Measuring disability: a systematic review of the validity and reliability of the Global Activity Limitations Indicator (GALI)

Ageing of populations defies health and social policies. Population ill-health and especially disability are major challenges as there is currently no consistent evidence that the lengthening of life expectancy goes with a reduction in the total lifetime days of disability, the so-called compression of morbidity [1].

The concept of disability is complex and multidimensional. In initial medical models, disability was viewed as a problem residing solely in the persons affected. Disability referred to consequences of chronic or acute diseases or accidents on the functioning of specific body systems and on mental, physical and sensory functions in terms of (1) impairment or dysfunctions and structural abnormalities in specific body systems; (2) disability or restrictions in basic physical and mental actions and (3) handicaps or difficulties in doing activities of daily life [2‐5]. More biophysical-social models introduce the person-environment perspective of the disablement process: disability as the outcome of the interaction of a person and his environment [6] and the dynamics of disability which is affected by how a person’s capacity fits the environmental demand and results in participation [3]. Participation restriction is defined as limitations in the performance of roles and social involvement in different settings such as work and employment, school, leisure, parenting, housework, community, social and civic life [7]. Because participation is influenced by environmental factors and social norms, any measure of participation restriction cannot differentiate the impact of the impairment and functional limitations from the impact of accommodations and enabling environments [8, 9].

Disability can occur in any human activity and settings. Adding to this complexity, instruments measuring disability differ in the domains of functioning included, in their goals to measure either capacity (without any personal or equipment assistance) or performance (with assistance), or to measure also disability symptoms (pain, weakness, endurance, …), levels of severity or the duration of the disability. Traditional survey instruments measure a limited number of tasks (5 to 7) in the domain of personal care (ADL (Activity of Daily Living)) or in the domain of household management (IADL (Instrumental ADL)). Other survey instruments have either increased the number of disability questions by adding more and more tasks [10] or have developed short set of disability questions that have good coverage of activities [11, 12]. At the same time there has been a quest to measure disability with parsimony similar to the parsimony in measuring health using the global one-item survey instrument on self-rated health (SRH) [13, 14].

In response to the call for parsimony, a global survey instrument to measure disability, the Global Activity Limitation Indicator or GALI, was proposed. The development of GALI occurred in the framework of the creation of a coherent set of indicators to monitor health across Europe [15]. GALI was part of set of 10 survey instruments including three global one-item survey instruments were proposed covering distinct health concepts: perceived health, chronic morbidity and participation restriction [16]. The 3 global questions define the Minimum European Health Module (MEHM) [17]. At the time of the development of the GALI, the beta version on the International classification of Functioning, Disability and Health (ICF) served as conceptual framework [6, 18, 19]. Because of its implicit reference to the ability for societal participation in a variety of non-specified settings and non-specified domains of life (such as employment, school, housework, and leisure) using the wording “activities people usually do”, GALI is intended to be a global self-reported measure of participation restriction. Additional conceptual criteria were the health relatedness of the cause of disability, the generic normative comparison in the level of participation, the long-standing duration of the disability (a duration of at least 6 months) and the ability to measure levels of severity [15, 18]. To accommodate the results of cognitive testing in relation to the severity options in the answer categories, the wording “to what extent” was added in the final version [20]:

“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?

Being part of the MEHM, GALI is used in major European health and non-health surveys such as the European Health Interview Survey (EHIS), Survey on Income and Living Conditions (SILC) and the Survey of Health, Ageing and Retirement in Europe (SHARE). Since 2004, GALI is also the underlying measure of the European indicator “Healthy Life Years (HLY)”. HLY is a measure of disability free life expectancy and was presented in the set of structural indicators selected and defined to help measure progress in strategic European policies such as the 2000 Lisbon strategy and the European 2020 strategy on Active and Healthy Ageing [21]. HLY is one of the components of the Active Ageing Index [22]. At the national level, countries such as France, have selected HLY as one of their high level indicators for long term evaluation of their economic, social and environmental policies [23]. GALI also fits the requirement to follow-up European and United Nations disability policies, that stress the importance of full and effective participation as main policy outcome [24, 25]. More recently, GALI, as underlying health measure of HLY, contributes to the scoreboard indicators of the European Pillar of Social Rights [26] . Due to its high informational value, its relative simplicity and its compliance elaborated by the European Union, HLY has been proposed to be the instrument in designing social security solutions [27].

Given the use of GALI within the European Union, especially the fact that it is the measure underlying the European indicator HLY, the objective of current paper is to bring together what is known from published manuscripts on the validity and/or the reliability of GALI. Construct validity evaluation has been divided into translation validity, a more qualitative process and criterion validity, a quantitative approach [28]. Current review focus on the quantitative validity, as translation validity is linked to the conceptual criteria used for the GALI development [9, 15].

Two search strategies were combined by HVO and NB in January 2017 to identify peer reviewed manuscripts published in English with publication date 2000 (the period GALI was developed) or beyond. Following the PRISMA guidelines, manuscripts were independently evaluated by HVO and NB first on the titles and abstracts and in a second stage on the text. The result of the search and manuscript selection is summarized in a PRISMA flow chart (Fig. 1) [29]. First, PUBMED database (https://​www.​ncbi.​nlm.​nih.​gov/​pubmed) was used with {“global activity limitation”[All Fields] OR (GALI[All Fields] NOT GALI[Author])} as search protocol. Of the 69 publications 54 were excluded: content not related to the topic (e.g. Gali as part of a name of a butterfly “Calisto franciscoi Gali”), language other than English, Gali in name, email or contact address of one of the authors). Of the remaining 15 articles, 9 articles were retained as manuscripts studying the validity or reliability of GALI. A second search used the Google Scholar database (https://​scholar.​google.​com/​) using as search protocol “global activity limitation”. Of the 208 references 107 were excluded (content not related to the topic, language other than English, only an abstract, and citations). The remaining 101 articles were screened to identify manuscripts published with the objective to estimate the validation or reliability of the GALI (N = 11). Double publications were excluded (N = 2). The double publications were a result of the publication of institutional working papers or reports prior to the publication of a manuscript in a scientific journal. All 9 manuscripts retained were found in both databases. Manuscripts were grouped as reliability studies, concurrent validity studies (cross-sectional studies measuring an association) or predictive validity studies (ability of GALI to predict an outcome) [28]. We further distinguished between national and international studies, as international studies may be more sensitive to total survey error due, for example, to the lack of international harmonization at different stages of the study [30].

The classification of the manuscripts by type of study is given in Table 1.

Defining disability is not easy. Because it interweaves medical and social domains [40], the concept of disability has led to divergent interpretations and uses [41]. People may experience disability due to health in any human activity; yet, activities included in traditional instruments with focus on ADL and/or IADL cover only a fraction of all activities [12]. The challenges of measuring disability have been tackled by two distinct approaches. One option is to include more activities, more specific answer categories, more aspects such as disability symptoms and disability in more specific settings of life [10, 42, 43]. This option induces increased respondents burden, increased survey cost, more complex analysis in order to provide condensed indicators for end users. The other option seeks short sets and/or a one single global instrument [11, 15]. GALI has been nominated the champion in parsimony [12], but the lack of simplicity and the high density of concepts in one single question may hamper its acceptability [9, 44]. Three different alternatives (decomposing GALI using filtered and routed questions or through omitting features such as the duration of the disability and/or the health relatedness) and the original GALI were evaluated against the short version (including 4 functional limitation questions) of the Washington Group on Disability instrument [11]. The four variants were randomly assigned to survey participants (N = 3009). The results, indicating a substantial higher sensitivity of GALI, no evidence for a better understanding of the simplified alternatives but possibly, a small advantage in specificity when bringing the duration of the participation restriction to subsequent questions, should be balanced against the cost of breaking an established chronological series [45]. Currently, Eurostat follows for the upcoming EHIS wave III (2018–2019), the recommendation of the EHLEIS working group on the blueprint for an internationally harmonized Summary Measure of Population Health [9]. The EHLEIS working group, including experts from the EU, Japan, USA, OECD and WHO met 3 times (2012, 2013 and 2014) in Paris and proposed that of the different components of disability, participation restriction in the first place and, in addition, functional limitations should be the main goals for internationally harmonized global measures [9]. GALI fits the six conceptual characteristics specified by the working group: 1. comprehensive content of participation; 2. measure of participation performance with current accommodation; 3. health relatedness of the cause of participation restriction; 4. normative comparison in the level of participation; 5. long-term duration of restriction; 6. measure severity of restriction in the response scale (at least three levels). As mentioned above, this comes at a cost of lack of conciseness and simplicity.

Of the 3 global questions that constitutes the MEHM [15], studies reporting on the concurrent and predictive validity and reliability of SRH have the longest history [46‐48], while less evaluations have been done with respect to the global question on chronic disease [49].

In this manuscript, we summarize for the first time the current evidence of the validity of GALI including concurrent and predictive validity studies and reliability studies.

Current review has limitations. A first limitation of the review is that it only included peer-reviewed manuscripts published in English, identified using only one bibliographic database in addition to Google Scholar. Google Scholar was used to search also the grey literature. The authors were in close contact with the international research network on health expectancies and the disablement process (REVES). Members of the REVES network were invited to participate in a survey [9] with the aim to identify additional manuscripts. Secondly, all studies included rely on self-reported measures and the precision of the validity and reliability estimates relies upon accurate reporting. Thirdly, no quality related weighting was applied in describing the different manuscripts. E.g in contrast to the mortality follow-up in Belgium [36], the mortality follow-up within SHARE [38] did not use register data resulting in an under-numeration of the number of deaths. A possible effect of a selection bias on the predictive power of GALI and SRH on mortality cannot be excluded, but if any, the text of the manuscript claims it should be limited [38]. Next, the review was hampered by the methodological heterogeneity of the different studies: e.g. the association of GALI with health indicators in the two national concurrent validations studies [31, 32] was measured using different health indicators and different statistical modeling. Finally, the review was not registered.

To summarize, in none of the 9 studies included, there was evidence of a failing validity. The concurrent validity was evaluated in 4 studies. The two national concurrent validation studies indicated a dose-response effect by GALI severity level on the association with other health status measures with a somewhat weaker association related to the mental well-being score.The two international studies did not consider the GALI severity level. They concluded that the odds of reporting participation restriction were higher in subjects with self-reported or observed functional limitations. The strength of the association varied more strongly between the countries using the EHIS compared to the SHARE survey. The difference in homogeneity can in part be explained by the variation in the implementation of the EHIS, including different wording across countries [33]. International comparability of data submitted to Eurostat, including those on GALI is hampered as EU regulations does not include guidelines on the exact formulation of the questions within one and between surveys [20] nor on the data collection mode increasing the likelihood for differential total survey errors [30]. This under-valorizes the substantial efforts done by Eurostat to harmonize instruments in surveys [49]. Differences in accounting for the institutionalized population may further have affected the heterogeneity in the size of the association of the GALI with other health measures [50]. Although, without focus on validity, a recent study, using the 2013 Danish Health and Morbidity survey and 31 health-related indicators, corroborates the concurrent validity describing a trend towards poorer health and wellbeing in subjects with more participation restriction as well as a trend toward a less healthy life style or less contacts with social relations in function of more participation restrictions [51]. Using follow-up data, GALI proved both in national and international studies to be a consistent predictor of future health outcomes both in terms of mortality and health care expenditure. As predictors of mortality, the two distinct concepts - SRH and GALI - acted independently and complementary of each other. Only one reliability study was identified indicating a sufficient reliability of GALI.

The strength of GALI as an inclusive one-question instrument is that it fits all conceptual characteristics specified for a global measure on participation restriction. The review indicates that current version of GALI has a good and sufficient concurrent and predictive validity and reliability.