it take to improve care for chronic illness? Eff Clin Pract. 1998;1(1):2-4.
11. Waxman E, Thompson M, Aron LY, Kaur M, Hoy-Rosas J, Haag E. Improving chronic illness management in Harlem: leveraging community health coaches to address the challenge of medication management. Urban Institute. October 2019. Accessed April 15, 2024. https://www.urban.org/sites/default/files/publication/101262/improving_chronic_illness_management_in_harlem_0.pdf
12. Federman AD, O’Conor R, Mindlis I, et al. Effect of a self-
    management support intervention on asthma outcomes in older adults: the SAMBA study randomized clinical trial. JAMA Intern Med. 2019;179(8):1113-1121. doi:10.1001/jamainternmed.2019.1201
13. Federman AD, O’Conor R, Nnemnbeng J, et al. Feasibility trial
    of a comprehensive, highly patient-centered COPD self-
    management support program. Chronic Obstr Pulm Dis. 2024;11(1):13-25. doi:10.15326/jcopdf.2023.0419
14. Association of Diabetes Care and Education Specialists; Kolb L. An effective model of diabetes care and education: the ADCES7 Self-Care Behaviors. Sci Diabetes Self Manag Care. 2021;47(1):30-53. doi:10.1177/0145721720978154
15. Jones-Smith E. Motivational interviewing and the stages of change theory. In: Theories of Counseling and Psychotherapy: An Integrative Approach. 2nd ed. SAGE Publications; 2016:317-344.
16. Johnson TL, Rinehart DJ, Durfee J, et al. For many patients who use large amounts of health care services, the need is intense yet temporary. Health Aff (Millwood). 2015;34(8):1312-1319. doi:10.1377/hlthaff.2014.1186
17. Modarai F, Powers BW, Palakodeti S, Garg V, Jain SH. How to improve care for high-need, high-cost Medicaid patients. Harvard Business Review. February 27, 2020. Accessed April 15, 2024. https://hbr.org/2020/02/how-to-improve-care-for-high-need-high-cost-medicaid-patients
18. Baum A, Batniji R, Ratcliffe H, DeGosztonyi M, Basu S. Supporting rising-risk Medicaid patients through early intervention. NEJM Catal Innov Care Deliv. 2024;5(11). doi:10.1056/CAT.24.0060
19. Kangovi S, Mitra N, Grande D, Long JA, Asch DA. Evidence-based community health worker program addresses unmet social needs and generates positive return on investment. Health Aff (Millwood). 2020;39(2):207-213. doi:10.1377/hlthaff.2019.00981