How Would Low-Income Communities Prioritize Medicaid Spending?

Abstract

Context:

Medicaid plays a critical role in low-income, minority, and medically underserved communities, particularly in states that have expanded Medicaid under the Affordable Care Act. Yet, the voices of underresourced communities are often unheard in decisions about how to allocate Medicaid’s scarce resources, and traditional methods of public engagement are poorly suited to gathering such input. We argue that deliberative public engagement can be a useful tool for involving communities in setting Medicaid priorities.

Method:

We engaged 209 residents of low-income, medically underserved Michigan communities in discussions about Medicaid spending priorities using an exercise in informed deliberation: CHAT (CHoosing All Together). Participants learned about Medicaid, deliberated in small groups, and set priorities both individually and collectively.

Findings:

Participants prioritized broad eligibility consistent with the ACA expansion, accepted some cost sharing, and prioritized spending in areas—including mental health—that are historically underfunded. Participants allocated less funding beyond benefit coverage, such as spending on healthy communities. Participants perceived the deliberative process as fair and informative, and they supported using it in the policy-making process.

Conclusion:

The choices of participants from low-income, medically underserved communities reflect a unique set of priorities and suggest that engaging low-income communities more deeply in Medicaid policy making might result in different prioritization decisions.

Budget constraints for states and the federal government generate frequent policy debates about Medicaid’s funding structure, payment rates, enrollee cost sharing, and eligibility. These debates are important, but they obscure a central underlying issue: Given scarce resources, what benefits should Medicaid prioritize? Should priority be given to those with the greatest health needs, or should children in poor families have wellness and prevention covered to promote a healthy start in life? Should out-of-pocket spending be kept low, to decrease financial obstacles to care seeking, or should some individuals face copayments or even premiums? Should priority be given to individual benefit coverage or to community-wide health promotion? And what process should be used to make these decisions?

Consider California’s decision in 2009 to eliminate nine optional benefits, most notably dental coverage for adult recipients, during the great recession. The decision to focus spending cuts on adult dental services, made unilaterally by the Schwarzenegger administration as part of a much larger package of required budget cuts, was possible because the federal government designates most dental services as optional benefits. Subsequently, adult Medicaid recipients in California dramatically reduced their use of routine and preventative dental services and increased their use of emergency departments for dental problems, effects that were most sharply felt by members of minority groups (Singhal et al. 2015). The change increased demand for safety net services and stressed the capacity of safety net providers, reducing their ability to serve their communities (Wides, Rab Alam, and Mertz 2014). Yet, this decision was made with little public debate or attempt to include the perspectives of the people or communities most affected.

In contrast to this process in California, and the process underlying many similar decisions about Medicaid coverage and eligibility around the country, we argue that setting priorities for Medicaid spending should include the voices of those most affected: Medicaid recipients, their families, and other residents of low-income and medically underserved communities. Making decisions about Medicaid priorities requires more than technical expertise; justice demands that these decisions incorporate the needs and values of the population served. Involving lay (nonprofessional) members of the public in decision making about Medicaid priorities can enhance the quality of these decisions by incorporating the perspectives and situated knowledge of those directly affected by these decisions (Goold 1996; Fleck 2009). Despite this, little research has examined how the public in general, or members of low-income communities in particular, would prioritize Medicaid spending.¹

In this article, we argue that deliberative public engagement can provide a way for the public to play a role in complex health policy decisions like the prioritization of Medicaid resources. Deliberative public engagement involves members of the lay public in a structured process of education, discussion, and recommendation. This combination of education and discussion with others encourages reasoning, reflection, and consideration of the knowledge and values of others. The result is informed opinions from members of the lay public that can serve as an input into policy processes on even esoteric or highly complex policy topics. We adapted an existing deliberative exercise, CHAT (CHoosing All Together), to facilitate deliberation among members of the lay public about Medicaid spending priorities. We report on the CHAT-derived priorities of low-income communities in Michigan, a state whose political and demographic composition make it of particular interest to Medicaid policy makers (Ayanian 2013). We find strong support for expanding Medicaid, generous funding for some services classified as optional by Center for Medicare and Medicaid Services (CMS) such as dental services, and less generous coverage for some particularly expensive areas, such as facility care and specialty care.

Why Deliberation about Medicaid?

Discussions about priority setting in health policy often contain calls to include the views of the broader public. These calls are often made from a moral or ethical perspective. Fleck (2009) argues that since any health care prioritization decision involves placing some needs above others, a process of open, public deliberation is necessary prior to setting priorities (see also Goold 1996; Solomon, Gusmano, and Maschke 2016). Reaching beyond technical experts to incorporate the views of the broader public might yield more tangible benefits as well. Policy decisions that incorporate public input may be viewed as more legitimate, and thus enjoy greater public support. Incorporating the views of the lay public can also incorporate the situated knowledge of members of the lay public that may be inaccessible to experts. Members of affected communities often know the needs and special circumstances of their communities in ways that outsiders cannot know (Young 2004).

Including the views of those most affected by health policy decisions is particularly important for programs like Medicaid. Medicaid recipients are almost by definition not represented among the various communities of experts—health care professionals, federal and state bureaucrats, academics, and the like—who tend to make decisions about how to prioritize Medicaid’s resources. This means that, while such experts can attempt to do what they think is best for Medicaid recipients, they lack a direct connection to the lived experiences of those who rely on such programs. Advocacy organizations can fill some of this gap but may do so imperfectly (Grogan and Gusmano 2007). To the extent that communities with large numbers of Medicaid recipients tend to be politically disenfranchised, these communities will also be less able to push back on prioritization decisions they disagree with through the normal political process. Including input from low-income communities in prioritization decisions can make these decisions more just, and more suited to the needs, values, and experiences of these communities.

Despite this, there is a dearth of research about what members of low-income communities, or even the public as a whole, think about how Medicaid should balance competing priorities. Some work examines the role of public opinion in support for the program overall (Shaw 2009; Barrilleaux and Rainey 2014; Grogan and Park 2017; KFF 2017). However, this research sheds little light on what the public thinks about specific elements of Medicaid. Even less work examines the views of low-income or minority populations about Medicaid. What work exists is focused on what recipients think about controversial program features like work requirements or required health savings accounts (Sommers et al. 2018), or the ACA expansion itself (Epstein et al. 2014). While useful for evaluating these policies, this work tells us little about more common and consequential prioritization decisions.

This gap in knowledge about public priorities for Medicaid coverage may well reflect the difficulty of soliciting the informed opinions about prioritizing scarce health care resources. Discussions about health care and financing can be complex, so people may be frustrated or intimidated by attempts to probe their knowledge of these topics, and few citizens can offer more than top-of-the-head opinions about policies like Medicaid (Zaller 1992; Weinfurt et al. 2006). Health concerns lack salience for healthy citizens and talking about health insurance requires people to think about illness and death, creating emotional tension when money appears to be pitted against health (Danis et al. 2014). These challenges can be particularly difficult when trying to incorporate the views of those most affected by Medicaid policy, as disadvantaged populations often lack resources of time, money, and education, and can thus be particularly difficult to engage using traditional methods like polls and public meetings (Bonevski et al. 2014). Finally, many common engagement methods tend to treat people as isolated individuals, failing to truly incorporate a community perspective into the decision-making process for shared, pooled resources.

Deliberative methods of engagement can address these concerns, allowing the public to play a role in complex health policy decisions like the prioritization of Medicaid resources. Deliberative engagement brings together members of the lay public to engage in an intensive process of learning about, discussing, and rendering judgement on important public questions. By combining education with discussion among members of the community, deliberative methods encourage reasoning, reflection, and consideration of the knowledge and values of others. Moreover, incorporating discussion among members of the affected community encourages participants to ask both what we should do as a community and what I want as an individual. Deliberative methods have been shown to change individual attitudes, increase knowledge (Fishkin 2009) and trust (Boulianne 2019), and, in some cases, they show evidence of public-spiritedness (Goold et al. 2004). As we show in this project, deliberative engagement can be conducted in a way consistent with the principles of Community Based Participatory Research (CBPR) (Israel et al. 1998; Israel et al. 2012). Like CBPR, deliberative engagement aims to include voices from communities that are often treated as the subjects of policy making rather than partners in the policy-making process.

Deliberative methods have increasingly been used to incorporate the views of the public into bioethical and health policy decision making (Abelson et al. 2013; Carman et al. 2013). However, these methods have rarely been used to inform the design of states’ Medicaid programs. The most notable exception is Oregon’s attempt in the early 1990s to prioritize health care services using a process of extensive public engagement, including 47 public meetings involving over a thousand state residents (Garland and Hasnain 1990). While this effort has been cited as an example of deliberative engagement applied to health care priority setting (IOM 2012: 110), critics noted that, despite the considerable expense of these public engagement efforts, the program largely failed to represent the views of the poor and minorities. (Goold 1996; Young 2001). Similarly, Grogan and Gusmano (2005, 2007) analyzed Connecticut’s Medicaid Managed Care Council from the perspective of deliberative democratic theory but argued that this forum sidelined the voices of Medicaid recipients, particularly minority recipients, and that professional advocates for recipients failed to faithfully represent their interests in the process. Moreover, both of these examples predate the ACA expansion, and thus provide little guidance about the informed views of the public about current policy decisions facing Medicaid policy makers.

Methods

To facilitate deliberation about Medicaid priorities constrained by limited resources, we adapted the existing exercise, CHAT, which aims to promote informed, reasoned dialogue among ordinary citizens about complex and value-laden allocation decisions (Goold et al. 2005). CHAT presents participants with a game board that resembles a pie chart (figure 1). Each wedge of the circle represents a category of spending. Within each category, different amounts of spending are represented by different levels within each wedge. For most categories, this includes the option to spend no money on the category. Participants select the level of funding for each category by allocating the number of markers required for the level they select. However, the number of markers available to participants does not allow them to select the highest level of spending for every category. Thus, allocation decisions involve tradeoffs; choosing high spending in one category requires lower or no spending in another. The costs and coverage offered at each level were chosen to reflect actual costs and benefits; thus, the cost-benefit tradeoffs faced by CHAT participants mirror those faced by policy makers.

Figure 1.

Medicaid CHAT board.

CHAT is played in four rounds in a group of roughly a dozen participants. In the first round, participants set priorities as individuals; in the second round they set priorities in small groups of two to four participants. After rounds 1 and 2, the whole group hears and discusses scenarios (“events”) that illustrate the consequences of their choices. In round 3, which generally takes up the bulk of the CHAT session, the entire group deliberates to come to a collective decision about how to set priorities. Group deliberation is led by a trained facilitator, who ensures broad participation, asks participants to articulate reasons for their preferences, and encourages them to make fair decisions on behalf of fellow community members. In the fourth and final round, participants again make individual allocations, which do not need to match the allocations made by their group and may differ from individual priorities they selected in round 1.

Adapting CHAT to Medicaid Priority Setting

Adapting CHAT to the task of setting priorities for Medicaid spending requires adapting the categories of spending, the possible levels of spending within these categories, and the descriptions of each category and the benefits provided under each level of spending to reflect the Medicaid program. Since Medicaid budgets and costs are state-specific, we adapted CHAT based on the actuarial information from a single state, Michigan. Michigan is a particularly apt location for such a single-state study. It is a medium-sized “purple” state that, at the time of this study, had a Republican governor who supported expanding Medicaid, making it a useful case to study the politics of Medicaid in the Affordable Care Act era (Ayanian 2013).

To adapt CHAT to the task of setting priorities for Medicaid spending in Michigan, we used a CBPR approach that involved working with community partners in all phases of the project. This process was led by one academic and one community codirector (Susan Goold and Zachary Rowe) and a steering committee of eight members from community groups representing minority populations, tribal groups, and urban and rural social service providers, and organizations from medically underserved communities, along with four members from research organizations focused on health policy. This geographically and culturally diverse steering committee represents an attempt to bring the principles of CBPR to the state level (Goold et al. 2016). Members of the steering committee drew on their own experience to help shape the project, and also served as points of contact with a broader set of community groups and individual community members from their constituencies and geographic regions. This allowed the steering committee to draw on a wide range of perspectives, experiences, and expertise when shaping the research project.

Working with this steering committee, we developed a series of categories that captured current and potential areas of Medicaid spending (e.g., coverage for primary care, spending on community health). These categories were designed to reflect categories of spending that were meaningful to both health policy makers and to the communities who would be involved in the research. This process involved the examination of current spending in Michigan’s Medicaid budget and the Medicaid budgets of several other states, interviews with providers and community members, consultation with our actuarial partner Milliman, Inc., and extensive deliberation within the steering committee itself. The result was sixteen categories of spending. Most categories reflect areas of health insurance coverage but also some categories that are not about coverage, such as spending on community health or initiatives to promote quality and equity, since they generally accrue to the same governmental units.

Within each category, we aimed to have a level that reflected current spending in Michigan, and levels that reflected both more and less spending in those areas. For instance, in the Who Is Covered category (eligibility), level 1 reflected pre-expansion eligibility; level 2 reflected current eligibility, and thus spending, for Michigan’s current Medicaid expansion under the ACA; and level 3 added spending to include legal noncitizen residents and adults up to 200% of the federal poverty level. Levels that represented increased or decreased spending were designed in cooperation with our steering committee to ensure they described realistic policy options that addressed the health needs of affected communities. Categories and levels were reviewed by our actuarial partner, who provided estimates of the relative costs of each option and advice on adjusting descriptions of levels in each category to present credible cost estimates for tradeoffs.

The result of this process is displayed in the CHAT board shown in figure 1 and the category and level descriptions shown in table 1. Each category had up to three spending levels that could be selected; the cost of each level is cumulative, so that higher levels (toward the center of the wheel) cost more markers and provide better coverage. The game board had 177 open spaces for markers; however, participants were only given 100 markers, reflecting current spending levels in Michigan.² Participants were given the choice to spend no money, and thus provide no coverage, on most categories. However, three categories (Chronic Conditions, Cost Sharing, and Who Is Covered) required participants to select at least level 1. For Chronic Conditions this was because actuarial estimates suggested that choosing no coverage in this area, rather than level 1 coverage, would actually increase overall costs. For the categories of Cost Sharing and Who Is Covered, we required a choice of at least level 1 spending to reflect the need to have an explicit description of the lowest level of spending (as opposed to no coverage) in those categories.

Table 1.

Categories and Levels in Medicaid CHAT

Category	Definition	Level 1	Level 2	Level 3
Chronic conditions	Extra care for those with chronic conditions. Chronic conditions last at least one year and need a lot of attention. They can be physical (e.g., diabetes). Or, they can be mental (e.g., depression). This may include preventive counseling, case management, community health workers, and lower copays for services that help control the condition.	(1 marker) Co-pays for these conditions are $0. People with chronic conditions that are at risk of going to the hospital have extra help. Services covered may include education, support groups, case management, community health workers, and home visits.	(3 markers) Co-pays for visits, medicines, and supplies are $0. People with any chronic condition can get extra help. Services covered may include education, support groups, case management, community health workers, and home visits.	n/a
Cost sharing	This option sets the amount that people on Medicaid have to pay healthcare. These costs include premiums or monthly payments, co-pays for services, deductibles, and out-of-pocket maximums. The value of services is based on scientific evidence and cost. This option may include discounts for those who try to improve their health.	(1 marker) Prescription copays $1–2. Doctor visits $3. Emergency department $50 (0 if admitted to hospital). Elective hospital stay $50. $0 for preventive services. Some people (e.g., disabled & children) have zero copays. Those over the federal poverty level pay premiums. Maximum total cost is 5% of income.	(3 markers) Same copayments as in 1. No premiums paid by anyone. Maximum total cost is still 5% but hardly anyone reaches it. Those who commit to improving a health behavior (e.g., quit smoking, lose weight, take medicines they need) can earn financial rewards.	n/a
Who is covered	You get to choose who qualifies for Medicaid.	(7 markers) Medicaid covers: children with special needs, children who live in households with incomes less than 200% of the federal poverty level, people below the federal poverty level who are disabled or pregnant, other adults who make less than 33% of the poverty level.	(10 markers) 1+ Adults who make less than 133% of the FPL are covered. Those between 100–133% pay a fee. At most, this will cost $25 a month per adult. People who spend a lot on medical care qualify. If they make more than 133% FPL, they can “spend down” subtracting health care expenses from their income.	(13 markers) 2+ People who make less than 200% of the FPL are covered. Legal immigrants who make less than 133% of the FPL can also get Medicaid.
Connecting to care	This benefit makes it easier for people to get care. It may include: outreach about Medicaid and other state programs, help with transportation, ways to get care without traveling (e.g. telehealth), expanding specialty services, improving communication, and providing useful health information to people.	(1 marker) Transportation to and from needed care is covered. Some money is used to spread information about Medicaid. Telehealth services are covered in some areas. But, you must live at least 50 miles away. Basic cellphone and service provided. Interpreters are available, at least by phone, in all settings.	(2 markers) 1+ Some device or phone-based care is covered. E.g., people can check their blood pressure between visits. They can live less than 50 miles away and use these services anywhere. Info is in many languages. Medicaid gives some resources to improve communication between patients and doctors.	(3 markers) Specialists can travel to remote areas at times. More resources help doctors and patients communicate. More device or phone-based care is covered. This care may include consultations with specialists, psychotherapy, and managing symptoms at the end of life.
Emergencies	This category lets you choose how to cover emergencies. This includes things like ambulances and ER visits. An emergency is an injury or illness that could lead to death or permanent harm. For example, emergencies include broken bones, severe bleeding, and severe and unfamiliar pain.	(3 markers) Unless you could die, you have to call your doctor before you go to the ER. Medicaid covers ambulances for true emergencies. If it is not a true emergency, you pay for it. You may be moved to an urgent care clinic. If you go to the ER for routine care (e.g., a cold) you may have to pay the bill.	(4 markers) Medicaid pays for ambulances. It does not have to be a “true” emergency. If you go to the ER for routine care, you pay the usual copay ($50). If you are not sure that you have an emergency, call your PCP to talk about your problem.	n/a
Medicines and supplies	You can choose to cover medicines, equipment, and supplies. Covered medications must be recommended by a doctor. Equipment can be things like hearing aids, glasses, and wheelchairs. Supplies refer to things like bandages and blood sugar test strips.	(12 markers) Some medicines are covered. So is standard equipment that meets basic medical needs. Hearing aids and eyeglasses are not covered. Over the counter items like aspirin are not covered either. Some supplies are not covered, although special supplies (e.g., syringes) are covered.	(20 markers) More medicines are covered. Expensive medicines may be covered if there is evidence that they work and cheaper meds do not. It is easier to qualify for special equipment like some wheelchairs. Basic hearing aids and eyeglasses are covered. Some over the counter medicines and supplies are covered.	n/a
Mental health	This covers spotting and treating mental illness and substance use. This is only for outpatient care. Inpatient care is covered in “Hospitals.”	(9 markers) You can visit a therapist up to 20 times in one year. But, Medicaid does not pay psychiatrists much. As a result, access to psychiatrists is limited. Services for people with disabilities are covered. But, these services should focus on keeping people out of institutions.	(16 markers) 1+ Long-term counseling is covered if needed. Psychiatrists get paid more and are easier to access. Medicaid covers intensive treatment for substance use. Copays are $0 for those who always show up. Mental health professionals sometimes work in primary care practices.	(25 markers) 2+ More people who have disabilities can get help. Services covered include assisted work, independent living services, vehicle modifications.
Specialty care	You choose how to cover specialty care. This includes: services of doctors, services of other specialists like physical therapists, imaging tests, lab tests, and outpatient surgeries and procedures.	(6 markers) Medicaid does not pay specialists very much. So, it is hard to find specialists that take Medicaid. Medicaid must approve any elective surgery. This includes things like hip or knee replacement. There are limits on the number of times patients can visit therapists.	(14 markers) 1+ Medicaid pays as much as Medicare pays for specialty care. But, they only pay this much for services and providers known to improve health. Medicaid does not need to approve elective surgery. Therapy visits are still limited. But, you can go more often.	(29 markers) 2+ Medicaid pays the same for specialty care as other insurance plans. Medicaid does not need to approve elective surgery. There are no limits on therapy visits.
Dental care	Care is covered for preventing and treating dental problems.	(2 markers) Preventive visits are covered. Teeth can be pulled and cavities filled. Payments to dentists for bridges, crowns and other more complicated services are very low. Dentures are provided if you pay half.	(7 markers) Preventive visits are covered. Teeth can be pulled and cavities filled with easy access to dentists. Payments to dentists for bridges, crowns and other services are about 2/3 what private insurance pays.	n/a
Hospitals	This covers care that people receive when they are admitted to the hospital. Care is provided for both physical and mental health needs.	(15 markers) People who have an emergency do not have to pay for inpatient care. Other admissions cost $50.	(21 markers) Admission to hospitals is covered at no cost.	n/a
Healthy communities	Funds help make communities safe and healthy. Efforts may deal with infectious and chronic diseases, mother and child health, mental health and substance use. This may involve: monitoring, controlling, treating and/or preventing disease; preparing for and responding to emergencies; improving the quality and safety of air, water, and food; and preventing violence and injuries	(1 marker) Medicaid provides money to: inspect buildings, air, and water for safety; track contagious diseases; provide basic care for those without access; reach out about diet, exercise, prenatal health, addiction; prepare for and respond to emergencies; and monitor and report causes of death and disease.	(2 markers) 1+ Additional money for: services that help people with addictions; screening for common diseases; inspecting more places (like homes); coming up with policies to improve health; outreach and helplines for abuse and violence; and providing some mental, physical and dental health care.	(3 markers) 2+ Medicaid helps: build and support spaces for safe exercise; improve access to healthy foods; reach out about prenatal and child health; develop and enforce policies that prevent violence; improve high school graduation rates, community social networks and other social influences on health.
Primary care	You can choose how to cover primary care. People may get primary care from a doctor, physician’s assistant, or nurse. Providers may help identify health problems. They may recommend tests and treatments. Primary care also includes reproductive health and maternity care. A big focus is on preventive health like vaccinations, screenings, counseling about healthy behaviors (like help with quitting tobacco).	(2 makers) Visits are covered, with payment rates less than half what private insurers pay. So, some visits are with a nurse or PA. Payment is even lower for pregnancy and deliveries. Vaccinations and screening tests are free if they are recommended in national guidelines; otherwise you pay for them.	(7 markers) 1+ Visits are covered with payment rates the same as Medicare. Providers can get bonuses for quality care. National rules make sure some vaccinations and tests are covered. Living will (aka advance directive) counseling is encouraged. Health behavior support is also covered.	(10 markers) 2+ Medicaid pays as much as other insurance plans for care. Vaccinations and screening tests are fully covered. But, if they are not recommended by the government, you pay 20%.
Home care	This is coverage for care provided in the patient’s home. Home care can be medical or supportive. Supportive care includes assistance with dressing, bathing, and using the bathroom.	(2 markers) Some necessary care will be covered. Skilled nurses and home health aides may provide some medical care in the home. Equipment, supplies and therapy may also be covered.	(6 markers) 1+ Help with activities of daily living is paid to someone you choose. But, there are limits to how much this person can provide care.	(9 markers) 2+ Help is available for difficult tasks around the house about once a month. Respite care in assisted living or other facilities is available for 2 weeks per year. There are small co-pays for each day of respite care.
Hospice	“Hospice” care is for those with incurable health conditions. Care is provided in addition to normal services, like home care or specialty care.	(2 markers) Hospice services at home increase the number of hours of nursing and help for the family. Attempts to extend life, like surgery, are not covered. Admission to hospitals to control symptoms or treat short-term problems like infections is covered. This level also provides grief support.	(3 markers) 1+ Hospice residences are covered with a daily copay to help cover room and board. Care to relieve patients and some services that may extend a patient’s life are covered. For example, dialysis and simple surgery are covered. ICU care is not provided.	n/a
Facility care	This covers care in a residence. It pays for care in a skilled nursing facility (SNF). It also covers care in rehabilitation facilities. Staff is always available to help.	(9 markers) SNF is covered at low rate for those who qualify. Some places don’t accept Medicaid. In assisted living only medical services (e.g., meds, therapy) are covered. Spouses can keep their homes and some income. After the patient dies, their estate may have to pay the state back.	(19 markers) 1+ SNF care is covered at higher rate for those who qualify. Most SNFs accept Medicaid. Payment for some help with activities of daily living in assisted living is provided. Spouses can keep their homes and some income. After the patient dies, their estate may have to pay the state back.	(21 markers) 2+ Program for all-inclusive care is covered for some patients. They must qualify for a SNF, but be able to live safely in the community.
(E)Quality	Funds help improve the health care system. This affects the quality of health care that people can get. Support is also given to programs that improve the health status of different groups or people. For instance, health difference can be based on socioeconomic status, gender, race or ethnicity, sexual orientation, and access to health care.	(1 marker) Medicaid finds and spreads info about health care quality. It also supports efforts to learn and spread info about differences in health.	(2 markers) 1+ Funds help improve safety and quality of care when and where it is the worst. Funds also to improve health in the groups of people worst-off.	(3 markers) 2+ More funds help improve the safety and quality of care. This includes more effort to improve health in less healthy communities.

The descriptions of levels and categories were designed with the help of the steering committee to be credible and comprehensible to members of the communities who would participate in the research. Final content (which included definitions and explanations of several scientific terms) had a Flesch-Kincaid readability score of 60.6 and was written at an 8th grade reading level. All content was translated into Spanish and Arabic.

Alongside the categories and levels, we worked with the steering committee to develop informational material for participants. The steering committee initially helped identify their community’s informational needs. Background information was drawn from public facing sources (e.g., the Kaiser Family Foundation) that seek to inform the public about Medicaid-related issues. In consultation with the steering committee, we selected material that was understandable to lay members of their communities, that met their informational needs, and that was seen as credible and unbiased. We also worked with the steering committee on how to present the information in a way that was culturally sensitive, respectful, and enabled participants with various backgrounds and educational levels to gain a baseline of knowledge about Medicaid. This information is listed in appendix B.

Sampling and Recruitment

Methods of recruitment for deliberative engagement are contested. Some deliberative practitioners advocate for recruitment strategies that aim for a demographic composition that matches that of the general population (e.g., Knobloch et al. 2013), or even a random sampling approach mirroring that of public opinion polls (Fishkin 2009). However, others worry that formal inclusion by itself will not lead to influence for disadvantaged groups (Sanders 1997; Young 2000). To address this, some argue that recruitment for deliberative forums should focus on representing diverse perspectives instead of mirroring the demographic composition of some underlying population (Davies, Blackstock, and Rauschmayer 2005; Dryzek and Niemeyer 2008). Decisions about Medicaid coverage most directly affect currently Medicaid-eligible residents of communities with large Medicaid-eligible populations,³ and those with particular health challenges such as mental health issues, precisely those whose voices might be lost in deliberation conducted with a demographically representative sample. Thus, we opted for a purposive recruitment strategy that was designed to include the perspectives of those most affected by Medicaid resource allocation decisions.

To this end, participants from low-income, minority, and medically underserved communities were recruited primarily through community-based organizations (e.g., Michigan Center for Urban African American Aging Research, Arab Community Center for Economic and Social Services) around the state as well as flyers posted in public locations, especially libraries and other community centers, a variety of local advertising (newspapers, Craigslist, radio) in English and Spanish, and through our community partners’ personal contacts. The steering committee had significant reach in their respective communities, which enabled them to utilize their relationships and partnerships with organizations and the community to support the recruitment process. Some participants were also recruited through the University of Michigan heath research recruitment website (UMHealthResearch.org). Exclusion criteria included currently working in health care or health research and being under 18 years of age. We aimed to recruit equal numbers of men and women, and to have disproportionately high representation of minority and low-income residents. All participants were paid a $45 honorarium. These recruitment efforts, constrained by our budget and staff capability, yielded a sample of 209 participants.⁴

Study Procedures

Groups were convened in locations familiar to, and convenient for, participants to maximize open and frank dialogue. To ensure representation of non-English-speaking communities, all study materials were translated into Spanish and Arabic, the two non-English languages most commonly spoken in Michigan. Two groups were conducted in Spanish, and three groups were conducted in Arabic. The Institutional Review Board at the University of Michigan deemed this project exempt human subjects research for posing no more than minimal risk to participants (HUM 00090413).

After arriving and being welcomed, participants completed a pre-session survey. To ensure that deliberations would be informed, and that all participants would start with some basic information about Medicaid, the session started with a video that addressed common questions about Medicaid, the Affordable Care Act, and the Medicaid expansion in Michigan (the Healthy Michigan Plan).⁵ Following the video, participants were given 15–20 minutes to review the informational materials about Medicaid (see appendix B). While we did not track what resources participants consulted, facilitators established clear norms that this time should be used to study the informational material (e.g., by asking participants not to do other things like talk or use their phones), and report that nearly all participants actively engaged with the background materials during this time. After this informational material was reviewed, participants were instructed how to play CHAT, and played through the exercise’s four rounds. Sessions lasted roughly three hours; deliberation in round 3 lasted an average of 48 minutes. After round 4, participants completed a postsession survey, were thanked for their participation, and received the $45 gift card as compensation for their time.

The center of the CHAT session was group deliberation in round 3, which was facilitated by a trained facilitator. To ensure the quality of deliberation, facilitators were recruited from the communities participating in the project if they had some previous experience facilitating group discussion. Prior to conducting any sessions, we brought together the facilitators for a day of training in specifically deliberative facilitation. Facilitators were trained to encourage dialogue and reason giving within the deliberative session, with reason giving defined broadly to include things like personal testimony, storytelling, and the like (Sanders 1997; Young 2000). They were also trained to seek, though not force, participation from group members who were less vocal, and to manage participants who threatened to dominate discussion. Finally, because the CHAT session involved setting a spending level for each of 16 categories, facilitators were trained to keep the group on task.

Deliberation in round 3 generally proceeded in the following way. At the start of the round, the facilitator asked for someone to suggest a category to start discussion. The participant who volunteered a category was asked what level of funding they thought was appropriate and why they thought that was the appropriate level of funding. Other participants then chimed in, usually on their own but occasionally prompted by the facilitator, to either support the original suggestion or suggest a different funding level. In either case, if the participant did not elaborate on why they held this opinion, the facilitator would prompt them to explain. Discussion on a category usually concluded when a group consensus was reached, though facilitators actively solicited disagreement with a group’s apparent consensus prior to concluding discussion, and also encouraged those who indicated support nonverbally or with a simple “yes” to provide reasons for their support. In cases where a consensus could not be reached, groups would put the decision to a vote, or agree to return to the category later in the deliberation. Discussion then moved onto a new category, again solicited by the facilitator. Groups could and frequently did revisit previous decisions, particularly as the number of markers dwindled.

Data Collection

We collected two forms of data about deliberators’ preferences for the allocation of resources within Medicaid. First, CHAT software recorded which categories and levels individuals and groups selected in each round. Our analyses focus on the group allocations made in round 3 as a measure of group preferences, and the individual allocations made in round 4 as measures of informed, postdeliberation individual preferences. Ten participants (4.8%) did not select priorities at round 4 and so were excluded from round 4 analyses.

In addition to the allocation of markers in the CHAT game, pre- and post-CHAT surveys measured attitudes and opinions about insurance and Medicaid adapted from existing surveys (Goold et al. 2005; Danis, Ginsburg, and Goold 2006; Kurtovich 2010). These surveys asked participants to choose up to three insurance features they considered to be most important as well as the copay that participants thought should be assigned to a variety of services. These surveys provide a second measure of informed priorities for Medicaid. The pre-survey also asked a series of questions about demographic and health characteristics of participants, and post-surveys included questions evaluating participants’ experience with Medicaid CHAT. Missing data for survey responses ranged from 0% to 7%.

Analysis

We describe the proportion of individuals and groups allocating money to each spending category at the end of deliberation as well as the proportion choosing each funding level. All analyses used Stata 14.2.

Results

This article reports on data collected from 209 participants in 22 deliberating groups from December 2015 to September 2016 across the state of Michigan. Deliberating groups ranged in size from 5 to 16, with most between 8 and 12 participants.

Participant Characteristics

The average age of the 209 participants was 44 years, ranging from 18 to 81, with slightly more women (61.7%) than men. Of the participants, 34.9% identified themselves as white, 25.4% as black/African American, 8.6% as Hispanic, 10.5% as Native American, 15.8% as Arab American, Arab, or Chaldean, and 4.3% not reporting a race or ethnicity. Most participants (80%) had annual incomes less than $35,000, more than half (57%) were under the federal poverty level, and one in five (20%) lived in rural areas. Participants had poorer self-reported health status than the general population, with 6.7% reporting poor health and 26% fair health compared to 3.3% with poor health and 19.7% with fair health in the contemporaneous National Health and Nutrition Examination Survey (CDC 2014), and about two-thirds (64.2%) had at least one chronic condition. Most participants (87.1%) were insured: 40.7% were covered by Medicaid, 25.8% by private insurance, 23.4% by Medicare, and 13.9% by other public insurance providers (e.g., VA, Indian Health Service, county health plans).⁶ Roughly one in five participants (7.7%) covered by Medicaid also reported being covered by Medicare, indicating that they were dual-enrolled. Full participant characteristics are listed in appendix A.

Spending Priorities

We report the priorities of individuals and groups in two ways: First, the categories of coverage that groups and individuals provided at least some funding for; and second, the level of funding provided in each category. We interpret the choice of whether or not to provide any funding for a category as qualitatively different from the decision of how much funding to provide. The former represents a choice about whether, given scarce resources, Medicaid should spend any money in this area; the latter a decision about what level of spending once that choice has been made.

What Should Medicaid Fund?

The left panel of table 2 shows the decisions by groups and individuals about which categories to fund at any level.⁷ All groups allocated at least some funding to Medications and Supplies, and Mental Health. Primary Care, Dental Care, and Hospital Care each had allocations from all but one group, while all but two or three groups allocated funding to Home Care, Hospice Care, and Emergency Care. Roughly three-quarters of groups provided funding for Specialty Care and (E)Quality (quality and equity), while slightly fewer provided coverage for the Connecting to Care category. Only 12 of 22 groups provided funding for Community Health programs.

Table 2.

Group Selection at Round 3 and Individual Selections at Round 4 in Deliberation Arm

Priority	Groupa (%)	Individualb (%)	Group selected; not individual	Group = individual	Group did not select; individual did	Markers assigned to levels	Level selected by groups (%)b N = 22				Level selected by individuals (%)b N = 199
	N = 22	N = 199	N = 199				0	1	2	3	0	1	2	3
Chronic	100	100	0	100	0	1, 3	NA	18.2	81.8	NA	NA	30.2	69.3	NA
Cost	100	100	0	100	0	1, 3	NA	40.9	59.1	NA	NA	37.7	61.8	NA
Covered	100	100	0	100	0	7, 10, 13	NA	13.6	45.5	40.9	NA	28.1	49.7	22.1
Connecting	68.2	78.8	7.6	70.2	22.2	0, 1, 2, 3	31.8	40.9	22.7	4.5	21.1	37.7	22.6	18.1
Emergencies	86.4	88.9	8.1	82.8	9.1	0, 3, 4	13.6	45.5	40.9	NA	11.1	39.2	49.2	NA
Meds	100	91.9	8.1	91.9	0	0, 12, 20	0	22.7	77.3	NA	8.0	35.7	55.8	NA
Mental	100	91.0	9.1	91.0	0	0, 9, 16, 25	0	27.3	63.6	9.1	9.0	38.2	43.2	9.5
Specialty	77.3	88.4	3.0	78.9	18.1	0, 6, 14, 29	22.7	31.8	45.5	0	11.6	53.3	30.7	4.5
Dental	95.5	93.5	6.0	92.0	2.0	0, 2, 7	4.5	45.5	50	NA	6.5	40.7	52.8	NA
Hospitals	95.5	92.0	6.5	89.5	4.0	0, 15, 21	4.5	81.8	13.6	NA	8.0	65.8	26.1	NA
Communities	54.6	75.3	8.1	69.7	22.2	0, 1, 2, 3	45.5	22.7	9.1	22.7	24.6	26.1	28.6	20.1
Primary	95.5	92.5	6.0	89.5	4.5	0, 2, 7, 10	4.5	40.9	50	4.5	7.5	38.7	45.2	8.5
Home Care	90.9	90.5	7.0	86.9	6.0	0, 2, 6, 9	9.1	59.1	27.3	4.5	9.5	47.2	31.2	12.1
Hospice	86.4	88.9	8.5	82.4	9.1	0, 2, 3	13.6	45.5	40.9	NA	11.1	42.7	46.2	NA
Facility Care	68.2	73.9	10.1	65.8	24.1	0, 9, 19, 21	31.8	63.6	4.5	0	26.1	55.8	11.6	6.5
(E)Quality	72.7	77.9	9.6	82.4	8.0	0, 1, 2, 3	27.3	13.6	27.3	31.8	22.1	23.6	29.1	25.1

Abbreviation: NA, is not applicable, where the marker level is not available as a choice.

^aPercentage of groups or individuals choosing each level in that priority.

^bNumber of groups or individuals choosing each level in that priority.

Coverage decisions by individuals after group deliberation largely mirrored these priorities, although there were some notable deviations. While all groups provided at least some funding to Mental Health and for Medications and Supplies, some individuals (9.1% and 8.1%, respectively) declined to provide coverage in these areas. Individuals were more supportive of providing coverage for Specialty Care, with 88.9% of individuals allocating at least some money to this area, as compared to only 77.3% of groups. Individuals were also considerably more likely than groups to provide funding for Connecting to Care (78.8% vs. 68.2%) and Healthy Communities (75.3% vs. 54.6%). Individuals were most likely to disagree with their group’s funding decisions about Healthy Communities and Facility Care coverage; about one-quarter of individuals whose groups did not fund these areas chose, as individuals, to provide funding.

What Level of Funding Should Medicaid Provide?

Beyond the decision about whether to provide funding in a spending category, groups and individuals could indicate how generous they thought funding should be in an area by selecting it at a higher or lower level. The right panel of table 2 shows the percentage of groups choosing each coverage level, while figure 2 displays this graphically.

Figure 2.

Levels of funding selected by groups at round 3, sorted by percentage selecting level 2 or 3.

Among groups, the highest levels of coverage were provided for Chronic Conditions, and Medications and Supplies, where roughly four out of five groups provided the highest possible level of coverage. A plurality or near plurality of groups also provided the highest level of coverage for Dental Care, Emergency Care, Hospice Care, and Cost Sharing. For some categories, most groups favored some coverage, but generally at a middle level, including Mental Health, Primary Care, and Specialty Care. Of particular interest are the categories where most groups favored coverage at the lowest possible level. These included Facility Care, Home Care, and Hospital Care. The categories with the highest variance between groups tended to be those outside of core medical services, including Healthy Communities, (E)Quality, and Connecting to Care.

Again, individual choices of the level of funding largely mirrored the funding levels chosen by groups. We report individual priorities in table 2 and graphically in Figure 3. However, individuals generally favored higher funding in the Connecting to Care and Healthy Communities categories. In contrast, fewer individuals than groups favored high levels of coverage for Mental Health, Chronic Conditions, and Medications and Supplies. The biggest drop in support from individual compared to group deliberation, however, was for the Who Is Covered category. Support for the most expansive eligibility requirements dropped from 40.9% of groups to 22.1% of individuals.

Figure 3.

Levels of funding selected by individuals at round 4, sorted by percentage selecting level 2 or 3.

Survey Measures of Insurance Preferences

In addition to allocating scarce resources to areas of coverage, standard survey items measured participant’s preferences for health care plans, shown in table 3. First, participants were asked to choose up to three insurance features they considered the most important. The top priority, selected by 70% of participants, was having health insurance pay for as many services as possible. This was followed by having small copayments, selected by 50% of participants. Participants placed a lower priority on features related to the breadth of the provider network as well as on being treated with respect by health care providers.

Table 3.

Importance of Insurance Features

Of the following, which are the most important to you (choose up to three)
Having health insurance pay for as many different services as possible	.70
Having very small (or no) co-payments for doctor visits and medicine	.50
Being able to get a doctor appointment immediately	.29
Having a good selection of primary care doctors to choose from	.29
Being treated with respect by my healthcare providers	.29
Having a good selection of specialties to choose from	.18
Having a choice of which hospital I go to	.22
Having doctors available who are close to where I live	.22

Note: When calculating the percentages, those who did not participate in the survey were excluded from the denominator.

We also asked participants about copayments for specific services. Participants rated a range of medical services on a 1 to 4 scale where 1 indicated they thought there should be no copayment for the service and 4 that they thought there should be a high copayment for the service. Full results are shown in table 4. Participants were least accepting of copayments for vaccines, basic treatments for chronic conditions, primary care visits, and ER and hospitalization for urgent problems. They were most accepting of high co-payments for services that were presented as optional or nonurgent, such as brand name medicines when a generic is available or ER visits for nonurgent problems.

Table 4.

Assigning Co-payments in the Post-Survey^b

	Mean (SD)	Standardized item random effecta	p a
Vaccines (measles, flu, etc.)	1.4 (.7)	−9.41	<.001
Hip replacement surgery	2.2 (.9)	5.65	<.001
Admission to a hospital for pneumonia	1.7 (.8)	−4.51	<.001
Inhalers for asthma	1.5 (.7)	−7.50	<.001
Medicine for heartburn	2.0 (1.0)	1.90	.06
Generic medicine	1.8 (.9)	−4.23	<.001
Brand name medicine when a generic is available	2.7 (1.0)	11.66	<.001
Brand name medicine when there is no generic available	2.1 (.8)	1.22	.22
Primary care visits	1.7 (.8)	−5.53	<.001
Specialist care visits	2.0 (1.0)	.98	.33
ER visit for a non-urgent problem	2.5 (1.1)	9.93	<.001
ER visit for an urgent problem	1.6 (.8)	−6.09	<.001
Stop-smoking services	1.9 (.9)	−.59	.55
Certain scans (such as MRI or CAT scans) that guidelines say aren’t necessary	2.3 (1.1)	6.41	<.001
Equipment for home testing (e.g., blood pressure, blood sugar)	1.7 (.8)	−4.76	<.001
Motorized wheelchair	2.1 (.9)	4.37	<.001

^aStandardized random co-payment service intercepts from fitting two-way crossed-effects model to copayment level responses for all 16 co-payment services from all participants. The crossed-effects are CHAT groups and the 16 copayment services. A positive (or negative) value means higher (or lower) copayment levels for that service relative to average level across all 16 different services. The corresponding p-value tests for the significance of each service’s random intercepts, that is, whether the level of support for a copayment is significantly higher or lower, relative to overall mean across all 16 services.

^bQuestion text: “Copayments (also called copays) are what you pay yourself for a specific health care service, for example a visit, prescription, or procedure. . . . Please check the co-pay you think each service should have. You need to make sure there are some in the Medium and High columns; not everything can be $0!” Responses can range from 1 = no copay to 4 = high.

Participants’ Evaluation of Medicaid CHAT

The democratic legitimacy provided by deliberative engagement depends in part on the quality of deliberation among participants. While a full examination of the quality of discussion in these sessions is beyond the scope of this article, other recent work has shown that CHAT sessions generally succeed in meeting standards to be considered deliberative (Goold et al. 2005; Goold et al. 2019). Here, we report one measure of deliberative quality: participants’ perceptions of the quality of discussion during deliberation, the quality of information and choices offered in the exercise, and whether participants supported using the results of their group’s deliberation in the policy-making process. These were measured with twenty questions in the post-deliberation survey; table 5 shows mean responses to these items, which we grouped based on a factor analysis of the same battery of questions administered in a different deliberative exercise (Goold et al. 2019).

Table 5.

Perceptions of Deliberative Quality

How much do you agree or disagree with the following statements? (5 = strongly agree; 1 = strongly disagree; reverse coded indicated with [-])	Mean (SD)
Sufficient Information and Choices	3.9
Information given to us was believable.	4.4 (0.1)
Choices offered in the exercise were realistic.	4.3 (0.9)
Choices in the exercise included all the choices I could have wanted.	3.8 (1.2)
We did not have enough information to make good decisions. [-]	2.7 (1.5)
Views of Deliberation	3.9
A few people dominated the discussions. [-]	3.0 (1.4)
The way in which the group reached its decision was not fair. [-]	2.3 (1.4)
The discussions were superficial. [-]	2.4 (1.4)
There was too little time to discuss. [-]	2.7 (1.4)
People in the group argued by referring to what would be best for themselves. [-]	3.2 (1.3)
Our discussion included responding to each other’s arguments.	4.0 (1.1)
I gained an understanding of arguments that opposed my own.	4.1 (1.0)
My views were considered and taken into account.	4.3 (1.0)
I had lots of chances to share my views.	4.4 (0.9)
The participants in the group argued by referring to what would be best and most fair for all people.	3.9 (1.2)
Discussion during the game was open and honest.	4.6 (0.8)
All positions in the group were considered with equal respect.	4.4 (0.8)
The arguments of other participants were useful in forming my own position.	4.2 (1.0)
During the exercise, I was treated with respect.	4.5 (0.8)
Support for using results	4.1
I would support using our group’s discussion to inform decision makers.	4.2 (0.9)
I would trust a process like this to inform funding decisions.	4.1 (1.1)

Participants reported generally positive views of all aspects of deliberation. The average ratings for the battery measuring the quality of information and choices and the battery measuring discussion quality were both 3.9 out of 5 (1 = strongly disagree, 5 = strongly agree). Ratings were the highest for items related to the openness of discussion and mutual respect shown by participants for one another. Participants also strongly supported using the results of the deliberation in policy making, with an average response of 4.1 on the two items measuring whether the results of the Medicaid CHAT exercise should inform future decision making.

Discussion and Conclusion

The ACA, and subsequent efforts to limit or repeal the act, created one of the most turbulent policy-making environments in Medicaid’s history (Smith and Moore 2015; Grogan and Park 2017). Changes to Medicaid proposed by state and federal lawmakers include changes to eligibility, scope of services, cost sharing, and other features as well as sweeping changes to how the federal government shares in the cost of Medicaid through block grants. Decisions like these, which involve tradeoffs between competing needs for shared, limited resources, could be substantially improved by incorporating the insights, experiences, priorities, and concerns of those most affected. While there is broad recognition that such decisions are improved by including the voices of those most affected (Goold 1996; Fleck 2009; Solomon, Gusmano, and Maschke 2016), attempts to do so through broad public engagement efforts or the inclusion of advocates for those most affected often fail to accurately represent these voices (Goold 1996; Grogan and Gusmano 2007). Medicaid CHAT provides a flexible tool that allows public officials, advocates, and academics to gather the views of communities most affected by Medicaid prioritization decisions and incorporate them into policy making and public debate.

Our results show how low-income and medically underserved communities in one state would allocate Medicaid funding. Their choices reflect a unique set of priorities. Nearly all groups and three-quarters of individuals elected to cover at least as many people as under the ACA Medicaid expansion, with a sizable minority choosing to provide coverage to additional groups. They also chose to cover a breadth of services, including some, like dental services and medications and supplies, that are often the first to be cut in funding crises (Singhal 2015). To pay for this breadth of services, participants were less likely to fund areas that were more collectively focused, such as improving general health care quality and equity, supporting community health, and making care more accessible. Many also selected lower levels of funding for particularly expensive areas of coverage, including Facility Care, Specialty Care, and Home Care.⁸ Finally, selections in CHAT and survey results showed that participants were willing to accept some cost sharing and limits on access to achieve other priorities.

Medicaid CHAT is the first project to specifically frame CHAT deliberation around a state’s full Medicaid program. However, these choices are like those made by other low-income populations using CHAT to select basic benefits packages. For example, uninsured Ohioans selected a breadth of coverage and some cost sharing (Ohio Department of Insurance 2009). Disabled Medicaid beneficiaries in California chose to include a range of services, even if that meant restrictions on access or higher out-of-pocket costs for some (Danis et al. 2006). Participants in CHAT exercises in a Texas program for the uninsured selected fewer types of services, though with a more constrained budget (Danis, Ginsburg, and Goold 2010).

How do these priorities compare to how resources are currently allocated? The strong priority given to eligibility fits generally with the decision of Congress, Michigan’s government, and a majority of states to expand Medicaid. High priority for medications, supplies, and chronic conditions also validates current policies in Michigan and many other Medicaid programs (Gifford et al. 2017). The high priority given to mental health services by our participants parallels the emphasis, in many communities, on improving access to and quality of mental and behavioral health care (Crane-Ross, Roth, and Lauber 2000), and states’ expansions of such coverage, although it stands in contrast to the greater difficulty Medicaid recipients have accessing mental and behavioral health care (Antonisse et al. 2018; Atiga et al. 2016; Goold et al. 2018).

Importantly, these priorities should be interpreted as conditional on current levels of Medicaid funding. Choosing priorities for the use of limited resources fundamentally changes when resources become more—or less—limited. An economic recession may constrain revenues and increase demand for Medicaid, while economic growth can increase state revenues and decrease demand. Resources can also be expanded or constrained by changes in taxation or spending in other areas (e.g., education, law enforcement), or decisions at the federal level to change how Medicaid funds are allocated (e.g., Valentine et al. 2017). Medicaid CHAT could easily be adapted to allow participants to increase or decrease the pool of markers available, simulating these broader policy decisions (see, e.g., Goold et al. 2004). We chose to structure the current study without this option because we wanted to focus discussion on tradeoffs under current funding levels, thus assuring useful, actionable guidance to officials working in the current fiscal structure. However, future work should examine the effects of giving deliberators the option of increasing or decreasing overall Medicaid funding.⁹

Future work may also build on this research by varying other elements of the research design, such as the recruitment strategies, the length of deliberation, and the number of participants. Recruitment that aims for demographic representativeness, like those employed in Deliberative Polls^® or Citizen Initiative Review processes, runs the risk of minimizing the voice of those most affected by Medicaid allocation decisions. However, these recruitment strategies might increase the degree to which state legislators or members of the general public perceive the results as democratically legitimate (Fishkin 2013; Warren and Gastil 2015; Gastil et al. 2016). Furthermore, the difference between how those in poor, minority, and underserved communities would prioritize Medicaid spending and how the public at large would do so is itself an interesting research question. Similarly, while the length of our Medicaid CHAT sessions was designed to make them accessible to a wide range of possible participants, a longer deliberative process might allow for more intensive deliberation or in-depth education prior to deliberation. Finally, while our sample of 209 participants represented a range of perspectives, a larger sample might allow for an even broader range, a more reliable measure of the preferences of these communities, and might be perceived as more democratically legitimate. Notably, any of these variations involve tradeoffs, such as between the length of deliberation, the number of people engaged, and the amount of resources expended on engagement efforts.

As states seek innovative approaches to optimize the clinical and fiscal outcomes related to their Medicaid programs, state and federal policy makers have much to gain from the insights of those most affected by these decisions. Current CMS policy appears aimed at granting states increased flexibility in the design of their Medicaid programs through, for example, new 1332 waiver concepts intended to foster “innovative approaches to meet the unique needs of [a state’s] population” (CMS 2018; Keith and Fleming 2018). Determining these unique needs and how to meet them without actively engaging those most affected by any policy change would be a grave mistake. These populations, after all, have great experience living within severely constrained resources, and on-the-ground knowledge and experience that can inform judgments about what spending provides the most value for low-income persons. The examples presented in this article suggest that a deliberative exercise like Medicaid CHAT could play an important role in a fluid policy-making environment.

The relatively short duration and abstract nature of prioritization in Medicaid CHAT means that it is not a substitute for existing policy-making structures as, for example, deliberation among non-elites substitutes for bureaucratic budgeting decisions in participatory budgeting processes (Wampler 2007). Instead, Medicaid CHAT can serve as an input to the policy-making process that introduces the voices of affected communities as a counterbalance to the expert and elite voices that generally dominate these decisions. This could happen in a number of ways. Like this study, Medicaid CHAT can provide a snapshot of priorities at one point in time to inform major policy changes. However, Medicaid CHAT could also be incorporated as a regular part of a state’s annual Medicaid budgeting process. This would provide an ongoing, institutionalized voice for affected communities as states react to changes in fiscal condition, federal rules, and the needs of Medicaid eligible populations. Medicaid CHAT could be also be useful for nongovernmental organizations that seek to increase the influence of affected communities over Medicaid policy, much as such groups currently use polling to advocate for policy change (Greenwald et al. 2003). While their cost and impact would vary, any of these uses of Medicaid CHAT would mark an improvement on current policy making by introducing the knowledge and experiences of those most affected to the policy-making process.

Biographies

C. Daniel Myers is an assistant professor of political science at the University of Minnesota. He received his PhD in politics from Princeton University and completed postdoctoral training at the University of Michigan as part of the Robert Wood Johnson Scholars in Health Policy Program. His research lies at the intersection of political psychology, political communication, and deliberative democratic theory. His work has appeared in the Journal of Politics, Political Analysis, Political Behavior; the Journal of Public Deliberation; the Journal of Theoretical Politics; Milbank Quarterly; Health Service Research; Health Expectations; and the Oxford Handbook of Political Psychology. cdmyers@umn.edu

Edith C. Kieffer is a professor of social work at the University of Michigan. She conducts community-based participatory research addressing health and health care disparities in community and health care settings, including randomized clinical trials demonstrating the effectiveness of community health workers in improving diet and exercise behaviors, depressive symptoms, glucose control, and other outcomes among Latina/os in Detroit. She is a member of the Institute for Healthcare Policy and Innovation team that is evaluating Michigan’s Medicaid expansion, leading qualitative interviews exploring the experiences of beneficiaries and providers and the impact of Medicaid coverage on health care use, beneficiary health, and other outcomes.

Mark Fendrick is a professor of internal medicine and a professor of health management and policy at the University of Michigan. He conceptualized and coined the term Value-Based Insurance Design (V-BID) and currently directs the V-BID Center at the University of Michigan (www.vbidcenter.org), the leading advocate for development, implementation, and evaluation of innovative health benefit plans. His research focuses on how clinician payment and consumer engagement initiatives impact access to care, quality of care, health care disparities, and health care costs. He remains clinically active in the practice of general internal medicine.

Hyungjin Myra Kim is a biostatistician and research scientist at the University of Michigan’s Center for Consulting for Statistics, Computing, and Analytics Research (CSCAR) and an adjunct professor of biostatistics. Her work has focused on applying advanced statistical methods to assess health outcomes. She is currently a co-investigator and principal statistical investigator on NIH, Department of Defense and Veterans Health Administration–sponsored multicenter randomized clinical trials, cluster randomized trials, and observational cohort studies. Her research interests are in longitudinal data analyses, health services research, and pharmaco-epidemiology, with additional interest in mental health treatment medication use in health systems.

Karen Calhoun is a community engaged research program officer for the Michigan Institute for Clinical and Health Research at the University of Michigan. This includes diverse interpretations of lived experiences in her work collaboratively with other partners to improve community and population health with a National Institutes of Health Clinical and Translational Science Award–funded consortium. She self-identifies as a community scientist and community partner, is a senior associate editor with Progress in Community Health Partnerships, and participates in a leadership capacity on other efforts locally and nationally. For the majority of her career, she has used community planning/problem solving and community engagement to address health disparities.

Lisa Szymecko is an evaluation specialist with the System exChange at Michigan State University. She has expertise in multiple research methodologies, including community-based participatory research and qualitative data collection and analyses. She has participated in large interdisciplinary and community teams to evaluate health policy. She has worked extensively in efforts to engage the voice and the perspective of vulnerable populations related to the local conditions perpetuating poor health conditions. She has been published in the Journal of General Internal Medicine, Milbank Quarterly, and Progress in Community Health Partnerships: Research, Education, and Action.

Lynnette LaHahnn is currently a naturopathic practitioner after serving in both the public health and mental health fields. She has experience as a health educator and administrator and was heavily engaged in community-based assessment of priority health needs, building rural health networks as well as the strategic financial and program planning and evaluation for addressing those identified needs. Most recently she worked to build consumer lead recovery and community inclusion programming for the sustainability of community-based mental health support. After two decades of providing leadership to consumers of health and human services, the most meaningful role currently is advocating for consumers.

Charo Ledón is a Latino community leader, advocate, and certified legal interpreter. Having lived in the Ann Arbor/Ypsilanti area since 1974, she has seen the community grow and diversify. She has owned and operated several businesses in the area and has had the opportunity to serve the community in a variety of settings for three decades. She is particularly well acquainted with the challenges, struggles, and strengths of Latinos. She currently works with a group of Latino community members calling themselves Buenos Vecinos to promote resilient Latinx leadership and powerful connections across Washtenaw County by igniting ideas, strengthening trust, and building relationships.

Marion Danis holds BA and MD degrees from the University of Chicago and trained in internal medicine at the University of North Carolina at Chapel Hill. She is head of the Section on Ethics and Health Policy in the Department of Bioethics and chief of the Ethics Consultation Service at the Clinical Center of the National Institutes of Health. Her research focuses on reducing health disparities by addressing the social determinants of health and finding strategies for the fair rationing of limited health care resources by involvement of the public.

Zachary Rowe is the executive director of Friends of Parkside, a grassroots community-based organization on Detroit’s East Side, which provides supportive services and programs to residents of the Villages at Parkside, a public housing complex, and the surrounding community. He has been involved with various Community-Based Participatory Research (CBPR) projects for more than 24 years and is a founding member of the Detroit Urban Research Center (URC) Board. He has served as coprincipal investigator on NIH and PCORI funded projects.

Susan Dorr Goold is a professor of internal medicine and health management and policy at the University of Michigan. She studies the allocation of scarce health care resources, especially the perspectives of patients and citizens. Results from projects using the award-winning CHAT (Choosing Healthplans All Together) allocation game have been published and presented in national and international venues. She serves on several editorial boards as well as the AMA Council on Ethical and Judicial Affairs and has held leadership positions in the American Society for Bioethics and Humanities and the International Society on Healthcare Priority Setting.

Appendix A: Participant Characteristics

Participant characteristics were measured in presession surveys and are reported in appendix table A1. Appendix table A2 reports associations between participant characteristics and individuals’ round 4 priorities. In general, we found few strong relationships between participant characteristics and priorities selections.

Appendix Table A1.

Participant Characteristics

Participant characteristics	Number (%)
Female (n = 206)	127 (61.7)
Age in years (n = 204), mean (SD, range)	43.8 (14.8, 18–81)
Age, categorical
<30	46 (22.0)
30–44	55 (26.3)
45–54	50 (23.9)
55–64	39 (18.7)
>65	14 (6.7)
Self-identified race/ethnicity (n = 209)
White	73 (34.9)
Black/African American	53 (25.4)
Arab/Arab American/Chaldean	33 (15.8)
Hispanic	18 (8.6)
Native American	22 (10.5)
Any other or self-identified as other	1 (.5)
Missing	9 (4.3)
Education (n = 206)
High school/GED or less	87 (42.2)
Some college	66 (32.0)
Bachelor’s degree or higher	53 (25.7)
Income (n = 199)
Less than $15,000	110 (55.3)
$15,000–$34,999	49 (24.6)
$35,000 or more	40 (20.1)
No. of people in household (n = 197), mean (SD; range)	3.0 (2.0, 1–12)
At or below federal poverty level (n = 194)	110 (56.7)
Living alone (n = 197)	45 (22.8)
Perceived health status (n = 200)
Poor	14 (6.7)
Fair	52 (26.0)
Good	64 (32.0)
Very good or excellent	71 (35.5)
Work or worked in health care (n = 202)	68 (33.7)
Currently covered by health insurance (n = 201)	175 (87.1)
No insurance	26 (12.9)
Other insurance	92 (45.8)
Medicaid	83 (41.3)
Comorbid illness
Hypertension (n = 201)	67 (33.3)
Asthma (n = 200)	30 (15.0)
Chronic bronchitis, COPD, or emphysema (n = 200)	15 (7.5)
A substance use disorder (n = 199)	15 (7.5)
Arthritis or rheumatism (n = 199)	54 (27.1)
Cancer other than skin cancer (n = 200)	15 (7.5)
Diabetes (other than during pregnancy) (n = 201)	29 (14.4)
Digestive problems (n = 200)	42 (21.0)
A heart condition or heart disease (n = 200)	23 (11.5)
HIV or AIDS (n = 201)	0 (0)
Kidney disease (n = 199)	11 (5.5)
Liver problems (e.g., cirrhosis) (n = 200)	7 (3.5)
Stroke (n = 200)	5 (2.5)
Total number of comorbid illnesses,a mean (SD)	1.6 (1.7)
At least one comorbid illness (n = 201)	129 (64.2)
Urbanity (n = 194)
Urban	140 (72.2)
Suburban	16 (8.2)
Rural	38 (19.6)

Note: Cell values are n (%) unless otherwise described; n does not add to the total when some responses are missing.

^aOf participants who checked at least one comorbid condition as yes or no.

Appendix Table A2.

Participant Characteristics Predictive of Round 4 Mean Marker Investment Level for Each Priority (N = 199)

Cell values are adjusted parameter estimates associated with each predictor for each priority.
				Race/ethnicity [ref = white]						Health status	Insurance [ref = no insurance]				≥1
	Age	Age2	Male	Black	Arab	Hispanic	Native	Other or missing	3-level education	Good or better	Medicaid	Other insurance	Suburban or urban	FPL ≥30%	Comorbid illness
Chronic
Cost				0.48†					−0.17
Covered					−1.65‡			2.30‡	0.70‡
Connecting						0.68†
Emergencies										0.64‡				0.98
Meds				−2.55†		−3.82†			1.05
Mental
Specialty				−2.71							2.85	3.11†			3.59‡
Dental										−1.01†					−1.43†
Hospitals
Communities	−0.002	−0.001‡				0.96‡									−0.34†
Primary	−0.01	−0.002		1.34†						1.21†
Home care	0.03						−1.64†						−1.91‡
Hospice				0.35
Facility care								−4.43
(E)Quality	−0.01†	−0.0008†												1.25‡

Note: Cell values correspond to beta coefficients associated with each predictor from fitting separate mixed-effects regression models with marker values selected at round 4 for each priority as response variables. Only the beta coefficients from predictors with p < .10 are shown, except when age-squared term is significant; beta coefficient from age is also shown regardless of significance. For each priority, only 2, 3, or 4 levels are available for marker values, but mixed-effects regression models were fit so that a beta estimate reflects the expected marker value change associated with the predictor, adjusting for covariates. A positive beta coefficient can be interpreted as higher investment associated with the predictor. Each model included chat group as random intercepts to account for potential correlation within chat group. For consistency, each model included the same predictors of age (centered at 45 years) and age squared, male (yes/no), race (defined as White, African American, Hispanic, Arab American/Chaldean, Native American, others or missing), 3-level education (0=high school or less; 1 = some college; 2 = college or more), 3-level medical insurance status (no coverage, Medicaid, or other insurance), having at least one comorbid condition (of hypertension; asthma, chronic bronchitis, COPD, or emphysema; a substance use disorder; arthritis or rheumatism; cancer other than skin cancer; diabetes or sugar diabetes other than during pregnancy; digestive problems such as ulcer, colitis, or gall bladder disease; a heart condition or heart disease; HP or AIDS; kidney disease; liver problems such as cirrhosis, or a stroke), self-perceived health status (good or better [vs. fair or poor]), rurality (suburban or urban [vs. rural]), and FPL (federal poverty level) (33% or higher [vs. not]).

^†p < 0.05

^‡p < .01

Appendix B: Educational Materials

We provided participants with links to already existing websites and media about Medicaid. At focus group locations without internet access, materials were printed and given to participants in paper copy. Text from websites was translated into Spanish or Arabic when not available and provided as a paper copy. Consulted websites and media are listed below.

MNT Editorial Team. What is Medicare/Medicaid? Medical News Today website. January 5, 2016. www.medicalnewstoday.com/info/medicare-medicaid#medicaid. Archived at web.archive.org/web/20191109122130/https://mnt-prod.medicalnewstoday.com/articles/323858.php

Medicare vs. Medicaid. Medicare Made Clear. www.medicaremadeclear.com/basics/medicare-vs-medicaid. Archived at web.archive.org/web/20170203032336/https://www.medicaremadeclear.com/basics/medicare-vs-medicaid.

Eligibility. Washington, DC: Centers for Medicare and Medicaid Services. www.medicaid.gov/chip/eligibility-standards/index.html. Archived at web.archive.org/web/20170612200743/https://www.medicaid.gov/chip/eligibility-standards/index.html.

Federal Poverty Level (FPL) Guidelines. ABC for Health, Inc. January 13, 2018. www.safetyweb.org/fpl.php. Archived at https://web.archive.org/web/20180118032130/http://www.safetyweb.org:80/fpl.php.

Wikipedia contributors. Medicaid. Wikipedia, The Free Encyclopedia. April 17, 2018. en.wikipedia.org/wiki/Medicaid. Archived at web.archive.org/web/20180419053235/http://en.wikipedia.org/wiki/Medicaid.

Medicaid: Health Care for People with Low Incomes. WebMD. November 10, 2016. www.webmd.com/health-insurance/medicaid-health-care-people-low-incomes. Archived at web.archive.org/web/20161114023825/http://www.webmd.com/health-insurance/medicaid-health-care-people-low-incomes.

Do I Really Need Insurance? WebMD. img.webmd.com/dtmcms/live/webmd/consumer_assets/site_images/aca/healthcare_infographic.jpg. Archived at web.archive.org/web/20161203174723/http://img.webmd.com/dtmcms/live/webmd/consumer_assets/site_images/aca/healthcare_infographic.jpg.

Quiz: Do You Know Your Health Insurance Basics? WebMD. July 25, 2016. www.webmd.com/a-to-z-guides/condition-15/insurance/rm-quiz-insurance-basics. Archived at web.archive.org/web/20160805063103/www.webmd.com/a-to-z-guides/condition-15/insurance/rm-quiz-insurance-basics.

Kffhealthreform. Health Reform Explained video: “Health Reform Hits Main Street.” YouTube. September 17, 2010. www.youtube.com/watch?v=3–Ilc5xK2_E. Also available in Spanish at www.youtube.com/watch?v=FAf2H3pFtzs.

Kaiser Family Foundation. Health insurance explained—The YouToons have it covered. November 11, 2014. www.youtube.com/watch?v=−58VD3z7ZiQ&list=PLBfqBziFkXIfmUHxQ0teDRppDa-VctX7c&index=5. Also available in Spanish at youtu.be/mDPhCo11z0E.

Health Michigan Plan. Lansing, MI: Michigan Department of Health and Human Services. www.michigan.gov/healthymiplan. Archived at https://web.archive.org/web/20160601000000*/www.michigan.gov/healthymiplan.

Policy Basics: Where Do Our State Tax Dollars Go? Washington DC: Center on Budget and Policy Priorities. April 24, 2017. www.cbpp.org/cms/index.cfm?fa=view&id=2783. Archived at https://web.archive.org/web/20160313081250/www.cbpp.org/cms/index.cfm?fa=view&id=2783.

In addition to these materials, we translated our “What Is Medicaid?” video into Spanish and Arabic.