Retrospective life course data from European countries on how early life experiences determine health in old age and possible mid-life mediators

Abstract

The data presented in this article is related to the research paper entitled “The long arm of childhood circumstances on health in old age: Evidence from SHARELIFE” (E. Pakpahan, R. Hoffmann, H. Kröger, 2016) [1]. It presents the distribution of socioeconomic status (SES) and health from childhood until old age in thirteen European countries. In order to capture the characteristics of longitudinal data, which resembles life course data, we divide the data into three schematic periods: childhood (up to 15 years old), adulthood (30 to 60 years old), and old age (61 to 90 years old). This data set contains respondents’ life histories, ranging from childhood conditions (such as housing and health) to detailed questions on education, adult SES (working history, income, and wealth) and old age health. The data can be used not only to understand on how early life experiences determine health in old age, but also to recognise the importance of possible mid-life mediators.

Specification Table

Subject area	Ageing, Sociology and Public Health
More specific subject area	Life course, longitudinal data, health inequality
Type of data	Table and graph
How data was acquired	Data was available from SHARE database through registration.
Data format	Aggregated, analyzed, filtered
Experimental factors	Our sample is based on the SHARELIFE dataset and was extracted using STATA and reorganized using the Stata newspell package[2].
Experimental features	For each person, data on health and SES was aggregated for childhood, adulthood, and old age
Data source location	Germany
Data accessibility	The data was available from the SHARE database through official registration athttp://www.share-project.org/data-access-documentation/research-data-center-data-access.html

Value of the data

• •The data shows the overall distribution of SES and health for old people in Europe, and therefore allows us not only to disentangle how SES and health are related from childhood to old age, but also to examine the trajectories or dynamics of SES and health over the life course.
• •SES information is based on various measurable variables to reflect multiple dimensions.
• •Health data is also based on various variables and offers both subjective and objective measurements.
• •The data enables an assessment of the notion of the long arm of childhood and is useful for life course analysis on health inequality [1].

1. Data

The data is based on SHARELIFE, i.e. the third wave (2008/2009) of the Survey of Health Ageing and Retirement in Europe (SHARE), which is a household panel survey [3]. SHARE is a cross-national panel database of micro data on health, socio-economic status and social and family networks of more than 45,000 individuals aged 50 or over in Europe and respondents are interviewed biennially. It is representative for the non-institutionalised population in European countries. The first wave started in 2004 and 12 countries participated, and in the sixth wave in 2015, 18 countries took part (http://www.share-project.org/home0/overview.html). SHARE is harmonised with its sister studies: the US Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA). In our sample, as inclusion criteria, the respondents are those aged 50 and above and the necessary variables needed for our research questions are available and comparable across countries, which eventually captures the citizens of 13 European countries. The data focuses on people׳s life histories, starting in childhood right through to old age. We use the data to try to understand how childhood experiences are associated with health in old age. In addition, we also take into account how mid-life conditions – such as education, income and occupation, and behaviour risk – mediate the effect of childhood SES and health on old age health.

2. Experimental design, materials and methods

2.1. Data source

SHARELIFE is the third wave of SHARE which focuses on people׳s life histories and more than 30,000 men and women took part in this round of the survey. The respondents are representative for the population aged 50+. The retrospective part of SHARELIFE offers a complete history of changes in SES and health, beginning at childhood. It complements the SHARE panel data by providing life history information to enhance the understanding of how early life experiences – and events throughout life – influence the circumstances of older people. Thirteen European countries are included: Austria, Germany, Sweden, the Netherlands, Spain, Italy, France, Denmark, Greece, Switzerland, Belgium, Czech Republic and Poland. The distributions of our sample, separated into male and female, are presented in Table 1. In Table 2 we present the distribution of all variables used in our dataset.

Table 1.

Sample size by country and by gender.

Country	Male	Female	Total
Austria	276	340	616
Germany	668	639	1307
Sweden	684	798	1482
The Netherlands	738	769	1507
Spain	678	521	1199
Italy	919	678	1597
France	714	845	1559
Denmark	647	726	1373
Greece	920	532	1452
Switzerland	415	462	877
Belgium	908	892	1800
Czech	589	740	1329
Poland	564	568	1132
Total	8720	8510	17,230

Table 2.

Descriptive statistics of the data set. Percentages are shown for categorical variables; actual numbers of cases are provided in parentheses. Means, standard deviation and ranges are shown for continuous variables.

Variables	Category	All (N=17,230)
Age	Mean	70.47
	Std. Deviation	7.62
Number of books	Up to 10 books (=1)	46.12 (7866)
	11–25 books	22.16 (3780)
	26–100 books	19.79 (3375)
	101–200 books	6.06 (1033)
	>200 books (=5)	5.88 (1003)
	Missing	173
Rooms per capita	Mean	0.71
	Std. Deviation	0.42
	Range	0: 8.75
	Missing	213
Father׳s occupation	Elementary (=1)	18.09 (2941)
	Skilled	69.42 ( 11,287)
	Associate	4.22 (686)
	Manager (=4)	8.27 (1345)
	Missing	971
Childhood SRH	Excellent (=5)	34.05 (5836)
	Very good	33.69 (5775)
	Good	23.92 (4100)
	Fair	6.15 (1055)
	Poor (=1)	2.19 (375)
	Missing	89
Ever missed school	No (=1)	88.84 ( 15,250)
	Yes (=0)	11.16 (1916)
	Missing	64
Ever in hospital	No (=0)	94.05 ( 16,172)
	Yes (=1)	5.95 (1024)
	Missing	34
Education	Mean	10.34
(years of schooling)	Std. Deviation	4.31
	Range	0: 25
	Missing	2115
Occupation	Elementary (=1)	19.40 (3280)
	Skilled	57.09 (9654)
	Associate	8.78 (1484)
	Manager (=4)	14.74 (2493)
	Missing	319
Income	Mean	16,779.89
	Std. Deviation	17,720.65
	Range	0: 586047.1
	Missing	1916
Wealth	Mean	141,249.20
	Std. Deviation	227,077.80
	Range	− 341,522.3: 6,932,346
	Missing	2094
Smoking	Current (=1)	16.46 (2715)
	Former	31.57 (5207)
	Never (=3)	51.97 (8573)
	Missing	735
Physical activity	Non-active (=1)	11.05 (1828)
	Active (=2)	88.95 ( 14,716)
	Missing	686
Old age health	Poor (=1)	14.18 (2441)
	Fair	29.29 (5042)
	Good	36.34 (6256)
	Very good	13.95 (2402)
	Excellent (=5)	6.24 (1075)
	Missing	14

2.2. Data description.

See Table 1, Table 2.

2.3. Data design and measurements

The life course data is divided into three age groups to represent the longitudinal feature of the life cycle: childhood, adulthood, and old age. In childhood we have two latent variables, childhood SES and childhood health. The childhood SES (CH_SES), as one of the key independent variables, has three indicators or observed variables: (a) number of books in the household, which represents the cultural background and parents’ education, (b) rooms per capita, which is a proxy for long-term household wealth, and (c) father׳s occupation, which we group into four categories according to ISCO (International Standard Classification of Occupation) skills levels: elementary occupations, skilled (service, shop or market sales worker, skilled agricultural or fishery worker, craft or related trades worker, and plant/machine operator or assembler), associate (technician or associate professional, clerk), and manager (legislator, senior official or manager, professional). All SES indicators refer to when the respondents were aged 10. The second latent variable is childhood health (CH_Health), which is constructed using three indicators: (a) childhood self-rated health, a five-point scale of health, from poor to excellent, (b) a binary variable indicating if the individual ever missed school for at least one month because of health, and (c) a binary variable indicating if the individual was ever hospitalised for at least one month. These health indicators refer to respondents aged up to 15 years old.

The second stage is adulthood. First, education is measured by years of schooling. The second covariate is a latent variable, adult SES, and we construct it using the indicators occupation (according to ISCO), household income and wealth – where income and wealth are corrected by purchasing power parities (PPP), relative to German Euros in 2006. Adulthood indicators refer to respondents who are between 30 and 60 years old.

The third stage contains the variables in old age, i.e. behavioural risks and old age health. We consider smoking (currently smoking, former smoker, and never having smoked) and physical activities (non-active and active) which require moderate level of energy such as gardening, cleaning the car, or taking a walk. Old age health is based on the question “Would you say your health now is…”, with the possible responses: “poor”, “fair”, “good”, “very good”, and “excellent”. The old age variables refer to respondents aged 60 and above.

3. Method

With this data it is possible to employ a structural equation model (SEM) approach, which allows us to test direct and mediating (indirect) effects via path analysis, and to combine this with measurement models for SES and health to reduce measurement error [4], [5], [6], [7], [8]. Estimating direct, indirect and total effects in our model represents the two ways in which childhood exerts its influence (the long arm) on health in old age. By direct effect we mean the extent to which childhood SES and childhood health affect old age health directly – that is, unmediated by any other variables – whereas the total effect is the sum of the direct and the indirect effects (i.e., those mediated by at least one intervening variable). Data preparation is performed using Stata 14.1, including the newspell package [2].

Transparency document. Supplementary material

Supplementary material

.