For many years we’ve been saying that 15% of the world’s population has a disability. WHO recently updated this to 16%.
I dialled in some statistician friends to understand these numbers. Longtime Debriefers will remember Jennifer Madans, and behind the scenes we had help from Dan Mont and Julie Weeks.
This edition explores how to think about prevalence when disability itself can't be reduced to a “yes” or “no” question. It turns out those top-line statistics aren't comparable. Jennifer asks us to be more careful with disability statistics, and shows us the questions we need to ask about the data we use.
What are your data dilemmas? We’ll be happy to get further into disability statistics – let us know your questions.
A new number doesn’t mean that there’s been change
Peter: For many years we’ve used 15% as the global prevalence of persons with disabilities. A new report says it’s 16%. How come?
Jennifer: A change in disability prevalence, or any population characteristic for that matter, shouldn’t come as a surprise. It also wouldn’t be surprising if there was no change. But you are right to ask “why”.
To understand a change in the numbers, I ask if the change in estimates reflects a real change in what we are interested in or a change in the methods used. From a policy point of view, changes that result from differences in data collection or analytic methods are not very useful. In fact, they can be detrimental as they cause confusion. This is why data collectors need to be very transparent in how data are collected and if any changes from previous collections were made.
Both the 15% and 16% come from modelling data, and so both depend heavily on the methods used. I wouldn’t take either of these numbers literally. Further, the methods for both differed, and that alone will impact estimates. We can’t conclude there was a change in population. As the Annex of the WHO report itself says, this means that: “…the estimates of the 2011 World Report on Disability [which gave 15%] and the current report cannot be compared”.
Oddly, the report then contradicts itself. Despite giving that warning, the main text of the report does make that very comparison: “This figure has grown over the last decade and will continue to rise due to demographic and epidemiological changes, underscoring the urgency for action.” If it isn’t possible to compare, how do we know that the prevalence has grown?
The percent of the population with disabilities might well have gone up, though. The demographic trend of population aging would be consistent with an increase in prevalence and this may increase further if the demographic trend continues. Older age is closely related to chronic disease and decreases in functional ability. But the 16% isn’t evidence of the increase, and any prediction about what might happen to relationships in the future is just that, a prediction.
Disability isn’t a “yes” or “no” question
Peter: And so if someone asked you about how many disabled people there are in the world, how would you answer?
Jennifer: I can’t answer without some discussion of how disability is being identified. Hopefully I would be able to stop myself, but I would want to start with a mini-lecture on the complexities of collecting disability data.
Prevalence is important for many characteristics. In the case of disability it’s complex. Disability isn’t just a “yes” or “no” question. There have been changes in how the term is understood over time and there is still variation in how it is understood by different people. There is also still stigma attached to the label in many places.
I strongly believe that providing an estimate without information on how it was obtained is equivalent to misinformation. For me, producers of data are ethically obliged to be transparent about how data are collected and analysed. And users, particularly advocacy groups, have a responsibility to ask for this information.
Peter: I’m glad you wouldn’t start with that lecture...
Jennifer: Just one more diversion since we are talking about a statistical definition. Individual characteristics exist on a continuum, and statisticians have to choose a “cut-point” to be able to define groups so we will know the percent of the population that is in that group. For instance, with age, you might choose 18 as the cut-point, defining those under 18 as children and those above 18 as adults.
Doing this with disability is much more complex than age because it is more complicated to create the continuum. But in both cases the choice of the cut-point is critically important. Small changes can have a big effect in the number, and you may want to use different cut-points for different purposes.
Peter: So what methodology do you prefer?
Jennifer: A full definition of disability includes the interaction of personal characteristics and the environment. The statistical approach I prefer is that used by the Washington Group. It gathers information on a person’s capabilities, through focusing on six domains: walking and climbing stairs, self-care, communicating, hearing, seeing, and cognition. In each area there will be a range of abilities and difficulties. The degrees of difficulty people have in those areas can have very different impacts on social participation.
This approach addresses different parts of the disability definition separately and doesn’t directly measure the impact of functional difficulties on participation.
For the purpose of international comparisons, the standard used to create the statistical group of “disabled people” is those having a least “a lot of difficulty” in at least one of those core functional domains. Information on capabilities in these areas can then be brought together with information on social participation, to see how they are related, and how disabled people face barriers to social participation.
How many disabled people are there?
Peter: And, so, the answer?
Jennifer: I will reluctantly share a guesstimate of the prevalence: between 8-12% of the population, using the above definition. I’m uncomfortable to even give this range, as it’s not a “scientific” answer. It comes from my familiarity with what happens when we use the Washington Group questions.
I don’t want to reduce the complexity of disability worldwide to one number. There is no one worldwide data collection that would produce the disability prevalence. To create a prevalence using an agreed upon definition it would be necessary to combine data from many different sources. Data are still missing for many countries and even if available the data are not produced using the same methods.
And I would still ask what you are going to use the information for. If you’re trying to estimate the cost of a social benefit, then it will be a very different answer. Most important, I would try to move the conversation away from the estimates and more to how to use the data to investigate inclusion.
Peter: say more about what the data show.
Jennifer: We can see some general patterns. Disability almost always increases with age, for example. At a younger age impairment is more likely to be caused by injury, and at an older age, by chronic illness. Rates of disability tend to be higher among women, and to some extent this reflects their longer life expectancy in many places in the world.
The relationship with wealth is more complicated. Disability is generally related to poverty, but we can also find higher disability rates in richer countries. Sometimes this reflects different methods and definitions used as part of more extensive statistical system. In richer countries populations often have greater access to health and people with higher incomes might be more likely to report their difficulties. Further, richer countries might also have lower mortality rates and so people may live for longer with a disability.
Different meanings of “disability”
Peter: This statistical division of “disabled” and “non-disabled” is different from the ways we use those terms in other contexts. It’s different from the population that call themselves “disabled”, for example, which is a much lower percent of nearly all populations.
Jennifer: You are absolutely correct. I think of it this way: the word “disability” is used in everyday conversation and, most important, the word has different meanings across cultures, people within cultures and in different contexts. The advantage of a statistical definition, especially a standard definition, is that it is clearly stated and will mean the same thing to everyone. If the definition is accepted as an international standard, then, overtime, there will develop a more universally-shared interpretation of statistics based on that definition.
The statistical definition that I describe is based on evaluating functioning across a continuum of difficulty and then selecting a useful cut point to define the population with or without disability. It can be very different from how an individual thinks of themselves.
This is no different from the way other social indicators are defined. For statistical purposes we have to create groups – based on age, sex, or poverty, say – that don’t necessarily reflect each individual’s orientation or identification.
With disability, for example, two people might report the same level of functional difficulty and even the same type and level of barriers to participation, but one will self-identify as a person with disability and one will not. Older people, who have acquired functional limitations slowly as a result of chronic conditions, often do not self-identify as having a disability.
This is why asking on a survey “Do you have a disability?” results in low prevalence. From a civil rights perspective, whether someone identifies as a person with disability may not be relevant to whether or not they need support or face discrimination. But from an identity or advocacy point of view, that question can be very important.
Going beyond prevalence
Peter: The prevalence statistic is very important to the disability movement and we mention it in any document. How does that differ from the way you’d use it yourself?
Jennifer: I see my role as constantly reminding people not to accept a number, any number, at face value, but to think about how it helps them do their job. So, I find it frustrating to see speeches that start by saying there are so many disabled people and we will work for them, but without understanding how to get from that number to identifying and solving more specific challenges those very people face.
Prevalence provides context and that is very useful, for any group. Knowing how big something is will affect how we think about it. The interest in the prevalence of disability also reflects the past where there was the perception that disability was rare and therefore not of great political or policy interest.
To address this misconception, it was necessary to more accurately estimate the prevalence to show that the number is not small. The work that has been done to advance how the social model of disability gets translated into data collection tools has provided more useful data for policy makers by clarifying the size of the population. Prevalence estimates are important advocacy tools.
But it is important to remember that they are starting points. As we discussed, the population that the prevalence is based on can be quite diverse in the extent and type of their functional limitations depending on the cut point used.
Peter: What kind of questions do you have beyond prevalence?
Jennifer: The experience of disability varies. We could look at different types or categories of disability, rather than a dichotomy between disabled and non-disabled. We can look at how disability intersects with other characteristics, like sex, age, and ethnicity.
Once we have defined the population with disabilities then we need to get information on the specific barriers and facilitators that produce restrictions in fully participating in society.
Peter: What are our opportunities for growth as users of disability data?
Jennifer: We all need to become more sophisticated and more demanding when it comes to data. One prevalence estimate, even if broken down by other demographics, is not enough to identify the population with disabilities.
Mostly, we need to broaden the discussion to become more comfortable with how the complex nature of disability is translated into data. There will never be just one prevalence of disability because there will always be more than one way of identifying the population of interest.
Peter: Thanks Jennifer.
Jennifer: Thank you Peter. Let’s do it again. There are lots of data issues to discuss.
The illustration is by Tan Kuan Aw.
Thanks to Jennifer, Dan and Julie for contributing their time on this edition.
Jennifer Madans is a statistics consultant and senior associate at the Center for Inclusive Policy (CIP). Dan Mont is CEO of CIP, who also support the Debrief. Julie Weeks is a statistician at the United States’ National Center for Health Statistics.
Dan, Jennifer and Julie have all been involved in the Washington Group that designed the questions Jennifer describes.
Disability Debrief is made by me, Peter Torres Fremlin, with support from the individuals and organisations that read it.