Introduction
In many countries, people with disabilities face significant disadvantages in areas such as employment, education and democratic participation (Amilon et al. 2021; Burkhauser, Houtenville and Tennant 2014; Lauer and Houtenville 2018; Rubio-Valverde, Nusselder and Mackdenbach 2019). The disparities between people with and without disabilities stand in contrast to the goals of the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD), which reaffirms the rights of people with disabilities in these areas (United Nations 2020). To close the gaps between people with and without disabilities, policymakers need accurate statistical data as well as a detailed understanding of the population with disability (McDermott and Turk 2011).
However, identifying this population is challenging (Altman 2014; Myers et al. 2020), as disabilities are a result of the interaction between the individual’s health conditions and environmental and contextual factors that influence how the individual experiences disability (WHO 2013). Consequently, according to the World Health Organization’s (WHO) International Classification of Functioning, Disability and Health (ICF), a disability entails impairments, participation restrictions and/or activity limitations and the environmental and personal factors that can either hinder or support a person’s functioning across these three levels (Clarke et al. 2011). Impairment refers to limitations of body structures or functions, participation restriction refers to restrictions in a person’s involvement in society, and activity limitation refers to an individual’s difficulties in executing activities (Shandra 2022).
Designing survey questions that reflect the multifaceted nature of disability has proven difficult. While a wide range of survey measures designed to identify and quantify the population with disabilities exist, different measures result in diverging estimates of disability prevalence (Bourke et al. 2020; Brandt et al. 2014; Burkhauser, Houtenville and Tennant 2014; Hall et al. 2022) and who is identified as having a disability differs depending on which measure is used (Amilon et al. 2021; Brandt et al. 2014; Burkhauser, Houtenville and Tennant 2014; Hall et al. 2022; Landes, Swenor and Vaitsiakhovich 2024; Hugaas and Tøssebro 2012). Given these irregularities, understanding the implications of using different disability measures is of vital importance for policymakers who wish to identify and eliminate disadvantages experienced by people with disabilities.
While disability is generally understood as a permanent state, impairments have a high degree of variability over time, and whether an impairment is disabling or not may depend on the person’s statuses and roles within the social and physical environment and may change with medical and technological advances (Barnartt 2016). Thus, while some severe impairments may be permanently disabling (in all contexts and environments), disability is often a fluid state that may vary over time and place.
This fluidity of disability is reflected in studies that demonstrate that significant proportions of respondents change disability status across survey waves (sometimes referred to as transitory disability (Myers et al. 2020; Ward et al. 2017)). Changes in disability status have been found to correlate with changes in self-reported health (Myers et al. 2020). Thus, inconsistent answers to disability questions over time reflect real health-status changes. However, as emphasized in the ICF, the experience of disability may not only involve physical or mental impairments but also participation restrictions and/or activity limitations (Shandra 2022; WHO 2013). Thus, changes in the experience of disability may, in part, be driven by changes in factors related to societal in- and exclusion (Aitken et al. 2022).
Against this backdrop, this study makes two novel and important contributions to the literature on disability measures: Building on data on 10,586 respondents aged 16–64 from the Survey of Health, Impairment and Living Conditions in Denmark (SHILD) merged with information from administrative registries, we first investigate and compare the prevalence of changes in disability status over time for two internationally recognized and widely used disability measures: the Washington Group Short Set on Functioning (WG-SS) and the Global Activity Limitation Indicator (GALI). Second, we investigate if changing response patterns to these measures reflect changes in participation restrictions, activity limitations and registry-based health measures, in addition to changes in self-reported health. Thus, our analysis not only contributes to clarifying the implications of using different disability measures, it also improves our understanding of how individuals’ experiences of disability (as reflected by their responses to the WG-SS and GALI) are associated with societal barriers to inclusion. Such knowledge is important, as it sheds light on potential targets for policy interventions.
The conceptual understanding of disability and the selected measures
There are two main conceptual understandings of disability based on two opposing types of models: medical models that frame disability as an impairment at the individual level and social models that point to limited societal inclusion as the primary source of disability (Shandra 2018). Both models have been criticized: medical models for ignoring societal discriminatory practices and social models for failing to recognize how the individual’s impairments affect their experience of disability (Altman 2001).
The WG-SS was developed by the United Nations Washington Group on Disability Statistics to provide internationally comparable data on disability consistent with the ICF (Lauer, Henley and Coleman 2019; Madans et al. 2001). From 2010 to 2024, over 75 countries used the WG-SS in censuses or surveys (Lauer, Henley and Coleman 2019; Washington Group on Disability Statistics 2025).
With cross-national comparison as its primary raison d’être, the WG-SS survey questions concern impairments at the individual level rather than disparities in participation, as opportunities for participation may be susceptible to cultural and economic differences across countries (Madans, Loeb and Altman 2011). Accordingly, the WG-SS defines disability as functional limitation(s) in six core domains: vision; hearing; walking or climbing stairs (mobility); remembering or concentrating; washing or dressing (selfcare) and communicating. Answers are given on an ordinal scale with four categories (no difficulty, some difficulty, a lot of difficulty, cannot do at all).
GALI—a single-item survey for measuring participation restriction—is included in several major European surveys such as the European Health Interview Survey (EHIS), the Survey of Health, Ageing and Retirement in Europe (SHARE) and the Survey of Income and Living Conditions (SILC) (Van Oyen et al. 2018). GALI reflects a social understanding of disability, as the question asks whether a health problem limits the respondent’s participation in activities:
For at least the past 6 months, to what extent have you been limited because of a health problem in activities people usually do? Would you say you have been severely limited, limited but not severely, or not limited at all?
Given the widespread international use of the WG-SS and GALI for research and policy purposes alike, a thorough understanding of the populations that are defined as having changing or consistent disability status based on these measures is paramount.
Hypotheses
The differences between the WG-SS and the GALI measures may result in differences in the shares of respondents with changing and consistent disability status, and differences in the association between changes in disability status and changes in health, participation restrictions and activity limitations, for the following three reasons:
First, the GALI survey question specifies that the duration of the health problem should be at least six months, whereas the WG-SS questions do not involve requirements regarding duration. Thus, people with temporary health problems may report having a disability in the WG-SS but not in GALI. These differences lead us to our first hypothesis:
H1. The proportion of respondents who change their response to disability questions across survey waves is larger when using the WG-SS than when using GALI.
Second, both measures are designed to reflect respondents’ experiences of impairments or health problems. We therefore expect changes in health to be reflected in both measures—even though we cannot predict the magnitude of these associations or any differences therein. We therefore hypothesize as follows:
H2. Changes in disability status are associated with changes in health for both the WG-SS and GALI.
Third, the GALI question asks about limitations in activities people usually do, whereas the WG-SS asks about functional limitations. Thus, a prerequisite for the GALI definition of disability is that a health problem causes activity limitations or participation restrictions, but this is not the case for the WG-SS definition. We therefore expect the associations between changing disability status and changes in activity limitations and participation restrictions to be larger when using GALI than when using the WG-SS. We thus form two additional hypotheses:
H3a. Changes in disability status are associated with changes in participation restrictions and activity limitations for both the WG-SS and GALI.
H3b. These associations are stronger for GALI than for the WG-SS.
We test these hypotheses by comparing the variation across survey waves in the shares of respondents with changing and consistent disability status, and the associations between changing disability status and changes in health, participation restrictions and activity limitations for the two measures.
Materials and Methods
Data and sample
We use data from the 2016 and 2020 waves of the Survey of Health, Impairment and Living Conditions in Denmark (SHILD), a representative, longitudinal survey covering a broad range of topics including living conditions and disability in the Danish population aged 16–64 (Amilon et al. 2021). The 2016 and 2020 waves included 20,451 and 17.935 respondents and response rates were 54% and 47%, respectively. A total of 17,447 respondents of the 2016 wave were eligible for participation in 2020 (i.e., were not older than 64, alive and residing in Denmark), and 10,586 (61%) participated. These respondents to both waves form the sample for our analyses. We merge the survey data from SHILD with comprehensive information from the central Danish health, population and employment registries. The data collection and linkage of data was approved by Statistics Denmark (reference number 706546).
Measures of disability
To construct the WG-SS measure of disability, we use the cut-off values recommended by the Washington Group and code individuals as having a disability if they answered either ‘a lot of difficulty’ or ‘cannot do at all’ to at least one WG-SS question (Washington Group on Disability Statistics 2020). We define individuals as having a GALI-defined disability if they answered that they had been ‘severely limited’ to the GALI survey question.
Definition of changing and consistent disability status
We define four disability status groups: never, consistent, declining and emerging (Table 1). The ‘never’ group consists of individuals who do not report a disability (in neither 2016 nor 2020), whereas individuals with ‘consistent’ disability report a disability in both survey years. Those with ‘emerging’ disability report a disability in 2020, but not in 2016, whereas those with ‘declining’ disability report a disability in 2016, but not in 2020. We define respondents with emerging or declining disability as experiencing changing disability status.
Measures for health, participation restrictions and activity limitations
We use both self-reported and registry-based measures for health status changes. First, respondents in the 2020 wave of SHILD were asked: ‘Is your health today better or worse than it was three years ago?’ and were given the response options: ‘better,’ ‘the same’ and ‘worse.’ We use this measure to construct an indicator for self-reported changes in health since the previous survey wave. Second, we use data from the Danish administrative health registers from 2017 to 2020 to construct a variable for registry-based changes in health or impairments. We construct a binary variable indicating whether a respondent received hospital care (in- or outpatient) pertaining to one of the following conditions: behavioral disorders or ADHD, autism spectrum disorders, mobility disabilities, sensory disabilities, learning disabilities, developmental disorders or mental disorders. The conditions were identified based on International Classification of Disease (ICD-10) codes and have previously been found to be associated with disability (see Supplemental files, Table A for details) (Amilon et al. 2021; Christoffersen 2019; Dean et al. 2018). This measure of worsening health is not without limitations, since hospital care can, in some cases, contribute to health improvements. However, we consider significant health improvements less likely given the chronic nature of the conditions in our measure. We also recognize that hospital visits for these conditions may not always signify new or deteriorating health issues, as individuals with these diagnoses may require regular check-ups or routine management. Nonetheless, we contend that, on average, individuals receiving hospital care for these diagnoses are more likely to experience declining health than those who do not.
While the ICF defines participation restrictions as problems that individuals with disabilities experience in involvement in life situations and activity limitations as difficulties individuals experience when performing activities (WHO 2013), a clear distinction between activity and participation has not been defined (Badley 2008). According to the WHO (2002) participation restrictions may include unemployment or not going to school due to disability-related prejudice, stereotypes or stigma, whereas activity limitations may involve being incapable of going out alone or of using public transportation.
We operationalize participation restrictions through respondents’ participation in employment or education. Employment and education are both key social policy success domains for which the equal participation of people with disabilities is protected by the UNCRPD. In Denmark, education is available free of charge and labor market participation rates are high for both men and women. In the period under study, unemployment rates were low (around 4% of the workforce (Statistics Denmark 2023)). Thus, restricted participation in education or employment may be associated with the experience of disability. We retrieved information from the Danish national registers on i) employment or ii) receipt of government student grant in week 12 (late March) each year from 2017–2020. We use this information to define restricted participation as being employed or in education for no more than one of the four indicator weeks versus being employed or in education for 2–4 indicator weeks, in the years 2017–2020.
We operationalize activity limitations through self-reported information from the question: ‘Do you experience problems with physical access to buildings that everybody uses?’ Not being able to access such buildings may hinder individuals from performing instrumental activities of daily living (i.e., activities that are necessary to lead an independent life, such as grocery shopping and moving about in one’s neighborhood). We construct a measure for the change in perceived activity limitations from 2016 to 2020 using the categories ‘both years,’ ‘2016 only,’ ‘2020 only’ and ‘neither year.’
We include registry-based socio-demographic information on sex (female, male), age and origin (non-immigrant vs. first- or second-generation immigrant) as control variables in the analyses.
Analytic strategy
We first present descriptive statistics on the shares of respondents with no, changing and consistent disability by survey instrument (the WG-SS and GALI), as well as background information on the analytical sample—in total and by disability status group. Second, we run linear probability regression models, estimated by ordinary least squares, to investigate how changes in disability status are associated with changes in health, participation restrictions and activity limitations. In the regression models, we compare individuals based on their disability status in 2016. Thus, for each survey instrument, we compare i) those who never experience disability to those with emerging disability and ii) those with consistent disability to those with declining disability. We investigate if and how individuals with changing disability status are different from individuals with consistent disability status in terms of changes in (self-reported and registry-based) health, participation and activity limitations. As the timing of the observed changes in disability, participation and health are unknown to us, and as variables that simultaneously influence health, disability and participation may be excluded from our model, coefficients cannot be interpreted as causal effects. We test for statistical differences in coefficients between models with z-tests (Kunzmann, Little and Smith 2000). As hypotheses 1 and 3b specify a direction (i.e., stronger associations for one of the measures), we apply one-sided z-tests with critical value 1.645 for p < 0.05. We consider and discuss only the differences between the two measures that are statistically significant. We use STATA version 18.0 for all analyses.
Results
Changes in disability status from 2016 to 2020
Table 2 shows percentages of respondents in the four disability status groups (never, declining, emerging, consistent) for the WG-SS and GALI, respectively. For the WG-SS, 82.8% of respondents never report a disability, and 5.2% have a consistent disability. A total of 6.7% have an emerging disability, whereas 5.2% have a declining disability.
Table 2
Descriptive statistics.
| VARIABLE | N | PERCENTAGE |
|---|---|---|
| Disability status changes for the WG-SS, 2016–2020 | ||
| Never (no → no) | 8,767 | 82.8 |
| Changing: Declining (yes → no) | 561 | 5.2 |
| Changing: Emerging (no → yes) | 714 | 6.7 |
| Consistent (yes → yes) | 553 | 5.2 |
| Disability status changes for GALI, 2016–2020 | ||
| Never (no → no) | 9,625 | 91.0 |
| Changing: Declining (yes → no) | 366 | 3.5 |
| Changing: Emerging (no → yes) | 421 | 4.0 |
| Consistent (yes → yes) | 165 | 1.6 |
| Female | 5,956 | 56.3 |
| Age in 2020 (years) | ||
| 20–24 | 675 | 6.4 |
| 25–34 | 1,232 | 11.6 |
| 35–44 | 1,710 | 16.2 |
| 45–54 | 3,083 | 29.1 |
| 55–64 | 3,886 | 36.7 |
| Danish-born | 10,040 | 94.8 |
| Immigrants (first and second generation) | 546 | 5.2 |
| Self-reported health status change in the last 3 years | ||
| Better | 1,623 | 15.3 |
| Same | 6,269 | 59.3 |
| Worse | 2,685 | 25.4 |
| Registry-based health status change | ||
| No disability-related hospital contacts, 2017–2020 | 9,284 | 87.7 |
| Disability-related hospital contact(s), 2017–2020 | 1,302 | 12.3 |
| Participation restrictions, 2017–2020 | ||
| Restricted (employed or in education <2 years) | 1,135 | 10.7 |
| Not restricted (employed or in education ≥2 years) | 9,451 | 89.3 |
| Activity limitation (Experienced limitations in access to buildings) | ||
| Both 2016 and 2020 | 247 | 2.3 |
| Only 2016 | 604 | 5.7 |
| Only 2020 | 598 | 5.7 |
| Neither year | 9,112 | 86.3 |
[i] Note: 10,586 individual respondents in total. The number of respondents is lower for some of the variables.
For GALI, 91% of respondents never report a disability, and 1.6% consistently report a disability. Moreover, 4% report an emerging disability, and 3.5% report a declining disability. Thus, as hypothesized (H1), the percentages with changing disability status are lower for GALI (7.5%) than for the WG-SS (11.9%). For both measures, the proportions with changing disability are larger than the proportions with consistent disability.
While the proportions of respondents with changing disability status are substantial for both measures, they differ in the shares that are defined as having a consistent, declining, emerging or no disability. Cross-tabulation between disability-status changes for the two measures (see Supplemental files, Table B.1) demonstrates a limited overlap between them. For instance, only 0.8% of the sample have a consistent disability according to both measures, whereas 0.6% and 1% have a declining and emerging disability, respectively. These figures reveal large discrepancies between the two measures in which respondents are defined as having a consistent, declining or emerging disability.
Overall, the analytical sample is largely representative of the Danish population, although older and female respondents are slightly overrepresented.
Tables 3 and 4 present descriptive statistics by disability status for the WG-SS and GALI, respectively. The tables reveal an association between disability persistence and adverse outcomes for both measures. People with consistent disability are the most likely to have had at least one disability-related hospital contact and to have experienced participation restrictions and activity limitations in between survey waves. Conversely, people without disability are the least likely to report such adversities, whereas the groups with emerging and declining disability come ‘in between’ the group with consistent disability and the group without disability. These results suggest that people with consistent disability status are more disadvantaged than those with changing disability status.
Table 3
Descriptive results, WG-SS, by disability status group.
| VARIABLE | NEVER | DECLINING | EMERGING | CONSISTENT |
|---|---|---|---|---|
| Female | 55.4 | 61.7 | 61.2 | 57.3 |
| Ref. | ** | ** | Ns. | |
| Age (years) | ||||
| 20–34 | 18.8 | 16.9 | 14.7 | 10.8 |
| 35–44 | 16.8 | 13.2 | 11.8 | 14.5 |
| 45–54 | 29.1 | 28.5 | 29.4 | 29.5 |
| 55–64 | 35.3 | 41.4 | 44.1 | 45.2 |
| Ref. | * | *** | *** | |
| Origin | ||||
| Danish born | 95.2 | 94.9 | 93.0 | 91.0 |
| Immigrants, descendants | 4.8 | 5.1 | 7.0 | 9.0 |
| Ref. | Ns. | ** | *** | |
| Participation restriction | ||||
| Restricted | 8.0 | 20.9 | 18.1 | 34.9 |
| Not restricted | 92.0 | 79.1 | 81.9 | 65.1 |
| Ref. | *** | *** | *** | |
| Self-reported health status change | ||||
| Better | 15.6 | 19.8 | 11.1 | 12.1 |
| Same | 63.3 | 44.7 | 40.2 | 34.4 |
| Worse | 21.1 | 35.5 | 48.7 | 53.5 |
| Ref. | *** | *** | *** | |
| Registry-based health | ||||
| No contact | 91.2 | 76.0 | 74.1 | 62.0 |
| At least one contact | 8.8 | 24.0 | 25.9 | 38.0 |
| Ref. | *** | *** | *** | |
| Activity limitation | ||||
| Both years | 1.3 | 3.5 | 4.1 | 15.9 |
| 2016 only | 5.3 | 8.2 | 6.3 | 9.2 |
| 2020 only | 4.6 | 6.6 | 13.0 | 11.9 |
| Never | 88.8 | 81.8 | 76.6 | 62.9 |
| Ref. | *** | *** | *** | |
| Observations | 8,767 | 551 | 714 | 553 |
[i] Note: 10.585 individual respondents in total. The number of respondents is lower for some of the variables. The significance tests relate to the difference between the reference group (Never) and respectively ‘declining,’ ‘emerging’ and ‘consistent.’ *p < 0.05, **p < 0.01, ***p < 0.001.
Table 4
Descriptive results, GALI, by disability status group.
| VARIABLE | NEVER | DECLINING | EMERGING | CONSISTENT |
|---|---|---|---|---|
| Female | 55.9 | 60.1 | 62.2 | 53.9 |
| Ref. | Ns. | * | Ns. | |
| Age (years) | ||||
| 20–34 | 18.4 | 15.3 | 14.3 | 9.1 |
| 35–44 | 16.2 | 13.7 | 17.1 | 19.4 |
| 45–54 | 28.9 | 34.4 | 29.2 | 29.7 |
| 55–64 | 36.5 | 36.6 | 39.4 | 41.8 |
| Ref. | Ns. | Ns. | * | |
| Origin | ||||
| Danish born | 94.8 | 95.9 | 95.7 | 92.7 |
| Immigrants and descendants | 5.2 | 4.1 | 4.3 | 7.3 |
| Ref. | . | . | . | |
| Participation restriction | ||||
| Restricted | 8.9 | 26.2 | 21.4 | 55.8 |
| Not restricted | 91.1 | 73.8 | 78.6 | 44.2 |
| Ref. | *** | *** | *** | |
| Self-reported health status change | ||||
| Better | 15.3 | 29.0 | 8.1 | 9.1 |
| Same | 62.3 | 38.9 | 22.6 | 22.4 |
| Worse | 22.5 | 32.1 | 69.4 | 68.5 |
| Ref. | *** | *** | *** | |
| Registry based health | ||||
| No contact | 89.6 | 74.3 | 68.4 | 54.5 |
| At least one contact | 10.4 | 25.7 | 31.6 | 45.5 |
| Ref. | *** | *** | *** | |
| Activity limitation | ||||
| Both years | 1.6 | 7.7 | 5.9 | 26.7 |
| 2016 only | 5.4 | 9.9 | 7.8 | 9.7 |
| 2020 only | 5.0 | 7.9 | 13.3 | 18.2 |
| Never | 88.0 | 74.5 | 72.9 | 45.5 |
| Ref. | *** | *** | *** | |
| Observations | 9,625 | 366 | 421 | 165 |
[i] Note: 10.577 individual respondents in total. The number of respondents is lower for some of the variables. The significance tests relate to the difference between the reference group (Never) and respectively ‘declining,’ ‘emerging’ and ‘consistent.’ *p < 0.05, **p < 0.01, ***p < 0.001.
Moreover, for both measures, we find systematic socio-demographic differences between disability status groups. People with changing disability status are more likely to be women whereas people with consistent disability status are more likely to be first- or second-generation immigrants, as compared to people without disability. In addition, regardless of disability duration, people with disability are on average older than people without disability. Thus, people with emerging, declining and consistent disability constitute distinct groups with different demographic and health profiles and with different experiences as regards societal in- and exclusion.
Emerging disability and health, participation and activity limitations
We now turn to the associations between changes in self-reported or registry-based health, participation restrictions and activity limitations and emerging disability status (as opposed to no disability) (Table 5). The results demonstrate that worsening health is associated with a higher probability of emerging disability. Thus, for both the WG-SS and GALI, worsening self-reported health or having had at least one hospital contact pertaining to disability between survey waves is associated with a transition from no disability in 2016 to disability in 2020 (H2).
Table 5
Regression results—association between emerging disability and changes in registry-based or self-reported health, participation and activity limitations.
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| WG-SS—EMERGING VS. NEVER | GALI—EMERGING VS. NEVER | WG-SS—EMERGING VS. NEVER | GALI—EMERGING VS. NEVER | |
| Registry-based health status | ||||
| No disability-related contact (ref.) | . | . | . | . |
| At least one contact | 0.12*** (0.01) | 0.08*** (0.01) | . | . |
| Self-reported health status | ||||
| Better | . | . | 0.01 (0.01) | 0.01 (0.00) |
| Same (ref.) | . | . | . | . |
| Worse | . | . | 0.10*** (0.01) | 0.10*** (0.01) |
| Participation | ||||
| Restricted | 0.07*** (0.01) | 0.05*** (0.01) | 0.07*** (0.01) | 0.05*** (0.01) |
| Not restricted (ref.) | . | . | . | . |
| Activity limitation in 2016 or 2020 | ||||
| Both years | 0.12*** (0.03) | 0.09*** (0.03) | 0.12*** (0.03) | 0.08*** (0.02) |
| 2016 only | 0.02 (0.01) | 0.02* (0.01) | 0.02 (0.01) | 0.02 (0.01) |
| 2020 only | 0.11*** (0.02) | 0.06*** (0.01) | 0.10*** (0.02) | 0.06*** (0.01) |
| Neither year (ref.) | . | . | . | . |
| Observations | 9,446 | 10,010 | 9,438 | 10,002 |
[i] Robust standard errors in parentheses. All models include a constant term and controls for sex, age group and origin (not shown). *p < 0.05, **p < 0.01, ***p < 0.001.
Moreover, experiencing participation restrictions and/or activity limitations is associated with a higher probability of emerging disability for both survey measures (H3a). Thus, we find a positive association between restricted participation between survey waves and emerging disability, as well as a positive association between experiencing activity limitations in both years, or in 2020 only, and emerging disability. Contrary to our hypothesis (H3b), these associations are of a similar magnitude or significantly larger for the WG-SS than for GALI (i.e., the variable ‘2020 only,’ Δz = 2.2, p < 0.01).
Declining disability and health, participation and activity limitations
We now turn to the associations between changes in health, participation restrictions and activity limitations and declining disability—that is, a change from disability in 2016 to no disability in 2020 as opposed to experiencing a disability in both years (Table 6). We find the expected negative association between having had at least one disability-related hospital contact in the period and declining disability status (models 1–2). Models 3 and 4 show parallel results when using self-reported data to measure health changes. Thus, for both health measures declining health is associated with a lower probability of declining disability status (H2).
Table 6
Regression results—association between declining disability and changes in registry-based or self-reported health, participation and activity limitations.
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| WG-SS—DECLINING VS. CONSISTENT | GALI—DECLINING VS. CONSISTENT | WG-SS—DECLINING VS. CONSISTENT | GALI—DECLINING VS. CONSISTENT | |
| Registry-based health status | ||||
| No disability-related contact (ref.) | . | . | . | . |
| At least one contact | –0.11*** (0.03) | –0.12** (0.04) | . | . |
| Self-reported health status change | . | . | ||
| Better | . | . | 0.03 (0.04) | 0.07 (0.04) |
| Same (ref.) | . | . | . | . |
| Worse | . | . | –0.13*** (0.03) | –0.22*** (0.04) |
| Participation | ||||
| Restricted | –0.09* (0.03) | –0.19*** (0.04) | –0.10** (0.03) | –0.19*** (0.04) |
| Not restricted (ref.) | . | . | . | . |
| Activity limitation in 2016 or 2020 | ||||
| Both years | –0.32*** (0.04) | –0.31*** (0.07) | –0.31*** (0.04) | –0.27*** (0.07) |
| 2016 only | –0.08 (0.05) | –0.08 (0.06) | –0.06 (0.05) | –0.06 (0.06) |
| 2020 only | –0.17*** (0.05) | –0.24*** (0.07) | –0.15** (0.05) | –0.19** (0.06) |
| Neither year (ref.) | . | . | . | . |
| Observations | 1,101 | 530 | 1,100 | 529 |
[i] Robust standard errors in parentheses. All models include a constant term and controls for sex, age group and origin (not shown). *p < 0.05, **p < 0.01, ***p < 0.001.
Moreover, we find a negative association between restricted participation and declining disability status. That is, those who experienced restricted participation in employment or education in between survey waves are less likely to transition out of disability as compared to those who did not experience restricted participation (H3a). Results are parallel for activity limitations, as we find negative associations between experiencing activity limitations in both years, or in 2020 only, and declining disability status (H3a).
The results are mixed as regards our hypothesis of a stronger association for GALI than the WG-SS (H3b). Contrary to our hypothesis, the coefficients for activity limitations are of a similar magnitude for GALI and the WG-SS. However, as hypothesized, restricted participation is more strongly associated with declining disability status for GALI (–0.19) than for the WG-SS (–0.10) (Δz = 1.8, p < 0.05))—a finding that may be due to the incorporation of participation restriction in the GALI question (i.e., if respondents have experienced limitations in activities people usually do).
Discussion
This study is the first to investigate and compare the prevalence of changes in disability status over time, as reflected in responses to two international disability survey measures: the Washington Group Short Set (WG-SS) and the Global Activity Limitation Indicator (GALI). Our analyses demonstrate that changes in the experience of disability, as indicated by response patterns for GALI and the WG-SS are common. Indeed, for both measures, changes in disability status are more common than consistent disability—a finding that corroborates those from studies on other disability survey measures. For instance, more than half of respondents who reported a disability in the American Current Population Survey did so inconsistently (Ward et al. 2017). Our findings point to the methodological challenges that relate to the measurement of disability—challenges that arise because disability, due to its inherent combination of individual and societal dimensions, is difficult to define both theoretically and conceptually (Hugaas and Tøssebro 2012).
As we hypothesized (H1), the share of respondents with changing disability status is larger for the WG-SS than for GALI—a finding that may be due to the incorporation of duration in the GALI survey question (i.e., asking whether respondents have been limited For at least the past 6 months…). Moreover, the overlap between the measures when it comes to the experience of changing or consistent disability is limited. Thus, those who experience a (changing or consistent) disability as defined by one survey instrument most likely do not experience a (changing or consistent) disability as defined by the other. Studies have demonstrated a limited overlap between the two measures in a cross-sectional context (Amilon et al. 2021). Our longitudinal findings establish that the choice of survey instrument is decisive, not only for who is defined as having a disability but also for whether that disability is consistent or not.
In parallel to previous research (Myers et al. 2020; Ward et al. 2017), and as hypothesized (H2), we find that changing disability status for both measures is associated with the corresponding changes in health. Thus, worsening health—if self-reported or identified via medical records—is associated with a higher probability of emerging disability (as compared to no disability) and a lower probability of declining disability (as compared to consistent disability).
Moreover, our study is the first to investigate if changes in disability status correlate with changes in participation restrictions and activity limitations. As hypothesized (H3a), we find that those who experience participation restrictions or activity limitations have a higher probability of transitioning into disability and a lower probability of transitioning out of disability. As we control for health, or health status changes, these findings do not simply reflect increased societal inclusion due to improved health (or, vice versa, decreased societal inclusion due to worsening health). Rather, these findings indicate that improving access to the built environment and removing barriers to participation in education and employment may not only improve the situation of people with disabilities; it may also reduce the share of people who experience disabilities in the population.
While we find some support for a stronger negative association between restricted participation and declining disability for GALI than for the WG-SS (H3b)—that may be due to the incorporation of participation restrictions in the GALI survey question—we generally find few systematic differences in associations between the two measures. Thus, while the measures result in different disability rates, they consistently show similar patterns in the associations between changes in disability status and changes in health, participation restrictions and activity limitations. Moreover, our findings demonstrate that changes in disability status, for both measures, reflect real changes in the health and societal participation of individuals. Our results are in line with Barnartt (2016), who views disability as a fluid state that may vary over time and place and with the individual’s statuses and roles within the social and physical environment. These results beg the question of whether the current understanding of disability as largely being chronic is limiting, as it may not capture the experiences of many people with disabilities.
Alternatively, one might argue that the populations defined through the WG-SS and GALI encompass different groups of people with disability, where those with changing disability status may have various types of (long-standing) health problems, rather than a chronic disability. Notwithstanding that some health problems can be strongly disabling for an extended period of time, our findings demonstrate that people with changing disability status are distinctly different—and in a relatively more advantageous situation—as compared to people with consistent disability.
Our findings thus demonstrate that the populations captured by the survey measures comprise heterogeneous groups of individuals with different types of disability profiles. Depending on the objective of the policy or research question that is posed, the inclusion of people with changing disability status in the population with disability may be more or less appropriate. In many cases, being able to distinguish between people with changing and consistent disability will be desirable, as their needs, opportunities and societal position, for example, are likely to differ. However, neither the WG-SS nor GALI provide users with tools for separating between the two groups. As a minimum, policy makers and researchers that use the WG-SS and GALI to identify people with disabilities should be aware of the heterogeneities within the resulting population. Combining disability information from several survey waves or asking supplementary questions in surveys about the duration of disability can be useful in identifying individuals with consistent disability, who may be in a particularly exposed situation.
Notwithstanding the important and novel findings of this study, it has limitations that need to be considered. First, by measuring disability status with a four-year interval, we cannot account for changes in disability status that respondents may have experienced in between survey waves. Studies indicate that disability status changes may take place with shorter time intervals (Myers et al. 2020). Thus, some of the respondents in the present study probably experienced additional disability status changes that we cannot account for. Hence, we may underestimate the share of respondents with changing disability status.
Second, conversely, we may overestimate the share of respondents with changing disability status by defining those who experienced a disability in 2020 but not in 2016 as experiencing emerging disability, as this transition to disability may be permanent for some respondents (i.e., they may not transition out of disability again).
Third, individuals with the most severe disabilities may not be able to participate in a survey or may be more likely to drop out, leading us to underestimate the shares with consistent disability. As we do not know the magnitude of these potential sources of bias, we cannot determine whether and to which extent they counterbalance. Nevertheless, they are likely to influence both the investigated measures in a similar manner. Therefore, we do not expect them to influence the overall conclusions as regards the limited differences between them.
Fourth, our measures of (registry-based) health, participation restrictions and activity limitations are crude and likely do not fully capture respondents’ experiences of health changes or changes in societal in- or exclusion. In particular, defining activity limitations as not being able to access buildings that everybody uses is restrictive, as it most likely will not encompass all activity limitations that are experienced by people with disabilities. Nevertheless, we find parallel results for the self-reported and the registry-based health measure, as well as the expected associations between changes in activity limitations and participation restrictions and changes in disability status. Potentially, we might have found even stronger associations between changes in activity limitations and participation restrictions and changes in disability status had we been able to include more comprehensive and precise measures.
Finally, as the exact timing of the observed changes in disability, participation and health are unknown to us, and as variables that simultaneously influence health, disability and participation may be excluded from our model, we cannot disentangle the causality of our results. Thus, we cannot tell whether, for example, a change in disability leads to a change in participation or the other way around. Nevertheless, our findings are the first to establish significant associations between changing disability status and the corresponding changes in participation restrictions and activity limitations for two internationally recognized and widely used disability measures.
Conclusion
For both the GALI and the WG-SS survey instrument, a significant proportion of respondents with disability experience a change in disability status across survey waves. Changes in disability status reflect real changes in health, activity limitations and participation restrictions. These results stress the importance of removing barriers to societal participation for people with disabilities. Moreover, including questions about the duration of disability in surveys can help identify respondents with consistent disability, who may be subject to accumulated disability-related disadvantage and hence may be in a particularly exposed situation.
While we establish independent and stable associations between health, activity limitations and participation restrictions on the one hand and changes in disability status on the other, we cannot establish the causal relationship between these factors. While disentangling causality in a disability context is challenging (Loprest and Maag 2007; Shandra 2018; Zhang, Hayward and Yu 2016), this is an important topic for future research.
Additional File
The additional file for this article can be found as follows:
Competing Interests
The authors have no competing interests to declare.
Author Contributions
The first author conceptualized the study and wrote the original draft. The second author analyzed the data and critically revised the original draft. Both authors contributed to the methodological design and interpretation of findings, approved the final version of the manuscript and agreed to be accountable for all aspects of the work.