Family life course and mortality a sequence analysis Maria Josefsson - - PDF document

Family life course and mortality – a sequence analysis

Maria Josefsson1, Emma Lundholm1,2, Gunnar Malmberg,1,2, maria.josefsson@umu.se, emma.lundholm@umu.se and gunnar.malmberg@umu.se

1 Centre for Demographic and Ageing Research, Umeå University, Sweden 2 Department of Geography and Economic History, Umeå University, Sweden

Abstract The aim of the proposed study is to explore patterns of family life courses and mortality for elderly people. We identify different states of social isolation, e.g. having and not having relatives and having them close by or far away. In addition we identify typical sequences of states in order to explore how this pattern changes over time by use of sequence analysis. Since influence from family ties may differ substantially between different socio-economic groups, we explore possible associations between socio-economic characteristics and the identified types of sequences. The empirical analyses are based on micro data-set covering the whole Swedish population, and include the information from various administrative registers provided by Statistics Sweden and the National Board of Health and Welfare. It allows us to examine different paths to relative social isolation, and go beyond cross-sectional analyses of the associations between, on one hand, being lonely and, on the other, mortality and health. The analysis will provide an overview of the social landscape surrounding elderly people in Sweden and how it has emerged over time. And moreover, inform us about the association between family life course, observed as sequences of states, socio-economic characteristics and the timing of mortality.

Introduction Increasing social isolation is often depicted as one of the most serious problem for older people today (UN 2015). In times of population ageing, fewer adult children, smaller sibships and long- distance migration, the lack of an accessible social network is a major problem for many elderly people, also in non-famialistic societies. Further, a huge body of research has demonstrated a detrimental impact of social isolation on health and survival (Holt-Lunstad et al., 2010; Steptoe et al., 2013). Social isolation may affect the subjective well-being, feeling of safety, sense of coherence and influence health outcomes and mortality. In addition, family networks may also influence life style (e.g. tobacco use) medication, health monitoring etc. and hence result in increasing mortality (e.g. Vink et al., 2003). There is however, limited knowledge of how social isolation arises and evolves over time, and how this influences the lives of elderly. Mortality risks are significantly elevated for singles in comparison to married individuals and this risk has increased in more recent years, especially for women (Roelfs et al., 2011). There is also strong evidence from many studies of an elevated mortality risk associated with widowhood (Shor et al., 2012; Moon et al., 2011) and divorce (Sbarra et al., 2011). This positive association between marriage and survival can either be seen as a causal effect or as a result of selection in and out of marriage. The casual protective effect could be attributed to a combination of the instrumental and economic benefits of sharing a household, or as an effect of social and emotional support. In addition, family networks can encourage healthy behavior and monitor the health of a spouse or other family member. The negative ‘widowhood effect’ tends to be highest in the nearest future of the event, but there is also evidence of a persisting higher mortality risk for widowed compared to married individuals (Shor et al., 2012; Moon et al., 2011). In addition to the emotional stress and grief, loss of emotional and instrumental support is often identified as mechanisms behind this effect. Also selection into widowhood is an acknowledged underlying explanation, this could be a result of shared risk factors like socio-economic position or lifestyles within couples (ibid). Mortality in relation to divorce is similar to that of widowhood (Sbarra et al., 2011). Studies have shown that men are particularly vulnerable to the effects of divorce, both in the long and short term (Metsä-Simula and Martikainen, 2012). There is evidence that the inclusion of marital history in addition to current marital status contribute to a better explanation

• f mortality risk, especially for men (Blomgren et al., 2010; Dondrovich et al., 2014).

The Scandinavian countries have been depicted as ‘weak family’ societies in comparison to other European countries (Reher, 2004). From surveys like SHARE it has been shown that economic and social support between generations is more frequent in Scandinavia although less intense compared to other European countries (Albertini and Kohli, 2013). It has been argued that extended families can be viewed as a latent source of support that steps in to help in times of need, for instance unemployment, divorce or bereavement (ibid). Moreover, we know from previous research that distance between parents and adult children influences intergenerational contact (Fors and Lennartsson, 2008; Hank, 2007; Michielin and Mulder, 2007). It is therefore important to also take the geographical proximity to family members into account. Intergenerational co-residence can be considered a non-normative support strategy in a Swedish context and is generally very uncommon (Albertini and Kohli, 2013). Access to family networks varies between socioeconomic groups. With higher education comes higher migration propensity and longer distances between family members. Previous studies on geographic proximity between elderly parents and adult children suggest that high educational level is associated with longer distances to kin. These differences are however smaller in an urban context (Lundholm, 2015), despite a trend of shorter geographical distances between generations in later cohorts (Malmberg and Pettersson, 2007). Moreover, cross-national comparisons show that distances between parents and adult children are larger and intergenerational contacts less frequent in Sweden (Hank, 2007). Yet, elderly Swedes report larger satisfaction with their social networks (Olofsson and Malmberg, 2016) and less feeling of loneliness (Sundström et al., 2009; Dykstra, 2009). In addition, trends regarding how patterns of marriage, childlessness, divorce and re-partnering vary across socio-economic groups are also important to consider. Here we can also observe a shift in the behaviours related to the second demographic transition where the social gradient of e.g. fertility and family stability has shifted (Esping-Andersen and Billari, 2015). While the divorce transition was set off by individuals with higher socio-economic positions, the socio economic gradient has gradually shifted, and divorce/separation is more common among persons with lower socio-economic position (Esping- Andersen, 2016). Additionally, higher educational achievement also increases the chances of re- partnering, at least for men (de Graaf and Kalmijn, 2003). It is therefore essential to investigate

changes in the socioeconomic and demographic composition of elderly in relation to social isolation. Despite evidence of differences between demographic and socio-economic groups regarding the elevated mortality risks among widows/widowers, socio-economic factors have been inadequately controlled for in many studies and there has been a call for research that pay attention to the mediating role of individual and contextual factors (Moon, 2012; Sullivan and Fenelon, 2013). One example of a study of contextual factors in Wright et al. (2015) who found rural-urban differences, where the initial effect was stronger in rural areas, while the long term ‘widowhood effect’ was most detrimental to men in urban areas. It is suggested that men in rural areas are more socially integrated than urban men and therefore more resilient to the ‘widowhood effect’. Studies of the ‘widowhood effect’ generally compare widowed to married men and

• women. In contrast to such approach treating change in marital status as a single event we strive

in this project to investigate the impact of different patterns of family life-course trajectories on mortality by examine how marital status in combination with extended family network. The aim of the proposed study is to explore patterns of family life courses in relation to mortality for elderly people by use of sequence analysis and register data. We identify different states of having and not having relatives and having them close by or far away. Explore patterns of family life-courses in old age in relation to mortality. Methods With the main objective to explore how family life courses, including marital histories and proximity to adult children, are associated with risks of mortality, we will initially use register data to map patterns of family life courses and examine how these patterns have changed over

• time. In a second step, we will use of sequences analysis to examine how the family life course

sequences are associated with mortality risks, socio-economic position and geographical contexts. Data In the empirical analyses we use a micro data-set, based on the information from various administrative registers provided by Statistics Sweden. The database contains anonymous

individual information on every registered resident in Sweden, with annually updated and linked individual data from 1990 to 2013. It includes a large variety of demographic and socio-economic attributes, links to family members (partner and adult children) and information about marital status and extra-marital cohabitation and, moreover, geo-coded data on place of residence. To this data is also linked on individual level information about deaths by death causes, provided by the National Board of Health and Welfare (Malmberg et al 2010). Data analysis and methodological challenges Life-course analysis in demographic research refers to an approach that focuses on sequences of events in an individual’s life as the unit of study (e.g. life-course trajectories), allowing a holistic perspective of the life courses. A common objective in these studies is to identify individuals with similar patterns of trajectories. However, life-course trajectories can be very detailed, making it difficult to manually classify the sequences. Sequence analysis (Abbott and Tsay, 2000; Billari and Piccarreta, 2005; Eerola and Helske, 2012) consists in classification algorithms used to group individuals’ life courses into categories having similar sequence properties. Predictors of categories can then be studied by regressing the category indicator against covariates of interest, our case various socio-economic characteristics. The analysis procedure involves three steps: First we identify annually, for each individual, different states e.g. (a) having or not having a partner, (b) having children close by, far away or no children. In sum we identify 12 possible states every individual could have for each year over 10-20 years preceding death. In a second step we use an Optimal-matching algorithm, a classification method originally used in biology for research on DNA, to produce pairwise distances between sequences. Dissimilarity between two sequences is defined by the minimum effort (cost) that is required to transform one sequence into the other one. The output is a distance matrix that is used as input for the third step, where cluster analysis or multidimensional scaling is implemented to group individual’s sequences on the basis of pairwise similarities. Next, we compare the different groups of sequences in relation to mortality risks. And in a further step we identify associations between a number of individual variables (gender, education level, marital status, income and urban vs rural residence) and different sequences.

There are however several methodological challenges that need to be addressed when it comes to analyses where the trajectories are the unit of study. In particular, the life trajectories will need to be aligned to each other. In this project, life courses will be aligned by the event of death, and since not all individuals die at the same age, we will have trajectories of different length. To

• vercome the problem one can either impute missing states based on available information

(Liew, et al., 2011) or using variable costs for unequal sequence length (Stovel and Bolan, 2004). Also, that data sources to be used in this study are high dimensional, and a known problem for calculating distance matrices, is an issue that has to be dealt with. Despite these methodological challenges, previous research has revealed the potential of using sequences analysis for large-scale analyses of life trajectories (see e.g. Billari and Piccarreta, 2005). In previous research on family life-courses and retirement, the classification analyses have identified useful categories that were possible link to theoretical concepts (Svensson et al., 2015). Having these challenges in mind, we maintain that sequence analysis is a very useful method for a longitudinal and holistic approach to research on social isolation and patterns of individual family life-courses in relation to mortality risks. Thus, suitable for the proposed research. Results Baseline characteristics of the 96,042 individuals alive and 70 years old in 1990 are found in Table 1. Overall, 93.8% were born in Sweden. 18.7% had no children, 44.1% had at least one child living within 5km distance, and 37.2% had children living farther away. Further, 62.2% were married, 20.2% were widowed, 8.7% were divorced, and 8.9% considered never married, at

• baseline. There were significant gender differences between the marital groups (Chisq 7,467.3,

p<0.001, a greater proportion women in the widowed and divorced group (80.9% and 56.8% respectively), while there were more men in the married and unmarried group (53.5% and 57.3% respectively). Table 1. Baseline characteristics across baseline marital groups.

Married Widowed Divorced Unmarried Women 46.5% 80.9% 56.8% 42.7% Education level: (primary/ secondary/ 64.5/25.8/7.7/5.6 71.6/20.8/4.6/2.9 61.1/26.3/7.0/5.6 66.3/19.4/7.2/7.1

post secondary/ NA) Adult children: (close by/ far

away/ no children)

48.5/40.4/11.1 49.1/37.2/13.8 40.8/46.3/12.9 5.2/6.7/88.1 Median age at event of death 84 84 82 81

Discussion It has been claimed that elderly care in Sweden is in a process of re-familialization where families need to step in when the tax-paid elderly care is experiencing resource strain (Sundström et al., 2006; Szebehely and Trydegård, 2012). In a situation where elderly are becoming increasingly dependent on family care, the presence and proximity to family networks is crucial. Older persons living in one person households are more vulnerable since spouses are the most frequent caregivers (Davey et al., 2014). Vulnerability is of course even greater in absence of geographically proximate network of family members outside the household. In addition, by analysing how social isolation correlates to socio-economic factors, the patterns of combinations

• f social and economic vulnerability among elderly can be identified.

From our analyses we expect to find two major results, on one hand, an overview of the social landscape surrounding elderly people in Sweden and secondly how this family network is associated with mortality. Although the observed social landscape only include the family members, and not the wider social network, this will be a major contribution to understanding the pathways and underlying processes leading to social isolation, when focusing on family

• members. An important contribution will be to acknowledge the pathways to social isolation,

something that can help identifying the most vulnerable groups of elderly. We can identify differences by gender, socio-economic groups, contexts (urban and rural) as well as changes between cohorts. Furthermore, the analyses will inform us about the association between family life course,

• bserved as a sequences of states and the timing of mortality. Rather than identifying specific

events or states in cross-sections as possible determinants of mortality, we explore how patterns and sequences of family life courses are associated with mortality and socio-economic characteristics. References: Abbott A, Tsay A. 2000. Sequence Analysis and Optimal Matching Methods in Sociology. Sociological Methods & Research 29: 3-33. Albertini, M., Kohli, M. 2013. The Generational Contract in the Family: An Analysis of Transfer Regimes in Europe. European Sociological Review, 29(4). Billari, FC., Piccarreta, R. 2005 Analyzing Demographic Life Courses through Sequence

• Analysis. Mathematical Population Studies, 12: 81-106.

Blomgren, J., Martikainen, P., Grundy, E., & Koskinen, S. (2012). Marital history 1971–91 and mortality 1991–2004 in England & Wales and Finland. Journal of epidemiology and community health, 66(1), 30-36. Börsch-Supan A. 2013. Survey of Health Ageing and Retirement In Europe (SHARE) Wave 4. Release version:1.1.1. SHARE-ERIC. Data set. Davey, A., Malmberg, B., & Sundström, G. (2014). Aging in Sweden: Local Variation, Local

• Control. The Gerontologist, 54(4), 525-532.

Donrovich, R., Drefahl, S., & Koupil, I. (2014). Early life conditions, partnership histories, and mortality risk for Swedish men and women born 1915–1929. Social science & medicine, 108, 60- 67. Dykstra, PA. 2009. Older adult loneliness: myths and realities. Eur J Ageing. 6: 91-100. Eerola, M. Helske, S. 2012. Statistical analysis of life history calendar data. Statistical methods in medical research, DOI: 10.1177/0962280212461205. Elder Jr, GH. 1977. Family history and the life course. Journal of Family History 2: 279-304. Esping‐Andersen, G., Billari, FC.2015. Re‐theorizing Family Demographics. Population and Development Review, 41: 1-31. Esping-Andersen, G. 2016. Families in the 21st Century. Fors S, Lennartsson C. 2008. Social mobility, geographical proximity and intergenerational family contact in Sweden. Ageing & Society 28: 253-270. Hank K. 2007. Proximity and contacts between older parents and their children: A European

• comparison. Journal of Marriage and Family 69: 157-173.

Holt-Lunstad J, Smith TB, Layton JB. 2010. Social relationships and mortality risk: a meta- analytic review. PLoS Med. 7. Kolk, M. 2016. A Life Course Analysis of Geographical Distance to Siblings, Parents and Grandparents in Sweden. Population, Space and Place.

Liew AWC, Law N.F., Yan H. 2011. Missing value imputation for gene expression data: computational techniques to recover missing data from available information. Brief Bioinform 12: 498-513. Lundholm, E. (2015). Migration and Regional Differences in Access to Local Family Networks Among 60-year olds in Sweden. Journal of Population Ageing, 1-13. Malmberg, G., Nilsson, L.-G., Weinehall, L. 2010. Longitudinal data for interdisciplinary ageing

• research. Design of the Linnaeus Database. Scandinavian Journal of Public Health 38: 761–767.

Michielin, F., Mulder, C. 2007. Geographical distances between adult children and their parents in the Netherlands. Demographic Research 17: 655-678. Moon, J. R., Kondo, N., Glymour, M. M., & Subramanian, S. V. (2011). Widowhood and Mortality: A Meta-Analysis. PLOS ONE, 6(8), e23465. Reher, D. S. 2004. Family ties in western Europe. In Strong family and low fertility: A paradox? (pp. 45-76). Springer Netherlands. Roelfs, D. J., Shor, E., Kalish, R., & Yogev, T. (2011). The rising relative risk of mortality for singles: meta-analysis and meta-regression. American journal of epidemiology, 174(4), 379-389. Sbarra, D. A., Law, R. W., & Portley, R. M. (2011). Divorce and death: A meta-analysis and research agenda for clinical, social, and health psychology. Perspectives on Psychological Science, 6(5), 454-474. Shor, E., Roelfs, D. J., Curreli, M., Clemow, L., Burg, M. M., & Schwartz, J. E. (2012). Widowhood and Mortality: A Meta-Analysis and Meta-Regression. Demography, 49(2), 575- 606. Silverstein, M., Giarrusso, R. 2010. Aging and Family Life: A Decade Review. Journal of Marriage and Family 72: 1039-1058. Steptoe, A., Shankar, A., Demakakos, P., Wardle, J. 2013. Social isolation, loneliness, and all- cause mortality in older men and women. Proc Natl Acad Sci U S A. 110: 5797-801. Stovel, K., Bolan, M. 2004. Residential Trajectories: The Use of Sequence Analysis in the Study

• f Residential Mobility, Sociological Methods & Research, 32: 559–598.

Sundström, G., Malmberg, B., Johansson, L. 2006. Balancing family and state care: neither, either or both? The case of Sweden. Ageing and Society, 26: 767-782. Svensson, I., Lundholm, E., Malmberg, G., de Luna, X. 2015. Family Life Course and the Timing

• f Women’s Retirement - a sequence analysis approach. Population, Space and Place.(online)

Szebehely, M., Trydegård, G.-B. 2012. Home care for older people in Sweden: a universal model in transition. Health & Social Care in the Community, 20: 300-309. UN (2015) United Nation Department of Economic and Social Affairs PD. Living arrangement

• f older persons around the world. New York: UNDESA / Population Division, 2005

Van Ham, M., Hedman, L., Manley, D., Coulter, R., Östh, J. 2014. Intergenerational transmission

• f neighbourhood poverty: an analysis of neighbourhood histories of individuals. Transactions of

the Institute of British Geographers 39: 402-417.

Vink, JM., Willemsen, G., Engels, RC., et al. 2003. Smoking status of parents, siblings and friends: predictors of regular smoking? Findings from a longitudinal twin-family study. Twin

• research. 6: 209-217.

Wright, D. M., Rosato, M., O’Reilly, D. 2015. Urban/rural variation in the influence of widowhood on mortality risk: A cohort study of almost 300,000 couples. Health & place, 34: 67- 73.