Improving Diabetes Control in a Medicaid Managed Care Population With Complex Needs

Abstract

Introduction

People enrolled in Medicaid managed care who struggle with diabetes control often have complex medical, behavioral, and social needs. Here the authors report the results of a program designed to partner with primary care teams to address those needs.

Methods

A nonprofit organization partnered with a Medicaid managed care plan and a Federally Qualified Health Center in California to enroll people with A1cs >9% in a 12-month program. The program team included a community health worker, certified diabetes care and education specialist/registered dietitian, behavioral health counselor, and registered nurse. They developed patient-led action plans, connected patients to community resources, and supported behavior changes to improve diabetes control. Baseline assessments of behavioral health conditions and social needs were collected. Monthly A1c values were tracked for participants and a comparison group.

Results

Of the 51 people enrolled, 83% had at least 1 behavioral health condition. More than 90% reported at least 1 unmet social need. The average monthly A1c among program participants was 0.699 lower than the comparison group post-enrollment (P = .0008), and the disparity in A1c between Hispanic and non-Hispanic White participants at enrollment declined.

Discussion

Participants had high levels of unmet medical, behavioral, and social needs. Addressing these needs resulted in a rapid and sustained improvement in A1c control compared to non-enrollees and a reduction in disparity of control among Hispanic participants.

Conclusion

By partnering with a primary care team, a program external to Federally Qualified Health Center primary care can improve clinical outcomes for people with complex needs living with diabetes.

Introduction

People with poorly controlled diabetes often have complex medical, behavioral, and social needs—such as undiagnosed behavioral health conditions, food insecurity, and housing instability—that are not met by the health care system.^1–3 Poor diabetes control is particularly evident in the Medicaid managed care (MMC) population. In 2010, the proportion of people enrolled in an MMC plan with diabetes who had a hemoglobin A1c >9% was 44.0%, compared to 27.3% in a commercial managed care plan population. In 2020, it was 45.4%, compared to 37.5% among those enrolled in a commercial managed care plan.⁴ Health care systems, payers, and providers are seeking solutions to meet the complex medical, behavioral, and social needs of this population to improve their diabetes control. Although some existing approaches have demonstrated success in engaging the MMC patient population,^5,6 there is a need for a scalable, effective solution that works in partnership with stakeholders across the health care system.⁷ Vayu Health, a nonprofit organization in California, designed and implemented a program to address these unmet needs among MMC plan enrollees who struggle with their diabetes control. This article describes the intervention and its preliminary findings.

Methods

PARTICIPANTS AND SETTING

Vayu partnered with Health Net, an MMC Health Plan in California, and Ampla Health, a Federally Qualified Health Center (FQHC) in California, to enroll participants from an Ampla primary care clinic. Health Net provided A1c values from Healthcare Effectiveness Data and Information Set measures submitted by Ampla to identify eligible patients. This was supplemented with chart reviews by Vayu staff and referrals from Ampla providers. Enrollment criteria included a diagnosis of type 1 diabetes for at least 3 months or type 2 diabetes for at least a year; an age between 18 and 64, at least 1 A1c >9% within the last 6 months, and enrollment in the Health Net MMC Health Plan. Patients were excluded if they had untreated psychotic or bipolar disorders without psychiatry support, had cognitive impairment without a caregiver, were pregnant, had complicated malignancies, were terminally ill, were receiving palliative care, were in hospice care, or had other end-of-life issues. Members of the Vayu team identified, recruited, and enrolled participants between November 15, 2021, and March 31, 2022, either face-to-face after a clinic visit or by phone or text message. A comparison group of 477 patients was created by identifying all patients with diabetes at any Ampla clinic location who were members of the MMC Health Plan and who had at least 1 A1c ≥9% in the 12 months prior to November 2021 (the start of recruitment and enrollment of participants into the intervention program).

Intervention

The intervention was based on evidence about the effectiveness of health care system navigation, referral to community services, and mental/behavioral health support.^2,8,9 The program team included a community health worker (CHW), a certified diabetes care and education specialist (CDCES)/registered dietician (RD), and a behavioral health counselor. This team was employed by Vayu, with the CDCES/RD and behavioral health counselor seeing patients virtually. The CHW was embedded in the clinic, attended primary care visits, and provided care navigation after and in between primary care visits for needs such as coordinating prescription refills, following up on referrals, connecting with community resources to address social determinants of health, and scheduling follow-up appointments. The goal of the team was to prioritize patients’ self-identified needs and goals, build trust, develop patient-led action plans, and support targeted changes to improve glycemic control and overall health. Table 1 provides additional details about the program structure. The team members had frequent contact with each other and with patents via text messaging and phone calls as needed. The team also met weekly to discuss patient cases, prioritize activities, and solve problems.

Table 1:

Phases and components of the Vayu program

Program Phase	Phase Components
1. Recruitment and enrollment	Identify potential patients for the program Verify eligibility Inform primary care physician Outreach to patient Complete initial onboarding visit
2. Understanding members and creating personalized care plans	First Vayu visit First primary care physician visit First CDCES/RD visit First behavioral health counselor visit Behavioral health and social determinants of health assessment surveys
3. Medical, behavioral, and social care pathways	Appointments with the CHW in the clinic, at the patient’s home, and/or in the community Virtual appointments with the CDCES/RD and behavioral health counselor Communication via text messaging with the CHW, CDCES/RD, and behavioral health counselor Primary care physician follow-up appointments, CHW attends with patient Monthly hemoglobin A1c test in clinic before primary care physician visits
4. Graduation	Connect patient to community and specialty resources Graduation visits with the CHW, CDCES/RD, and behavioral health counselor
5. Post-graduation follow-up	Post-graduation follow-up at regular intervals from the CHW

Data Collection and Measures

Participant characteristics were collected at enrollment, and validated surveys (ie, Protocol for Responding to & Assessing Patients' Assets, Risks & Experiences [PRAPARE] social needs screening tool,¹⁰ ACE Survey,¹¹ PHQ-9,¹² and PAM¹³) were used by program team members to identify medical, behavioral, and social needs. Comparison group A1c values were obtained from Healthcare Effectiveness Data and Information Set measurements submitted by Ampla to Health Net. The A1c values for participants in the program were abstracted from the medical record by the CDCES/RD on the team. All virtual, in-person, phone, or other contacts with the program team were also entered into a tracking database by team members.

Data Analysis

A monthly average of A1c values for enrollees and those in the comparison group was computed from January of 2021 through March of 2023. The authors conducted a difference-in-difference (DiD) interrupted time series (ITS) analysis¹⁴ using segmented regression.¹⁵ This approach is broadly used and accepted in public health and health policy analyses and is considered the strongest approach for evaluating the effects of quality improvement programs.¹⁶ Differences were calculated by subtracting the mean A1c in the intervention group from the mean A1c in the control group in each month. The authors hypothesized that the average monthly differences in the post-period would be larger than in the pre-intervention period. The authors included a parameter for the secular trend in mean A1c (time), an intercept parameter (to estimate the change in level), and a slope parameter (to estimate the rate of change). They also included an autoregressive parameter to control for serial autocorrelation. All analyses were conducted in SAS 9.4.¹⁷

As nearly half of participants identified as Hispanic (any race), and evidence indicates higher A1c values for these individuals compared with people who identify as non-Hispanic and White,^18,19 the authors compared the last A1c value before enrollment with the most recent A1c value 12 months later for these 2 groups. The Kaiser Permanente Washington Health Research Institute Institutional Review Board determined that this activity was designed to evaluate an existing program and did not meet the definition of human subject research per federal regulations (45 CFR 46).

Results

Of the 145 eligible patients at the Ampla primary care clinic who met enrollment criteria, 51 enrolled between November 15, 2021, and March 31, 2022. Participant characteristics are in Table 2. The mean age was 46.4 years (standard deviation, or SD, 12.0, range 19–64), and 9 of the enrollees (18%) had type 1 diabetes. As of March 31, 2023, the mean length of time in the program was 364.6 days (SD 74.9, range 188–496). Over that time span, participants averaged 22.1 (SD 11.0, range 5–58) contacts with a member of the Vayu team.

Table 2:

Characteristics of Vayu program participants (n = 51)

Gender	Participant % (N)
Male	31.3% (16)
Female	62.7 (32)
Transgender	2 (1)
Other/Unknown	4 (2)
Race and Ethnicitya
Non-Hispanic and White	39.2% (20)
Hispanic (any race)	35.3 (18)
Black	5.8 (3)
Asian	2.0 (1)
Multiracial	11.8 (6)
Education
Less than high school	25.5% (13)
High school graduate	21.6 (11)
More than high school	37.3 (19)
Unknown	15.7 (8)
Type 1 Diabetes	18% (9)

^a3 were missing race and ethnicity (program enrollees were asked at baseline to self-identify a single ethnicity category, as Hispanic or non-Hispanic, and a single race, as Black, Asian, multiracial, or other race). Participants were informed they could skip these questions.

A total of 8 of the early participants declined to participate in the psychosocial components of the baseline assessments because they did not understand the behavioral health support that was offered at the time of enrollment, although they did continue in the program. Of the 43 participants who completed the baseline assessments, 83% had at least 1 behavioral/mental health condition such as a depressive disorder (56%), trauma- and stress-related disorder (51%), anxiety disorder (26%), substance use disorder (20%), or bipolar disorder (5%). Of these, 44% had a diagnosis for at least 2 conditions. Approximately half (51%) had experienced at least 2 or more adverse childhood experiences. The most common social health needs identified through the PRAPARE screening tool were transportation insecurity (37%), utility insecurity (36%), and food insecurity (29%). More than 90% of patients reported at least 1 unmet social health need. As the Vayu team established rapport and trust with patients over the first 3 months, they reported increases for insecurities around food (63%), housing (49%), and employment (18%). These needs were prioritized, with the CHW providing connections to local resources.

Trends in A1c values for the participants in the program vs the comparison group are reported in Figure. Mean A1c values for the comparison group exceeded 9% in 7 of the 12 months in the post-enrollment period compared with only 1 month for program participants. Table 3 shows the results of the DiD ITS analysis. Controlling for the secular trend in mean A1c values and an autoregressive parameter, the mean difference in A1c between the comparison group and program participants was significantly larger in the post-implementation period by about 0.699 (p = 0.0008). In other words, the average A1c in the intervention group was lower post-implementation. The rate of change (ie, slope parameter) was not significantly different between the 2 groups. Additionally, the disparity in A1c control between participants who self-identified as Hispanic (any race) vs non-Hispanic and White decreased from 11.9% (SD 1.8) and 10.3% (SD 1.4) respectively, at the time of enrollment (p < 0.01), to 8.2% (SD 1.4) and 7.9% (SD 1.3), respectively, 1 year after enrollment (p > 0.10).

Figure:

A1c trend for participants and comparisons.

Table 3:

Difference-in-difference interrupted time series regression results

Variable	Estimate	Standard Error	T-Value	p Value
Intercept	−0.934	0.132	−7.1	<0.0001
Intervention (level)	0.697	0.180	3.87	0.0008
Time	0.023	0.017	1.35	0.189
Time after (slope)	0.029	0.021	1.38	0.18

Discussion

MMC plan enrollees with A1c values >9% who were cared for by an FQHC primary care team had high levels of complex, unmet medical, behavioral, and social needs. For example, the rate of depressive disorder (51%) was 4–5 times higher than that in the larger population of people living with diabetes.²⁰ A high proportion (51%) also reported adverse childhood experiences.^21,22

A1c values among patients who enrolled in the program showed rapid, sustained improvement and were significantly lower in the post-enrollment period compared to non-enrollees, although there was no significant difference in the rate of change between the 2 groups. It is possible that when patients reengaged with their primary care physician as pandemic restrictions were lifted during 2022, both groups experienced improvements in diabetes control. Even so, those enrolled in the program had significantly lower A1c values on average during the post-enrollment period.

In addition, the authors observed that the disparity in A1c values at enrollment between participants who self-identified as Hispanic (any race) vs non-Hispanic and White was no longer present 12 months after enrollment. Members of the program team attributed these findings to their work to prioritize addressing patient-identified goals, building patient trust, and enhancing access to diabetes, behavioral health, and social health management support. This attribution is supported by evidence that people with behavioral health needs are less likely to meet A1c targets,²³ and that providing integrated care similar to this program can improve their diabetes control.²⁴

This analysis had several limitations. The team continued to make iterative improvements in the program over the first year, so patients who joined the program early on may have had different experiences than those who enrolled later. Although there may also be subtle but important differences between those who enrolled in the Vayu program compared with those who were eligible but did not, a particular strength of the DiD ITS analytic approach is that because it evaluates changes in rates/values of an outcome in a cohort of patients, confounding by individual-level variables cannot introduce serious bias.²⁵ Although A1c data were obtained from different sources for the comparison group vs participants in the program, the ITS approach controlled for secular trends, used cohort data, and tested for differences in means and trends between the 2 groups.²⁵ Finally, Vayu launched enrollment as the COVID-19 pandemic began to wane and patients started to seek care again. This general reengagement in health care could have a confounding effect on the findings, although the comparison group experienced the same health care environment.

Conclusion

A program external to an FQHC designed to meet complex medical, behavioral, and social needs in partnership with a primary care team demonstrates promise in improving clinical outcomes for MMC plan enrollees who struggle with their diabetes control. Given the small sample size at a single clinic site, a more rigorous evaluation of the successful spread of this program to additional clinic sites and inclusion of a larger population of patients are needed. This evaluation should also include the costs of such a program, although given its impact on equity and A1c control, it may be less important to policymakers and stakeholders to demonstrate a downstream financial return on investment.