5

Contextual Factors: Policy Influences on Morbidity and Disability

OVERVIEW

The presenters assembled for this session were chosen for their expertise in addressing questions such as the following:

• What is the impact of the changing policy landscape (including public health and safety net programs such as Medicare, Medicaid, and Social Security Administration [SSA]/Social Security Disability

• Insurance [SSDI]) on the disabling effects of chronic illness and injuries?
• How does disability prevalence vary across regions and states within the United States?
• How can natural experiments, such as those created by state-by-state variation in policies or environment, contribute to our understanding of the causal relationship between contextual and structural factors on disability severity and benefit program enrollment?
• How do changes in the supply and type of housing and other aspects of the built environment affect disability prevalence?

Jennifer Karas Montez (Syracuse University and committee member) moderated the session. Philippa J. Clarke (University of Michigan) addressed the role of community and housing environments for measures designed to improve functioning with aging. Ellen Meara (Harvard University) presented her research on trends and consequences of structural barriers, including those related to health insurance, to disability. Emma Zang (Yale University) discussed the influences of state policy contexts on cognitive functioning and impairment.

THE ROLE OF COMMUNITY AND HOUSING ENVIRONMENTS FOR OPTIMIZING FUNCTION WITH AGING

Philippa J. Clarke (University of Michigan) began her remarks by defining the built environment as the features in the world that are created or modified by people. Examples of these features are steps to climb, elevators, automatic doors, distance to a public transportation stop, the number of lanes of traffic, and the length of time provided by traffic signals to cross a street.

Clarke described the optimizing function among people who have disability—that is, who have limitations in physical, sensory, or cognitive function—in terms of how the environment either limits or facilitates their ability to participate in society and maintain health. People with disability face a range of disparities regarding their health; they are twice as likely as the general population to be current smokers and 10 percent more likely to be obese. Cardiometabolic conditions are also much higher in people with disability, with the prevalence of diabetes twice as high. Clarke explained that health problems affect a person’s ability to engage in society, including with work, and can have adverse consequences for stable employment trajectories, income, and wealth accumulation. These disadvantages can be particularly consequential during emerging adulthood (ages 20 to the late 30s), a critical time window for attaining advanced education, starting a

career, getting married and forming a family, and acquiring assets. From a perspective of cumulative disadvantage, the progression of health problems over the life course is also a cyclical reinforcement of socioeconomic disadvantage.

Clarke also outlined her use of the Panel Study of Income Dynamics (PSID) to follow a cohort ages 20–34 (in 1979) through at least age 64. She and colleagues defined work disability as applying to people who, during these ages, reported at least four times that they were unable to work due to a physical, sensory, or cognitive limitation. While the difference in household incomes between those with and without disability (Figure 5-1) is not that great at the youngest ages, it begins to diverge with households headed by individuals in their late 20s and increases with age all the way through the prime working years. In terms of health, people who were defined as aging with disability were more likely to be sedentary and obese, and to have a host of chronic health problems including heart disease, high blood pressure, stroke, diabetes, arthritis, cancer, and psychiatric problems.

One of Clarke’s main points is that these conditions are not inevitable, but are affected by the physical, social, and attitudinal environment. More than half of adults with disability do not engage in physical activity, not because they do not want to but because they face equipment barriers in recreation facilities, negative attitudes in both the community and those

facilities, and inaccessible walking paths and sidewalks that limit their ability simply to get around (Rimmer et al., 2004). Adults with disability are also more likely than the general population to delay or forgo medical care due to physical barriers either leading up to a health facility or within it or due to attitudinal barriers in that facility.

The International Classification of Functioning, Disability, and Health (within the World Health Organization) classifies function at three levels: the body (impairment), the person (activity), and the society (participation). Clarke and colleagues (Khan et al., 2024) recently completed research using medical claims data to examine the density of neighborhood resources and the incidence of cardiometabolic disease in adults aging with physical disability. The data used in the analysis, provided by Optum Clinformatics Data Mart, include 15,467 people who were diagnosed with cerebral palsy or spina bifida. The outcome tracked was incident cardiometabolic disease, identified through International Classification of Diseases codes (ICD-9 and ICD-10) in the claims data, and the exposure was neighborhood context. To acquire access to individual ZIP codes and remain within confidentiality guidelines, Clarke excluded data on race and ethnicity, instead using crosswalks from the National Neighborhood Data Archive (NaNDA) to link neighborhood characteristics to individual ZIP code of residence.

Among the group in the sample with physically disabling conditions, over a three-year period 40 percent experienced an incident cardiometabolic diagnosis. The diagnosis was more likely to be hypertension, but hyperlipidemia and type 2 diabetes were also common. Clarke explained that living in an area with more opportunities for mobility and physical activity—whether parks, recreation facilities, or access to public transit—was associated with an almost 15 percent lower risk of incident cardiometabolic disease. The number of health care establishments was also protective, though somewhat less so. Broadband internet was a strong predictor as well, associated with greater than a 10 percent reduced risk of cardiometabolic disease. One element of this effect is that access to broadband provides access to health care. Telehealth, specifically, was the primary way that health care was delivered through the pandemic, and it is a critical resource for people with disability or mobility limitations who cannot physically access health care. Although a common perception is that broadband is less accessible in rural areas, Clarke’s team found it is also not equally distributed within urban areas.

Clarke also presented results from her research on the relationship between vision impairment and type 2 diabetes. Access to grocery stores, broadband internet, recreational parks, and optical goods stores was important for reduced rates of diabetes, but living in an area with more high-speed roads and greater intersection density was associated with increased risk. Since heavy traffic is the most common environmental factor associ-

ated with injuries and fatalities among disabled pedestrians, navigating the local environment can become more hazardous for people with vision impairment, especially as they age and develop more mobility limitations.

Clarke next outlined a study of adults with spinal cord injuries (Tan et al., 2023) that examined which environmental features limited their access to various community buildings or sites. The greatest barriers were the physical structure of buildings and limitations in parking. Nearly everyone in the study reported a need for more van-accessible spots that were wide enough to accommodate access into and out of vehicles with ramps. The study by Tan and colleagues examined features in the built environment not only to determine which were barriers but also to identify which were facilitators. It found that curb cuts, ramps, and automatic doors generally were helpful, while gravel surfaces and crowds generally were barriers. It also found variations: for example, 8 percent of adults with spinal cord injuries said that curb cuts limited their ability to get around. Barriers in the home environment can also create challenges, with steps at entry especially problematic. Economic factors are important in determining whether people can make the modifications they need to navigate their environments.

Clarke suggested that an immediate need for adequately studying the role of the built environment in people’s wellbeing is a greatly improved disability data infrastructure. Non-survey data are likely to become an increasingly valuable component of this infrastructure. The NaNDA data, noted above, are publicly available and shared, but they do not capture accessibility. Clarke shared some work by Erica Twardzik, a disability researcher at Johns Hopkins Bloomberg School of Public Health, who has created a Public Transit Disability Dashboard on the accessibility of 100 transit systems in the United States. Google Street View can be helpful in identifying features in the built environment. The Radical Access Mapping Project has audit instruments to conduct accessibility assessments of street view images, but these assessments are very time consuming. Artificial intelligence may be useful for using high-resolution aerial imagery to identify and quantify intersections, sidewalks, curb cuts, and crosswalks.

During open discussion at the end of the session, steering committee member Nicole Maestas raised the question of why estimates of employment rates among people with disabilities vary so greatly, with reported estimates based on cross-sectional snapshots from American Community Survey (ACS) and Current Population Survey being especially low (around 20%). Clarke responded that there is no good measure of disability in the PSID, so instead a crude measure was used based on not being able to work for four or more years due to a health problem. She thought that was probably the source of variation between these estimates and those based on more precise definitions.

STRUCTURAL BARRIERS TO PUBLIC DISABILITY AND HEALTH INSURANCE

Ellen Meara (Harvard University) spoke about how small policies can have big impacts on people with disabilities. Researchers and policy makers tend to pay attention to high-visibility laws, but often what is important but neglected is how those laws are implemented, what choices are made around key design features, and what people are actually doing in response to the laws. To illustrate that, Meara discussed three projects she is working on.

The first project concerns the relationship between age-specific program eligibility thresholds and insurance gaps. Insurance gaps are created when recommended care is not covered and is subsequently foregone, which leads to more emergency department visits, more hospitalizations, and worse self-rated health outcomes (Banerjee et al., 2010; Gresenz et al., 2007; Horne et al., 2022; Ross et al., 2006). People may encounter health insurance coverage gaps when they reach age 19 years (a time when coverage through Medicaid changes from child coverage to adult coverage) and at age 26 years (when adult-dependent or foster care coverage often ends). In their analysis of Colorado, Meara’s team focused on gaps in coverage of at least three months in duration, and the research design divided people into four groups: those with disabling conditions (e.g., Down syndrome, cerebral palsy, blindness, deafness); those with complex conditions (e.g., cystic fibrosis or sickle cell disease); those with chronic conditions (e.g., asthma, obesity); and all others. The analysis of these groups is based on claims data, so it relies on people actually accessing care and being diagnosed.

Among those with Medicaid in the first observed month, Meara reported, a spike in insurance gaps was observed for all four groups noted above at age 19. The greatest noncoverage was found to be among those with no conditions (10%); noncoverage was 8 percent among those with chronic conditions, 7 percent among those with complex conditions, and 6 percent among those with disabling conditions (Figure 5-2). Among these groups, noncoverage decreased somewhat after age 19. A different pattern appeared among those with commercial insurance in the first observed month; there was a small peak at around age 18 and a much larger peak at age 26. Youth with disabilities and commercial insurance were the least likely to experience coverage gaps as they approached Medicaid cutoffs, but they were the most likely to experience these gaps as they approached age 26.

A second research study discussed by Meara, undertaken by Ma and colleagues (2024), focused on subsidized drug coverage for low-income Medicare beneficiaries who were entitled to both Medicaid and Medicare. According to the study, if a person on Medicare applies for and is

accepted into Medicaid, the Social Security Administration auto-enrolls this Medicaid beneficiary in a low-income subsidy for the drug coverage. Those deemed eligible keep their low-income subsidy for at least six months and up to 18 months, even if they lose Medicaid at some point. Each year, about 7 percent to 8 percent of dually eligible Medicare/Medicaid beneficiaries lose Medicaid for a month or more. About half regain Medicaid coverage within 12 months, but half do not, resulting in interrupted Medicaid coverage. Everyone will have Medicare coverage, albeit with substantial cost-sharing burdens, while no one will have Medicaid, and people vary in whether they benefit from low-income subsidy coverage, which pays the premiums for the Part D plans. During that period, about 7.7 percent of dual-eligible individuals lost Medicaid for at least one month in 2023.

The SSA’s rules for implementing low-income subsidy coverage vary depending on whether Medicaid coverage ended in the first half of the year or the second half, with those in the latter group getting the subsidy for a substantially longer time. Meara reported that after 18 months, those who had lost Medicaid coverage early had a death rate that was higher (by 3.2 per 1,000) compared to those who had not lost coverage. Later, after both groups had lost Medicaid coverage, the differential went away. The differences were larger for those who were dual-eligible, as compared with those with just cost-sharing. Those taking drugs for chronic lung disease or cardiovascular disease, as well as those taking HIV antiretrovirals, were extremely vulnerable to death during the period when the low-income subsidy was interrupted. An intent-to-treat analysis implied that losing the low-income subsidy led to 8.8 per thousand more deaths.

During open discussion, Agree commented that the evidence about the mortality effects of not re-enrolling or pushing people off the rolls of Medicaid is powerful and that it is currently a major topical issue. She asked how the analyses discussed by Meara handled people who were taken off Medicare but later went back on the program. Meara responded that the analyses presented were limited to those who remained off.

The third study discussed by Meara assessed the decline in both applications and awards from SSDI since 2000. At every step of the process (being forwarded to the state Disability Determination Service, meeting medical criteria, and being approved at the hearing level or above), there was a decline in benefit allowances. The largest absolute declines appeared among racial and ethnic minorities and in ZIP codes with the highest levels of poverty and lowest levels of education.

Karraker commented that Meara compellingly drove home the point about how small variations in policy implementation and eligibility can matter a lot for outcomes and asked about the administrative burden involved. Meara responded that, historically, the United States has often created incredibly complex systems around health and disability, sometimes

deliberately to prevent false claims from being fulfilled. In some instances, as was often the case during the COVID-19 pandemic, people fail to realize that they are covered, and sometimes people may lose coverage without realizing they once were covered. Sometimes people may submit applications containing errors because the local field offices have been closed. She concluded that policy makers should step up efforts to streamline the key health care programs.

THE INFLUENCE OF STATE POLICY CONTEXTS ON COGNITIVE FUNCTIONING AND IMPAIRMENT

Emma Zang (Yale University) continued the discussion of policy impacts on the well-being of individuals with disability. She described research in which she is engaged that draws on residential history data from national surveys to study how trajectories of exposure to geographic context at the state level affect health outcomes. Her intention in the study is to both demonstrate the value of information on residential history and to urge the development of methodological tools to analyze these data.

As described by Zang, the study makes use of measures of state policy liberalism, as defined by (a) greater government regulation and welfare provision to promote equality and protect collective goods, and (b) less government effort to uphold traditional morality and social order at the expense of personal autonomy.¹ The policies cover a wide range of areas, including abortion, criminal justice, drug regulations, right-to-work laws, and employment discrimination based on sexual orientation. Zang stated that over the period 1932 to 2020, states have experienced polarization, with the liberal states becoming more liberal and the conservative states becoming more conservative (Figure 5-3). The underlying data for estimating these trends are drawn from the Dynamics of State Policy Liberalism dataset.

Zang reported that her research indicates that U.S. life expectancy would be 2.8 years longer among women and 2.1 years longer among men if all U.S. states experienced the health benefits of the more liberal policies she found in her analysis. However, she added that one of the biggest gaps in studying state policy and population health is the lack of information on lifetime exposure to state policies based on where people have lived. Most existing studies look at the connection between the current state policies and the current population health in the same year. Although people who are more likely to live in liberal states as children are also more likely to live in these states as adults, this is not always the case. Therefore, ideally, an analysis would be able to take into account both changes in the states and people’s movement from one state to another over time. Zang added

___________________

¹ The definition is adapted from Caughey and Warshaw (2016).

that exposure to these policies during early childhood may be especially important.

The research summarized by Zang is focused on cognitive change, with cognitive growth peaking in emerging adulthood and declining afterward. Dementia is the leading cause of death among older Americans. In 2022, 6.7 million Americans ages 65 and older were living with dementia, costing an estimated $345 billion annually. The number of people living with dementia is expected to grow to 13.8 million by 2060.

Using data from the Health and Retirement Study (HRS), Zang and colleagues examined two cognitive outcomes: cognitive functioning and cognitive impairment. They conducted linear random effects models with wave-, age-, and state-fixed effects, controlling for the number of waves in which respondents participated. One finding from the analysis was that a one-unit increase in the policy liberalism score between ages zero and 50 is linked to a 1.7 percent improvement in cognitive functioning and a 2.4 percent reduction in cognitive impairment risk. Zang pointed out that the improvement in cognitive functioning is comparable to what has been found elsewhere in the literature for Medicare Part D implementation. Timing, sequence, and duration of context all mattered: early childhood and adulthood exposures were both important, longer exposure was linked to better cognitive health, and a pattern of increasing

exposure to liberal policies was associated with better cognitive health. Zang’s analysis showed no significant differences based on gender, but racial and ethnic minorities were found to be much more affected by state policy liberalism overall compared to the White population.

Zang closed by commenting on how the associations in her research underscore potential areas for policy intervention to support aging populations. During early childhood, quality early education, nutrition assistance, parental support, health care access, safe housing, and affordable childcare are needed to foster a healthy developmental environment. In adulthood, policies promoting mental well-being, work-life balance, and financial security are central, as are stress-reduction initiatives and programs that provide continuous educational and cognitive engagement opportunities.

This page intentionally left blank.