Table 1 Studies estimating the validity or reliability of the Global Activity Limitation Indicator (GALI) by study design, period 2000–2017

Van Oyen et al. 2006 [32]

National HIS* 2001 Belgium Population aged 15+ N = 9168

Self-reported: ADL*, SF-36* physical domain score, number of self-reported chronic physical conditions out of a list of 29 conditions, number of mental conditions (depression, anxiety, somatization, sleep disorders) from the SCL-90R*, GHQ-12* scale for mental wellbeing and CMI*

Yes

Predicted probability distribution of GALI distribution, POR* from proportional odds models, heterogeneity across demographic variables

• The probability distribution of GALI by severity level fits appropriately against indicators measuring mental and physical illness both in subjects with or without ADL limitations;

• 95% of subjects without ADL limitations and no mental or physical health problems do not report participation restrictions;

• Subjects with ADL limitations report participation restrictions and the severity level of participation restriction is higher in function of the level of severity and the number ADL limitations and there is no evidence for heterogeneity across gender, age, education and language;

• The probability distribution of GALI by severity level is associated with the different physical and mental morbidity measures; A dose-response relationship is observed;

• The measures of associations are not as strong for mental health problems compared to physical health problems.

Cabrero-Garcia et al. 2014 [31]

National HIS 2006 Spain Population aged 65+ N = 7835

Self-reported: physical and mental morbidity, functional disability (ADL, IADL and mobility)

Concurrent comparison of the associations of GALI and the association of SRH* with the health outcomes:

FCI*, GHQ-12 scale for mental wellbeing, Functional disability*

Yes

Spearman correlation, predicted probability of GALI from fractional polynomial models, MOR* from multinomial logistical regression

• GALI is primarily a measure of functional status and is secondarily associated with physical and mental morbidity whereas for SRH physical morbidity and to a lesser extend mental morbidity are the main correlates;

• The odds of having (severe) participation restrictions increased with the level of functional limitations (number), the physical and mental morbidity, suggesting a dose-response relationship;

• Mental morbidity was as strong a correlate of GALI as of SRH, whereas physical morbidity was less strong a correlate of GALI compared to SRH.

Jagger et al. 2010 [34]

International SHARE* 2004 11 EU countries Population aged 50+ N = 27340

Measurement: maximum grip strength and walking speed (in subjects aged 75+)

Self-reported: ADL, IADL, and walking limitations

No

Predicted probability distribution of GALI,

ORs from logistic regression models,

Random-effects meta-analysis to assess heterogeneity of associations between countries

• GALI effectively capture disability as measured by both the self-reported as objective measures of functional limitations;

• The likelihood of reporting participation restrictions increases as the severity of functional limitations increases in both the self-reported as objective measures of functional limitations;

• The likelihood of reporting no participation restriction in subjects with limitations is non-zero, though small and a minimum for the most severe measure, the ADL;

• Cross-country comparison did not provide any evidence for heterogeneity for the OR of having participation restrictions in function of the self-reported ADL and the objective measures’;

• In all countries, the odds of having participation restrictions was higher in subjects with IADL limitations. The size of the effect was however more pronounced in some countries compared to others.

Berger et al. 2015 [33]

International EHIS* 2007–2010 14 EU countries Population aged 15+ N = 152,796

Self-reported: ADL, IADL and functional limitations

No

Predicted probability distribution of GALI,

ORs from logistic regression models,

Random-effects meta-analysis to assess heterogeneity of associations between countries

• GALI is significantly associated with ADL and IADL limitations and functional limitations;

• The likelihood of reporting participation restrictions increases as the number of ADL and IADL limitations and the severity of functional limitations increased;

• The likelihood of not reporting participation restrictions decreases as the number of ADL and IADL limitations and the severity of functional limitations increased;

• In all countries, the odds of having participation restrictions was higher in subjects with ADL, IADL and functional limitations. The size of the effect was more pronounced in some countries compared to others providing evidence for heterogeneity of the effect size.

Berger et al. 2015 [36]

National HIS 2001 linked with mortality and migration database (National Register), 2001–2010 Belgium Population aged 15+ N = 8583, 902 deaths

Mortality (follow-up to 10 years)

Yes

MRRs* from Poisson regression models;

Comparison of relative predictive ability of GALI compared to SRH*

• Compared to individuals without participation restrictions, subjects with moderate or severe participation restriction have a 1.8 to 3.0 increased mortality rate over the 10 years of follow-up;

• The effect does not vary significantly by gender, education or age, except in subjects under age of 50 years;

• SRH and GALI are complementary predictors of mortality, with some indications of a stronger effect of SRH;

• The predictive effect of SRH and GALI slightly decrease over time.

Van der Heyden et al. 2015 [37]

National HIS 2008 linked with mortality within the Health Insurance database, 2008–2010 Belgium Population aged 65+ N = 1894, 178 deaths

Mortality (follow-up to 2 years)

No

MRRs from Poisson regression models;

Comparison of relative predictive ability of GALI compared to SRH

• Subjects with participation restriction have a 2.4 increased mortality rate over the 2 years of follow-up;

• The effect does not vary by gender;

• In men, SRH and GALI are complementary predictors of mortality, whereas in women this is only so for GALI.

Van der Heyden et al. 2015 [35]

National

HIS 2008 linked with Health Insurance database including expenditure in 2008–2010 Belgium Population aged 15+ N = 7286

Health care expenditure (Health insurance, out of-pocket, supplement)

Yes

Linear regression after logistic transformation of costs;

Cost ratios were estimated to compare expenses to a reference;

Decomposition of differences in expenses using the Blinder-Oaxaca method

• Moderate and severe participation restriction increases all health expenses by 3 to 6-times;

• The increase is the more pronounced in the reimbursed health care expenditure;

• In absence of any chronic condition, moderate and severe participation restriction increases all health care expenditure by 2.5 to 4.5 times;

• Chronic conditions explain only 22% of the differences in health care expenditure by level of participation restriction.

Verropoulou et al. 2015 [38]

International SHARE* 2004 with follow-up to re-interview in wave 2006 /2007 11 EU countries Population aged 50+ N = 17,941, 696 deaths

Mortality (follow-up 2 to 3 years)

Yes

Hazard ratios from Cox proportional hazard models; Comparison of relative predictive ability of GALI compared to SRH

• Both GALI and SRH are significant predictors of mortality in separate models;

• When adjusting for specific health indicators (asthma, cancer, depression, mobility, IADL, orientation), GALI and SRH (only men) were significant but the magnitude diminished;

• GALI and SRH add information on top of specific health indicators;

• When GALI and SRH are included in one model, GALI was only significant in women, suggesting a partial conceptual overlap as there is a correlation between GALI and SRH;

• SRH and GALI represent different aspects of health.

Cox et al. 2009 [39]

National Food Consumption Survey 1st and 2nd visit Belgium, Population aged 15+ N = 170

Twice self-reported GALI within time window between 11 and 55 days

Yes

Pearson correlation coefficients, weighted Kappa coefficients

• Both Pearson (0.73) and Kappa coefficient (0.68) indicate an acceptable reliability;

• Agreement is significantly higher for males (Kappa = 0.82) compared to females (Kappa = 0.54);

• Agreements did not differ by education level, age, time span and language (French, Dutch).