Disability is most commonly measured at the nominal or ordinal level, depending on how the data is collected. If a survey simply asks whether someone has a disability (yes or no), that’s nominal data: categories with no inherent ranking. If a tool measures disability severity on a scale from mild to severe, that’s ordinal data: ranked categories where the distances between levels aren’t necessarily equal. In some research contexts, advanced statistical models can transform ordinal disability scores into interval-level measures, but this requires specialized processing that goes well beyond raw survey responses.
Nominal: Disability as a Category
The simplest way disability appears in data is as a yes-or-no classification. The U.S. Census Bureau’s American Community Survey, for example, asks respondents about six types of functional difficulty: hearing, vision, cognitive, ambulatory, self-care, and independent living. Each question produces a binary response. A person either reports serious difficulty walking or climbing stairs, or they don’t. There’s no ranking between these categories and no mathematical relationship between them. You can count how many people fall into each group, but you can’t average the results or say one category is “more” than another. This is nominal measurement.
The same applies when disability status is used as a demographic variable in research, much like race or marital status. Researchers might compare outcomes between people with and without disabilities, but the disability variable itself is just a label, not a number on a scale.
Ordinal: Measuring Severity on a Ranked Scale
Most clinical and survey tools that assess disability severity produce ordinal data. This means the scores can be ranked from least to most severe, but the gaps between levels aren’t guaranteed to be equal.
The Washington Group Short Set, used in international surveys across dozens of countries, asks people to rate their difficulty in areas like mobility and vision on a four-point scale: no difficulty, some difficulty, a lot of difficulty, and cannot do at all. These responses have a clear order, but the jump from “no difficulty” to “some difficulty” isn’t necessarily the same size as the jump from “a lot of difficulty” to “cannot do at all.” That makes the data ordinal rather than interval.
The Katz Index of Activities of Daily Living works similarly. It scores people on six self-care tasks (bathing, dressing, toileting, transferring, continence, and feeding), producing a total between 0 and 6. A score of 6 means full function, 3 to 5 indicates moderate impairment, and 2 or below signals severe impairment. These categories are ranked, but a drop from 6 to 4 doesn’t necessarily represent the same real-world change in function as a drop from 4 to 2. The Barthel Index, another widely used tool, scores people from 0 to 20 using a similar ordinal rating structure where each activity is scored as independent, needs help, or unable.
These scales follow a pattern first described by researcher Sidney Katz: as people age or develop chronic illness, they tend to lose abilities in a predictable order, from complex tasks like bathing down to basic ones like eating. This hierarchical structure, confirmed through a statistical method called Guttman scaling, means a person’s total score reliably predicts which specific tasks they can and can’t perform. Studies have found that everyday activity scales meet the threshold for this kind of reliable hierarchy, with scalability coefficients ranging from 0.65 to 0.91 in older adults. But even with this predictable ordering, the scores remain ordinal because the intervals between points aren’t uniform.
Why This Distinction Matters
The difference between ordinal and interval data isn’t just academic. It determines which statistical analyses are valid. Many common techniques, like calculating averages or running regression analyses, assume that the distance between each score is equal. When researchers treat ordinal disability scores as if they were interval data, they risk drawing incorrect conclusions. Using summative scores from ordinal scales in parametric statistics violates the assumption of equal-interval scaling and increases the likelihood of both false positives and false negatives.
This is especially problematic when comparing disability levels across different populations or tracking changes over time. If the gaps between score points aren’t equal, a three-point improvement for one person might not mean the same thing as a three-point improvement for another.
Converting Ordinal Scores to Interval Data
Researchers have developed ways to transform raw ordinal disability scores into interval-level measures. The most common approach uses Rasch modeling, a statistical technique that recalibrates each item on a scale so the distances between scores become genuinely equal. The model estimates both the difficulty of each task and the ability of each person on a shared ruler, making direct comparisons possible.
A large study published in BMJ Open applied Rasch modeling to disability data and found that the resulting interval-level measures were consistent across countries and demographic groups. The researchers demonstrated that these transformed scores produced more precise estimates when comparing disability severity between populations and when measuring change over time. Ordinal scores, by contrast, were less reliable at detecting small but meaningful differences between groups.
The practical takeaway: raw disability scores from clinical tools and surveys are ordinal. They tell you who has more or less difficulty, but not by how much. When precision matters, as in cross-country comparisons or longitudinal tracking, those raw scores need to be mathematically transformed before they can support the kinds of analyses most people assume they already support.
How the WHO Defines Disability Levels
The World Health Organization’s International Classification of Functioning, Disability and Health (ICF) provides a framework that organizes disability across three levels of human functioning. The first is the body level, covering impairments in body structures or functions, like a missing limb or reduced vision. The second is the person level, covering activity limitations, meaning difficulty performing tasks like walking or dressing. The third is the social level, covering participation restrictions, meaning problems engaging in life situations like employment or community activities.
This framework doesn’t prescribe a single measurement scale. Instead, it defines the dimensions along which disability can be assessed. Any given measurement tool might capture one or more of these levels, and the statistical properties of the resulting data depend on how the tool is designed, not on the framework itself. A binary question about whether someone has a hearing impairment produces nominal data. A graded scale measuring how much difficulty someone has participating in social activities produces ordinal data. The ICF provides the conceptual map; the level of measurement depends on the instrument used to collect the data.