Estimating social inequalities in Healthy Life Years in Belgium
Estimating social inequalities in HLE:
Challenges and opportunities
10 February, 2012
Rana Charafeddine
Scientific Institute of Public Health
Outline
1. Background
2. Mortality follow up of survey data
3. Policy recommendations
1. Background
Social inequalities in HLY in Belgium
Inequalities in HLY are not static over time
In the last decade, HLY have evolved differently according to the educational level. This generated an increase in the gap between educational groups (Van Oyen et al., 2011)
For instance, the difference in HLY at 25 years among the highest and lowest educated men was 17.0 years in 1997 and became 18.6 in 2004
Among women, this difference was 11.4 in 1997 and became 18.2 in 2004
Educational inequalities in HLY among women at age 25, Belgium
Ref: Van Oyen et al. (2011), Eur J of Public Health
Education 1997 2004 Diff..
Higher education 44.73 47.10 2.37
Higher secondary 43.41 41.27 -2.14
Lower secondary 40.88 42.01 1.13
Primary education 34.70 36.27 1.57
No diploma 33.31 28.92 -4.39
Total 38.91 40.42 1.51
Estimation of HLY by SES
Prevalence of health status by SES Disability indicator Survey data (e.g. HIS, SILC)
Mortality rate by SES Numerator: number of deaths by education Denominator: number of person years by education Mortality follow up of census using a unique identifier Linked approach: golden standard Alternative for the census?
Project: HEALTHY LIFE EXP
Funded by the Belgian Science Policy in the context of the AGORA program
Proposed by the Federal Public Service Social Security
Institutions involved: ISP (Herman Van Oyen, Rana Charafeddine, Stefaan
Demarest) VUB (Patrick Deboosere, Sylvie Gadeyne)
Started in January 2010 until June 2011
Objectives
To explore different possible methods as alternative to the census to be used in Belgium to estimate and update HLY by SES.
Method I: involves the use of mortality rates by SES generated from two different cross-sectional datasets.
Method II: involves the use of linked record studies other than the census such as surveys.
2. Mortality follow up of survey data
Evaluation of surveys as data sources to estimate HLY by SES Surveys considered: HIS, LFS, SHARE, SILC
Criteria of evaluation Indicators availability: Health and SES Mortality follow up Representativity of the sample Survey design aspects (sample size, response rate, periodicity)
Final Choice: HIS (2001) and SILC (2004)
Follow up mortality of these surveys
Data
Mortality follow up : 6 years for HIS, 5 years for SILC HIS: 12 111 individuals initially, 10 093 matched SILC: 10 146 individuals initially, 97 75 matched
SES variable HIS: Highest educational level in the household SILC: Highest individual educational level
Health outcome Global Activity Limitation Indicator (GALI) • “For at least the last 6 months, have you been limited because of
a health problem in activities people usually do?”
Estimation of HLY
Sullivan method: based on cross sectional data
Method of choice for estimating HE due to its simplicity, relative accuracy and ease of interpretation
HLY will be estimated with their standard errors
SES Inequalities is studied in both surveys by comparing the lowest versus the highest educational category using the z-statistics
Mortality by SES among women in HIS
2001 and SILC 2004
EducationHIS SILC
N % N %
Higher education 24 1.8 20 1.4
Higher secondary 41 2.6 39 2.6
Lower secondary 51 5.9 30 4.0
Primary education 91 17.7 123 12.1
Total 207 5.3 212 4.8
Prevalence of disability by SES among
women in HIS 2001 and SILC 2004
EducationHIS%
SILC%
Higher education 14.1 19.0
Higher secondary 20.3 24.9
Lower secondary 27.6 34.7
Primary education 42.8 49.7
Total 23.3 30.7
HLY and years of disability among women aged 25 years
HIS 2001
0
10
20
30
40
50
60
70
Primary education Lower secondary Higher secondary Higher education
HLY Years in disabil ity
SILC 2004
010
20304050
6070
Primary education Lower secondary Higher secondary Higher education
HLY Years in disabil ity
HLY and years of disability among women aged 65 years
SILC 2004
0
5
10
15
20
25
30
Primary education Lower secondary Higher secondary Higher education
HLY Years in disabil ity
HIS 2001
0
5
10
15
20
25
30
Primary education Lower secondary Higher secondary Higher education
HLY Years in disabil ity
Inequalities in HLY among women by
age and survey
Survey Primary
educationTertiary
education Difference
in HLYp
HIS-25 years 34.25 50.48 16.24 <0.01
HIS-65 years 8.62 16.95 8.33 <0.01
SILC-25 years 30.37 42.73 12.36 <0.01
SILC-65 years 7.56 12.57 5.01 <0.20
Comparison HIS and SILC
Comparable mortality rates by educational level in HIS and SILC
Disability rates are systematically higher in the SILC compared to the HIS
Systematically HLY are higher in the HIS
Significant inequalities are found in both surveys
At older ages, educational inequalities are significant in the HIS but not in the SILC
3. Recommendations
Conclusion
The golden standard for estimating mortality by SES (and subsequently HLY) is through census linkage with National Register
In the absence of the census: Linked approach using surveys is a possible alternative Selection bias Comparison with the 3 years mortality follow up of census 2001
Life expectancy among women aged 25 years, census, HIS, SILC
EducationCensus2001
HIS 2001
SILC 2004
Higher education 59.90 63.37 61.85
Higher secondary 58.52 59.67 59.10
Lower secondary 58.00 58.83 61.47
Primary education 56.17 58.5 58.56
No diploma 53.98 - -
Recommendations for the use of survey follow up (1) A choice has to be made concerning the survey to use
As we got comparable variances using the HIS and SILC, the choice should not be based on statistical stability but on other criteria (e.g. regional estimates, yearly estimates)
Estimates are not interchangeable between HIS and SILC
To monitor HLY by SES in Belgium we recommend the use of the SILC as it is yearly and it is used at the European level
Recommendations for the use of survey follow up (2) Use of educational level as the SES stratification variable
Use of the Global Activity Limitation Indicator (GALI) to estimate the disability free life expectancy
Use of the Sullivan method Calculate the variances with the estimates
More practical information for the estimation (statistical programs, request for the privacy commission) are found in the final report
Thank you for your [email protected]