How can we measure how common disability is?

By now, you’ve heard that 15% of the global population, or 1 billion people, has a disability. As you heard earlier this week, the proportion of people in a population with disability is called the prevalence. This means that the prevalence of disability is 15%, or 15 in every 100 people. Do you know where this estimate comes from? This estimate is usually referenced as being from the world report on disability. Within the report, the estimate is from the World Health surveys, which were completed between 2002 and 2004. These surveys have quite different questions and criteria to the most commonly used tool today, the Washington Group Short Set, and 15% was the all-age average across 59 countries.

This ranged from 12% in high-income countries to 18% in low-income countries, and between 6% in those aged 0 to 17 versus 30% in those aged 60-plus. Since then, much has changed that might have impacted on how relevant this estimate is today. The world is ageing, and as we’ve seen, disability becomes more common with age. Conversely, health system strengthening in many contexts means that less people are living with disability from avoidable impairments, such as visual impairment from cataract, hearing impairment from impacted wax, or impairments related to complex births. So how do we get more up-to-date information on how common disability is? And how do we know the estimates are accurate in our setting?

Every 10 years, most governments complete a population census. This is a complicated undertaking in which the whole population of a country is enumerated, meaning listed, usually on a single night. It provides a wealth of important information on population structure, living circumstances, and demographics. For the 2020 round of census, the Washington Group questions have been recommended by the United Nations for inclusion to capture information on disability. Because of the COVID-19 pandemic, many of these censuses have been delayed. However, the collection of comparable disability data will be of huge importance for estimating disability prevalence going forward and one that many other organisations will be able to draw on to understand how common disability is in their country.

Between censuses or where there is a need for more in-depth information on disability than a census can provide, population-based surveys are needed. The population-based survey is an example of a cross-sectional or single-time point study usually designed to estimate the prevalence of one or more outcomes.

Unlike in the census, where every single person living in the country is enumerated, surveys select a random sample who represent the whole population. By selecting a sample, it is possible to get information from large populations at lower cost, quicker speed, and increased accuracy. For this to work, the sample that we select has to be representative, which means if we did take a census, we would see the same characteristics in our sample as in the whole population, and the prevalence of the outcome in our sample would be similar to the prevalence in the whole population. Each person in the population has to have an equal chance of being selected in the sample for this to be the case.

And for this reason, we can’t estimate how many people with disabilities there are through measuring disability at service centres or in clinics. The number of people with disabilities attending a clinic may not represent all people with disabilities in the population, for example, because some people with disabilities may not be able to get to the clinic because of mobility or finances. There are a number of key steps in survey design which we will quickly go through now. Step 1 is to define the target population. First, you need to define who your target population is. Is this all people living in a sampling area, for example, a province or a country? Or all people above a specific age group?

We will see in a minute how this choice affects the sample size and therefore the time and cost of completing a survey. Step 2 is to choose the sampling method. How will you identify the people who will be in your sample? One option is simple random sampling. This is where you number all of the people in the sampling area and randomly draw out the required sample size. The list might be from a polling station or a school register. The method is simple, and you reduce the risk of errors in your sampling by simply drawing people from a complete list. But how easy is it to get a complete list of people? Both the examples just now might be incomplete.

For example, they will miss people who are not registered to vote or children who are not in school. This adds bias, which you will learn more about in week 3, particularly because people with disabilities may be less likely to be on both of these lists. Plus, even if a full list was available, practically, this could be hard to manage as people could be scattered about far apart, leading to lots of time travelling between participants to collect data. Instead, population-based surveys tend to use cluster-based sampling. Here, instead of randomly drawing individuals from a list, we randomly draw groups of people who are living more closely together in clusters.

So going back to our example above, we might draw whole schools or pop polling stations from a list of either rather than individuals.

If we were measuring disability prevalence in the population, we would lead a list of population data. Usually, this is in the form of enumeration areas. The lowest administrative area recorded by the country census. Enumeration areas tend to have a total population size of between 200 to 500 people, and often, the Census Bureau has maps of each available, which can be very helpful for managing data collection. Step 3 is to calculate the sample size. This tells you how many people you need to sample from the whole population to be able to accurately estimate the outcome of interest. If your sample size is too small, your results may not be precise enough to draw a meaningful conclusion.

Sample size is dependent on a couple of important parameters and tells you how many people you need to be able to include to answer your research question or questions. These parameters are explored in the See Also documents as we don’t have time to go into them all here. Like many other research groups, at ICD, we often use population-based surveys to estimate disability prevalence and learn more about the lives of people with disabilities in different contexts. Here you can see disability prevalence estimates from surveys that we’ve completed across the world using the Washington Group questions and how much this can vary between settings, as well as being related to population-age structure and access to health and rehabilitative services.

This variation can also be related to variability in how the questions were translated, how the questions were asked, how functional limitations are perceived in different settings. Usually, when we complete surveys, and as you can see from some of the report titles, we also include an extended questionnaire for people with disabilities, and people from the same community ask them the same age and the same sex. This is called a nested case-control study design. It allows us to explore the differences in the lives of people with and without disabilities in more detail.

For example, in a recent study we undertook in Vanuatu, we found that people with disabilities were twice as likely to live in a household with a female household head less than half as likely to have worked in the last week, and reported statistically lower satisfaction with their lives than people without disabilities. We can break these results down by age group and sex. For example, amongst people with disabilities, women and people with self-care limitations would at least likely to be thriving. These data are really important in helping us understand not just how common disability is but the lived experiences of people with disabilities, so services, programmes, and policies can be designed to fully meet their needs.

You will hear more about the measurement of lived experience later this week. In summary, you should now be familiar with how we measure how common disability is and how this data can be used. Later in the course, you will learn more about the risks associated with sampling and how to mitigate these.