Trends in inequalities in disability in Europe between 2002 and 2017

Background Monitoring socioeconomic inequalities in population health is important in order to reduce them. We aim to determine if educational inequalities in Global Activity Limitation Indicator (GALI) disability have changed between 2002 and 2017 in Europe (26 countries). Methods We used logistic regression to quantify the annual change in disability prevalence by education, as well as the annual change in prevalence difference and ratio, both for the pooled sample and each country, as reported in the European Union Statistics on Income and Living Conditions (EU-SILC) and the European Social Survey (ESS) for individuals aged 30–79 years. Results In EU-SILC, disability prevalence tended to decrease among the high educated. As a result, both the prevalence difference and the prevalence ratio between the low and high educated increased over time. There were no discernible trends in the ESS. However, there was substantial heterogeneity between countries in the magnitude and direction of these changes, but without clear geographical patterns and without consistency between surveys. Conclusions Socioeconomic inequalities in disability appear to have increased over time in Europe between 2002 and 2017 as per EU-SILC, and have persisted as measured by the ESS. Efforts to further harmonise disability instruments in international surveys are important, and so are studies to better understand international differences in disability trends and inequalities.


Online Resource 3 Table S3.1 Global Activity Limitation Indicator (GALI) Comparability to standard question, by country (27 countries) and year -European Union Statistics for Income and Living Conditions (EU-SILC 2005-2017)
GALI comparability for 2005-2012: obtained from the EUROSTAT document "Overview of the implementation of the GALI question in EU-SILC" https://circabc.europa.eu/webdav/CircaBC/ESTAT/health/Library/working_group_2012/documents/Item%209.1%20HLY%2 0annex%202%20-%20overview%20tables%20and%20notes.pdf From year 2012-2016: comparability was obtained by extending the assessment done by the prior document if the question was not changed. This is based three sources of information. The first is the document from the European Health and Life Expectancy Information System (EHLEIS) "EU-SILC Health questions 2014-2016 in national languages and back translations by country experts" that can be obtained from: http://www.eurohex.eu/pdf/Reports_2018/2018_TR4%206_SILC%20Questions_Backtranslation.pdf The second is the document "Health questions from the Minimum European Health Module used in EU-SILC in the 27 countries", and can be obtained from: http://www.eurohex.eu/pdf/Reports_2014/2014_TR4%205_Health%20Questions.pdf From year 2015-2017: when additional information was necessary, country specific questionnaires were consulted from the GESIS Microdata Lab documentation, available at: https://www.gesis.org/en/missy/materials/EU-SILC/documents/questionnaires Comparability was assessed based on the 4 GALI question criteria : 1)being limited, 2) in activities people do , 3) because of a health problem, 4) for the last 6 months. Countries that had 4/4 were deemed comparable, while 3 partially comparable and below not comparable. If a question fulfilled 4 criteria, but was separated in filters (like the UK in last years) it was also set to partially comparable. This last criterion did not change the assessment for any country.

. Global Activity Limitation Indicator (GALI) change in question relative to prior year, by country and year -European Union Statistics for Income and Living Conditions (EU-SILC 2005-2017)
GALI comparability for 2005-2012: obtained from the EUROSTAT document "Overview of the implementation of the GALI question in EU-SILC" https://circabc.europa.eu/webdav/CircaBC/ESTAT/health/Library/working_group_2012/documents/Item%209.1%20HLY%2 0annex%202%20-%20overview%20tables%20and%20notes.pdf From year 2012-2016: comparability was obtained by extending the assessment done by the prior document if the question was not changed. This is based three sources of information. The first is the document from the European Health and Life Expectancy Information System (EHLEIS) "EU-SILC Health questions 2014-2016 in national languages and back translations by country experts" that can be obtained from: http://www.eurohex.eu/pdf/Reports_2018/2018_TR4%206_SILC%20Questions_Backtranslation.pdf The second is the document "Health questions from the Minimum European Health Module used in EU-SILC in the 27 countries", and can be obtained from: http://www.eurohex.eu/pdf/Reports_2014/2014_TR4%205_Health%20Questions.pdf From year 2015-2017: when additional information was necessary, country specific questionnaires were consulted from the GESIS Microdata Lab documentation, available at: https://www.gesis.org/en/missy/materials/EU-SILC/documents/questionnaires Deviations from the standard GALI question phrasing in EU-SILC vary by country and by year, and they vary between omitting dimensions of the question (i.e , 6 month duration, activities people do), using a filter question (i.e the UK between 2005-11, Portugal 2005 and changing wording of the question.   Estimates are obtained from sex stratified logistic models using microdata as dependent variable a dichotomous GALI indicator: logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) The inclusion of the GALI comparability variable in EU-SILC decreases the rate of change over 1 year of both prevalence and absolute educational inequalities. The average effect of the inclusion of the comparability variable in GALI prevalence is statistically significant reduction of 2.65 %-points for males and of 1.87%-points for females for partly comparable relative to the standard phrasing of the question. For not comparable questions, there is a statistically significant decrease of 1.15% in prevalence for males and of 1.26% for females .

Females -EU-SILC (High Educated) Females -ESS(High Educated)
Estimates are obtained from country, sex stratified logistic models using EU-SILC and ESS microdata as dependent variable a dichotomous GALI indicator, stratified by sex, survey and country: ESS and EU_SILC country stratified models logit(GALI) = b0 + b1(age)*(age) + b2( education) +b3(year) + b4(year)*education EU-SILC Corrected logit(GALI)= b0 + b1(age)*(age) + b2( education) +b3(year) + b4(year)*education + b7(GALI Comparability) We compute the average marginal effects for an increase of 1 year for each education level using the STATA command margins,dydx. These correspond to the annual change in prevalence presented. The annual average change in prevalence difference is estimated by subtracting the average marginal effects of the low and the high educated BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) For low educated men, countries where both surveys show a statistically significant increase in disability prevalence include the Netherlands, Belgium and Germany, while Spain and Hungary show a statistically significant decrease in prevalence. Countries where the trend for low educated men in both surveys is not statistically different from zero include Finland, the UK, France, Italy and Slovakia. For many countries, EU-SILC detects a significant trend for this group, while the ESS estimates are often not statistically significant. Women also show substantial heterogeneity between surveys and countries.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  Estimates are obtained from country, sex stratified logistic models using EU-SILC and ESS microdata as dependent variable a dichotomous GALI indicator, stratified by sex, survey and country: ESS and EU_SILC country stratified models Logit(GALI) = b0 + b1(age)*(age) + b2( education) +b3(year) + b4(year)*education EU-SILC Corrected Logit(GALI)= b0 + b1(age)*(age) + b2( education) +b3(year) + b4(year)*education + b7(GALI Comparability) We compute the average marginal effects for an increase of 1 year for each education level using the STATA command margins,dydx. These correspond to the annual change in prevalence presented. The annual average change in prevalence difference is estimated by subtracting the average marginal effects of the low and the high educated

Females-EU SILC (RII) Females -ESS (RII)
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance  (1) and (1b) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) (2) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) **Slope Index of Inequality (SII) and Relative Index of Inequality (RII) are obtained through generalized linear modals that include education (low, medium, high) and age in 5 year age categories(30-34; 35-39;…) as covariates and are stratified by year. Trends are obtained through linear regression of RII/SII as independent variable and year as independent variable using robust standard error. a European Union Statistics on Income and Living Conditions (EU-SILC) annual microdata between 2005 and 2017. Member countries in EU-SILC use variants of the GALI question over time. b European Social Survey (ESS) biannual microdata between 2002 and 2016.ESS uses the same version of the GALI question for all countries and years. This version omits the 6 month time frame of the standard GALI question. c GALI comparability estimates include baseline model plus a 3 level categorical variable relative to standard question phrasing (comparable, partially comparable, not comparable) GALI question change estimates include baseline model and a binary variable that takes the value of 1 if question changed prior to last year. Both GALI variables were obtained for 2005-2012 from the EUROSTAT document "Overview of the implementation of the GALI question in EU-SILC", and can be obtained from:

Global Activity Limitation Indicator (GALI) educational inequalities (prevalence difference, ratio, slope index of inequality and relative index of inequality) change in 1 year (ages 30-79) for the pooled sample of 26 European Countries from , by education, sex and survey (European Union Statistics on Income and Living Conditions 2005-2017; European Social Survey 2002-2016) and T-tests between survey estimates
https://circabc.europa.eu/webdav/CircaBC/ESTAT/health/Library/working_group_2012/documents/Item%209.1%20HLY%20annex%202%20-%20overview%20tables%20and%20notes.pdf d Average age-standardized GALI prevalence over the corresponding period for each survey using the EU 2013 standard population for all countries included in the sample. e Two sample t-test between the EU-SILC and ESS coefficients . Significant at the 95% level in bold BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)  Estimates are obtained from sex stratified logistic models using microdata as dependent variable a dichotomous GALI indicator: (1) and (1b) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) (2) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) a European Union Statistics on Income and Living Conditions (EU-SILC) annual microdata between 2005 and 2017. Member countries in EU-SILC use variants of the GALI question over time. b European Social Survey (ESS) biannual microdata between 2002 and 2016.ESS uses the same version of the GALI question for all countries and years. This version omits the 6 month time frame of the standard GALI question. c GALI comparability estimates include baseline model plus a 3 level categorical variable relative to standard question phrasing (comparable, partially comparable, not comparable) for EU-SILC only. * Models include the percentage of education measured for each country, survey, sex and year as part of the regressions. **Models exclude information prior to 2008 ***Models use the normalized probability weights provided by EU-SILC and ESS in the regression analyses. Main estimates use product of population and probability weighs in analyses. ****Models exclude Romania and Latvia BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Estimates are obtained from sex stratified logistic models using microdata as dependent variable a dichotomous GALI indicator: (1) and (1b) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) (2) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) a European Union Statistics on Income and Living Conditions (EU-SILC) annual microdata between 2005 and 2017. Member countries in EU-SILC use variants of the GALI question over time. b European Social Survey (ESS) biannual microdata between 2002 and 2016.ESS uses the same version of the GALI question for all countries and years. This version omits the 6 month time frame of the standard GALI question. c GALI comparability estimates include baseline model plus a 3 level categorical variable relative to standard question phrasing (comparable, partially comparable, not comparable) for EU-SILC only. * Models include the percentage of education measured for each country, survey, sex and year as part of the regressions. **Models exclude information prior to 2008 ***Models use the normalized probability weights provided by EU-SILC and ESS in the regression analyses. Main estimates use product of population and probability weighs in analyses. ****Models exclude Romania and Latvia BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Estimates are obtained from sex stratified logistic models using microdata as dependent variable a dichotomous GALI indicator: (1) and (1b) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) (2) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) a European Union Statistics on Income and Living Conditions (EU-SILC) annual microdata between 2005 and 2017. Member countries in EU-SILC use variants of the GALI question over time. b European Social Survey (ESS) biannual microdata between 2002 and 2016.ESS uses the same version of the GALI question for all countries and years. This version omits the 6 month time frame of the standard GALI question. c GALI comparability estimates include baseline model plus a 3 level categorical variable relative to standard question phrasing (comparable, partially comparable, not comparable) for EU-SILC only. * Models include the percentage of education measured for each country, survey, sex and year as part of the regressions. **Models exclude information prior to 2008 ***Models use the normalized probability weights provided by EU-SILC and ESS in the regression analyses. Main estimates use product of population and probability weighs in analyses. ****Models exclude Romania and Latvia BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s)

59
Estimates are obtained from sex stratified logistic models using microdata as dependent variable a dichotomous GALI indicator: (1) and (1b) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) (2) logit(GALI) = b0 + b1(age) + b2(age)(age) + b3( education) +b4(year) + b5(year)*education + b6(country) + b7(GALI Comparability) a European Union Statistics on Income and Living Conditions (EU-SILC) annual microdata between 2005 and 2017. Member countries in EU-SILC use variants of the GALI question over time. b European Social Survey (ESS) biannual microdata between 2002 and 2016.ESS uses the same version of the GALI question for all countries and years. This version omits the 6 month time frame of the standard GALI question. c GALI comparability estimates include baseline model plus a 3 level categorical variable relative to standard question phrasing (comparable, partially comparable, not comparable) for EU-SILC only. * Models include the percentage of education measured for each country, survey, sex and year as part of the regressions. **Models exclude information prior to 2008 ***Models use the normalized probability weights provided by EU-SILC and ESS in the regression analyses. Main estimates use product of population and probability weighs in analyses. ****Models exclude Romania and Latvia Table S11.