2

Defining, Conceptualizing, and Measuring Disability

OVERVIEW

The opening session of the workshop was structured around a set of guiding questions:

• What are the conceptual and technical challenges in measuring disability and estimating disability prevalence for the working-age population?
• How do alternative definitions and measurement approaches affect estimates for specific populations?
• What measures are most appropriate for different purposes—e.g., for population monitoring, for establishing eligibility for programs, for informing policies, or for international comparability in reporting?
• What are the key data and infrastructure needs for improving measurement and statistics relating to people with disabilities? And, how can existing data investments, particularly in longitudinal sources, be enhanced to better study disability in working-age populations and how disability tracks with work experiences and trajectories?
• For understanding disparities between disabled and non-disabled people, what frameworks are available for measuring health disparities that are specific to the disability context?

Addressing these and other questions were Scott Landes (Syracuse University and committee member), who served as moderator; Bonnielin Swenor (Johns Hopkins University), who presented on key data issues and the overall measurement framework and also commented on how federal policy advances need to be considered in research supported by the National Institute on Aging; Kathleen J. Mullen (University of Oregon), who outlined issues and methods for measuring work capacity; Sarah F. Rose (University of Texas, Arlington), who discussed historical trends in working-class people with disability, in workers’ compensation, and in disability discrimination; and Anjali Forber-Pratt (American Association on Health and Disability), who explored strategies for ensuring equity in ongoing and future disability measurement.

In initiating the discussion, Scott Landes (Syracuse University) provided some conceptual framing about disability measurement. He explained that there is currently a robust debate about whether, in official government agency surveys and administrative contexts, to continue measuring disability solely as functional limitations or to invest the resources needed to develop new, more inclusive and accurate measures of disability. The two main instruments used by the agencies equate disability with limitations in

“core functions.” Those two instruments are the ACS-6, used in the Census Bureau’s American Community Survey (ACS) and the National Health Interview Survey, which is administered by the Census Bureau on behalf of the National Center for Health Statistics (NCHS), and the Washington Group Short Set, which has been proposed as a substitute for the ACS-6 questions, despite its known undercount of disabled populations.¹ Measures of functional limitations are helpful but do not include all disabled people, for example those with only intermittent functional limitations. Similarly, data sources tracking benefits exclude those who are disabled but not receiving benefits, and criteria based on work exclude those disabled people who are able to work, with or without accommodations.

Landes commented that it is idealistic to expect a measure that includes all disabled people due to the complexity and heterogeneity of the experience of disability. What can be done, however, is to develop better survey questions and to clearly articulate who is and is not captured by various definitions. Landes noted that some researchers do that well, but more typically researchers overgeneralize findings to the larger disabled population.

Landes closed his overview by noting the “problematic” tendency in research to view disability as a negative health outcome, rather than to realize that disability status is distinct from health and is a morally neutral identity for many disabled people.

KEY DATA ISSUES AND THE MEASUREMENT FRAMEWORK

Bonnielin Swenor (Johns Hopkins University) presented two images to depict her view of the current state of data on disability. The first, by the artist Keith Haring,² shows a face within a mosaic. The second (Figure 2-1) is Ursus Wehrli’s rearrangement of the same cells in Haring’s art by color and size and is titled The Statistical Analysis of Keith Haring. Swenor posited that the second image is representative of the current state of disability data: “devoid of context in which the interconnections can be seen, the data lose meaning and their connection to the people they are meant to represent.”

Swenor reviewed several types of models of disability in terms of the kinds of information they can capture:

___________________

¹ Elements of this undercount are documented in a number of sources, including here: www.urban.org/urban-wire/proposed-census-changes-would-drastically-undercount-disabled-americans

² Artwork can be viewed here: https://www.themarginalian.org/2008/11/20/tidying-up-art/. The National Academies was unable to obtain permission to reproduce the original artwork by Keith Haring.

• The medical model of disability expresses disabilities as being a consequence of a health condition or a trauma and places the focus on an attribute that should be treated or fixed.
• A rehabilitation or functioning model of disability is one in which disability is defined by someone’s limitations or their ability to perform daily activities.
• The social model of disability seeks to measure the gap between a person’s functioning or limitations and the environment or the social context in which they are living.
• The identity model of disability is expressed in terms of how a person self-identifies the way they are participating in the broader community, in terms of how a person self-identifies their participation in the broader community.

Swenor’s presentation proposed a model that embeds the idea that people with disability share an identity that, not unlike gender and race, may be used as a demographic characterization.

Returning to the approaches used by government agencies, Swenor pointed out that research by Jean Hall et al. (2024) helps to quantify the degree to which current measures undercount the number of people with disabilities: the ACS-6 and the Washington Group Short Set failed to identify 20 percent and 43 percent, respectively, of respondents who reported disabilities. Furthermore, the paper reports that these measures performed

especially poorly in capturing respondents with psychiatric disabilities or chronic health conditions. Landes and Swenor, comparing the two measures within the National Health Interview Survey, found only partial overlap: among respondents with any ACS disability status, 42.9 percent had their disability status counted in the Washington Group Short Set measures. Section 4302 of the Affordable Care Act (ACA) requires the Secretary of Health and Human Services to ensure that any federally conducted or supported health care or public health program activity or survey collects and reports specified demographic data regarding health disparities. Landes and Swenor pointed out this aspect of the ACA requiring identification of disability as an essential data element of a person’s demographic profile. As a result of section 4302, the ACS-6 questions were outlined as the standard to measure disability as a demographic variable, but Swenor argued this directive has not been upheld.

Swenor explained the work of a research team led by Landes, who outlined a roadmap with immediate, midrange, and long-term goals for disability measurement. The immediate goal is to continue using the ACS-6 disability questions because that is what has been outlined in the ACA, and doing so would provide continuity. The midrange goal involves expanding on the ACS-6 questions to better capture groups in the disability community that the ACS-6 is missing. The State of Oregon, for example, has sought to do this through its Race, Ethnicity, Language, and Disability initiative (McGee, 2020). Swenor continued by noting there is also a need to expand where disability data are being collected to more places, more surveys, and more cohort studies. Over the long range, there is a need to create a more representative disability measure grounded in the views of disabled people across the lifespan, and then to create new data standards based on these measures.

A common societal view is that people with disabilities are a group that should be reduced or prevented. Often the way that data are collected and analyzed reinforces the idea that non-disabled people are the ideal. Instead, new conceptual frameworks should, in Swenor’s view, be centered on data reflecting the perspectives of disabled people, putting the focus on environments, information, and policies, not on work and other capacities. Among other things, this would involve not conflating health and disability. Indeed, Swenor explained how “work capacity” as a construct is problematic and very much rooted in an ableist perspective.

MEASURING WORK CAPACITY

Work capacity, which is a context-specific measure of functional capacity, gauges one’s potential to work, apart from whether one works or not. Kathleen J. Mullen (University of Oregon) presented research investigating

several aspects of measuring work capacity, based on joint research undertaken with Italo Lopez Garcia (University of Southern California) and Nicole Maestas (Harvard Medical School).

Mullen began by noting that research measuring functional capacity often focuses on cognitive ability. In older populations, there are measures of cognitive functioning or dementia. At the other end of the age spectrum are broader achievement tests, such as the Armed Services Vocational Aptitude Battery, developed by the Department of Defense to measure young adults’ potential to succeed in the military. When economists relate wages to education and other factors, ability is often expressed as the unobserved component. To measure work capacity, economists have developed alternative approaches, such as looking at labor supply responses to policy variations like disability insurance benefits or social security earnings.

Mullen and her research colleagues have instead mapped individuals’ functional abilities to potential occupations in the national economy. By measuring abilities on the same scale that is used to measure ability requirements of occupations, each person’s ability can be compared with an estimate of the level of ability required to perform an occupation and derive a set of potential occupations that people can do; potential earnings can also be mapped to a given set of potential occupations.

For research such as that described by Mullen, the Occupational Information Network (O*NET) is a primary data source for occupational information (e.g., skills, knowledge and abilities required). O*NET differentiates skills from abilities—the former are obtained through training, while the latter are enduring talents that can help a person do a job. O*NET categorizes 52 abilities across more than four different domains posted in its online module. As an example, O*NET characterizes janitors as performing roughly 19 tasks. Some of the top abilities required for this job are near vision, oral comprehension, static strength, trunk strength, extent flexibility, and manual dexterity.

Using data from the RAND American Life Panel, Lopez Garcia (2022) collected ability measures harmonized with the O*NET data. O*NET is based on asking analysts to rate how important each ability (such as arm-hand steadiness) is to the performance of a job, assigning a five-point scale from not important to extremely important. If the ability is needed, then a seven-point scale is used to indicate what level of that ability is needed. For example, lighting a candle would require a low level of arm-hand steadiness (a “2”) while cutting facets in a diamond would require a high level (a “6”). Using this approach, researchers can calculate the percentage of abilities of any job that a person can do.

Mullen explained that work capacity can be characterized in multiple ways. For example, it can be expressed as the fraction of all jobs that an individual can perform, or as an estimate of the mean potential earnings

from the range of jobs possible to the individual. A strict measure might require that a person has 100 percent of the abilities required for the job, while a more generous measure might require that a person has 88 percent of the abilities.

One finding presented by Mullen is that workers often do not have all the abilities required to perform a given job. In the sample she studied, about 39 percent of the people have 100 percent of the abilities required to do their own job across all 52 abilities, while the 75th percentile is at 88 percent of the abilities required to do their job. Mullen also said that the importance of health to the ability to perform work tasks is most prominent at lower levels of health. She argued that there is little difference in what occupational sets can be performed when comparing a worker with excellent health versus one with very good health, and much more when comparing fair health with poor health. Looking across the age cycle, a decline in abilities is generally observed over time; but often these declines do not reach the point of falling below occupational requirements (Lopez Garcia et al., 2023).

Mullen shared that in some newer follow-up research, her team is comparing self-reported ability measures with objective measures of performance, seeking to measure the extent of systematic error in self-reports. For example, it could be that older people systematically overrepresent what they can do, which could make a big difference in where they fall on the spectrum of ability levels.

Mullen and co-authors adapted a Functional Abilities List used by the Dutch Social Security Administration for determining whether a person qualifies for disability insurance benefits for self-administration in the United States and then combined this with harmonized Dutch occupational requirements data to measure work capacity. This allowed them to characterize the types of jobs that people can do and the set of potential earnings that they have as related to their past earnings.

HISTORICAL TRENDS

Sarah F. Rose (University of Texas, Arlington) spoke about how the concept of disability has evolved over time. Disability historians have pointed to the late 1910s and early 1920s as a key turning point when disability emerged as an umbrella term covering populations that policy makers, employers, and the public had long viewed as distinct. People in the wage labor market, those who labored in rehabilitative institutions for no pay, and those who could not work and relied on public or private relief all fell under the terminology of disability. So too did people with a very wide range of impairments.

Rose shared how the threat and reality of disabling injuries has long been a normal part of the everyday worker experience. Until quite recently,

and still today in some more dangerous occupational sectors, the notion that a mature worker could have an intact, unscarred body was fanciful. Rose recounted the story of how an early Ford Motor Company worker recalled his punch press crew losing an average of 16 fingers a month—the possibility of acquiring work-related impairment was simply part of the job.

Next, Rose explained how workmen’s compensation laws passed by states, mostly in the 1910s, combined with the preemployment medical inspections required by insurance in the wake of these laws, had a devastating and long-lasting impact on many people’s access to the paid mainstream workforce, particularly the industrial workforce. Legislators had intended for compensation statutes to prevent industrial accident victims from falling into poverty. Ironically, workmen’s compensation laws contained financial incentives for employers to screen out disabled workers and job applicants out of the fear that they were more likely to have a second injury and thus become eligible for costly workmen’s compensation benefits.

By 1926, half of the nation’s 600 largest companies reported that they were unwilling to hire any disabled workers at all, largely due to concerns about workmen’s compensation. However, Rose noted that the Ford Motor Company took a different approach, using screenings and a more complex understanding of mechanization to place hires in jobs appropriate for their physical capabilities. By 1918, more than 18 percent of the company’s nearly 30,000-person workforce were men with physical disabilities. Rose explained that, the Ford case notwithstanding, in just one generation—with the help of industrial physicians, safety engineers, and policies to safeguard workers—disability went from being widely viewed as simply part of working life to being a barrier to paid work.

ENSURING EQUITY IN DISABILITY MEASUREMENT

The next presenter during the measurement needs session was Anjali Forber-Pratt (American Association on Health and Disability) who discussed strategies to ensure that all populations are captured in ongoing and future disability measurement, including those not well covered in statistical agency surveys, and in a way that captures disability as core demographic data. Forber-Pratt spoke of the frustration of not having adequate data for some basic analyses, and the need for a data infrastructure that consistently captures information that can be harmonized across systems, across federal agencies, and across publicly available datasets. As an example, Forber-Pratt pointed to a paucity of data about the prevalence of homeless or undocumented or incarcerated individuals with disabilities residing in the United States.

Forber-Pratt warned that it is important to know the perspective being represented when using administrative data. For example, data from pro-

vider agencies may omit the voices of those who use consumer-directed services, and data from employers or labor unions might miss the perspective of workers with disabilities. She also noted that equity and equality are not synonymous. Equity is the quality of being fair and just, and equality is the state of being equal, especially in status, rights, and opportunities. Equity means meeting communities where they are and then allocating resources and opportunities as needed to make up for the historical systemic oppression and ableism that the disabled population has experienced. Various subpopulations within the disability community, such as Black individuals, have, in Forber-Pratt’s view, long been missing or undercounted in datasets. Justice, she noted, is about the proactive removal of the systematic barriers and addressing the root cause or causes of the inequities.

Forber-Pratt stressed the importance of community partner engagement in future measurement efforts. Specifically, she believes that the perspectives of people with disabilities should be included in decisions at all levels and in the development and testing of future instruments as a regular and continuing process. Pilot testing with a diverse set of participants is needed to identify potential biases or misunderstandings before finalizing future questions. She further noted that researchers should incorporate universal design principles of learning into the design of future questions to ensure that they are as usable as possible by the widest range of individuals without the need for adaptation. Forber-Pratt also emphasized that the framing of questions is critical to avoid perpetuating stereotypes or ableist assumptions. For example, focusing on limitations would give a different perspective than focusing on what support needs or accommodations might be needed for a person to succeed. For such an evolution of the data infrastructure to occur will require a next generation of statisticians with disability expertise.

A critical point made by Forber-Pratt is that the relevant array of disability data includes not only health information but also data collection covering many areas of life, such as transportation, justice, and housing. A person’s accessibility is dependent on the interaction of functional abilities and the surrounding environment. A physical limitation may not be disabling at all in a well-designed building, park, or airport while it may be disabling in outdated facilities. As people with disabilities face hardships across multiple areas of daily life compared to those without disabilities, improved data are needed to provide a more comprehensive picture of these gaps and to provide a better understanding of the underlying causes.

Though it was beyond the scope of this workshop, Forber-Pratt advocated for a federal data standard or statistical policy directive to establish consistent measures of disability status. Regarding question design, she stated that disability needs to be embraced as a demographic. Because this is not the current practice, questions on federal surveys often exclude many individuals and subpopulations, such as those with mental health disabilities, intellectual

and developmental disabilities, and communication disabilities. Forber-Pratt noted that question design should also be sensitive to the presence of mobility or communication aids or support and how that affects what would be captured in question wording.

Another challenge to improving the overall data infrastructure highlighted by Forber-Pratt and other participants is that different definitions of disability are in regular use. Sometimes these different definitions of disability are driven by competing statutory definitions or specific programmatic requirements. The Interagency Committee on Disability Research recently updated a compilation of the types of disability definitions that are in use across the federal government, and right now there are dozens of variants in use across the United States. The use of multiple definitions and conceptualizations makes it difficult to compare findings across agencies and programs.

Forber-Pratt also pointed out that the presence of a coherent disability identity, which she described as both an internal part of who a person is and something that guides external relationships with the broader community, is believed to help individuals adapt to disability, including navigating daily social stressors or ableisms. However, not all people who meet criteria for having a disability self-identify as such. For example, someone might receive Social Security Administration disability payments but not claim to be disabled.

DISCUSSION

Ari Ne’eman (Harvard School of Public Health) opened the discussion, raising the issue of how the ACS-6 and Washington Group question sets define and measure disability. One complexity is that many who qualify for disability civil rights protections under the Americans with Disabilities Act of 1990 and Section 504 of the Rehabilitation Act of 1973—a national law that protects qualified individuals from discrimination based on their disability—would not identify themselves as disabled if asked. Ne’eman suggested that definitions need to allow for flexibility to fit the context in which they are being applied. For example, a research question related to income support or long-term services and supports may apply to a different population than one dealing with workplace protections, which would involve a much broader definition of disability. The ACS-6 (which, again, is used in the Census Bureau’s American Community Survey and NCHS’s National Health Interview Survey) is limiting because of its use of binary yes/no questions rather than the Washington Group’s continuum.

Forber-Pratt replied that both sets of questions have limitations and that the conceptualization of disability needs to be advanced if measurement is to be improved. She particularly objected to cutoffs for inclusion

in various categories used, with the Washington Group questions being too limiting, resulting in individuals misleadingly not being counted as having a disability. Swenor agreed that both measures have limitations, and stated there is a need to develop something better.

Landes responded that there is a need for multidimensional measures—of functional limitation, but also of experienced ableism, built environment, access to resources, and accommodations. He also considered it mistaken to consider any measure of functional limitation to be representative of the disabled population. He saw value in a measure of functional limitation that is scaled, especially within medical settings, but viewed the severity scale within the Washington Group measure as undercounting those with severe disabilities, especially for those who are blind and deaf. Mullen added that context shifts over time, as has been the case with occupational requirements. For example, robotic tools might lessen the need for arm-hand steadiness while benefiting all surgeons. Landes added that there is slippage on both sides: there are people who do not know to ask for accommodations, but also people who are being accommodated but do not identify as having a work-limiting health condition because it is being addressed. Capturing such nuances requires a reconceptualization of disability measurement beyond what currently exists.

This page intentionally left blank.