Health and mortality abstracts

Strand organisers: Dr. Paula Griffiths, Loughborough University & Professor Monica Magadi, University of Hull 

Health: Health inequalities in Africa: Monday 7 September - 1.30pm 

Maternal education and infant mortality decline in Egypt
Jona Schellekens, Hebrew University of Jerusalem 

There are three major explanations for the decline of infant mortality in less developed countries: improved living standards; advances in medical and public health; and better maternal education. A recent summary of the evidence suggests that maternal education is more important for infant survival than are living standards. This summary, however, is mostly based on cross-sectional studies. One common assumption is that the causal links detected at the cross-sectional level will also hold from a time-trend perspective. A hypothesis about mortality decline, however, cannot be tested by looking at cross-sectional correlations at a single point in time. The major aim of this study is to estimate the contribution of better maternal education to infant mortality decline in a longitudinal study, using data from Egypt. The study pools the infant mortality history of 20,739 women from six surveys: the World Fertility Survey and five phases of the Demographic and Health Survey. Between 1970 and 2006, 55,636 infants were born to these women. This study estimates that between 1970 and 2007 infant mortality in Egypt declined by about 90 percent. Even though young mothers received much better education, the results of a mixed logistic regression model of the probability of an infant death show that better maternal education only provides a marginal explanation for the decline in infant mortality. Instead the decline is strongly correlated with improved living standards, as measured by GDP per capita. Thus, our model leaves little room for explanations of the decline that are not correlated with the rise in GDP per capita, such as advances in medical and public health. 


Too poor to access free care: Have maternity care fee exemptions policies in Ghana benefited the poor?
Fiifi Amoako Johnson1, Faustina Frempong-Ainguah2, Sabu S. Padmadas1, 1Division of Social Statistics and Demography and Centre for Global Health, Population, Poverty, and Policy (GHP3), University of Southampton 2Regional Institute for Population Studies, University of Ghana   

Low and middle income countries have invested significantly in fee exemption policies to reduce financial barriers that limit the poor and marginalised from accessing skilled maternity care and also accelerate progress in reducing maternal and neonatal deaths. Ghana has implemented four major policies in the last two decades: full-cost recovery ‘cash and carry’ scheme, antenatal care fee exemption, delivery care fee exemption, and National Health Insurance Scheme. This study aims to investigate the impact of maternity-related fee payment policies on the uptake of skilled birth care amongst the poor in Ghana, focusing on the temporal trends referring to the periods that the policies were functional. Population data representing 12,288 births between November 1990 and October 2008 from 4 consecutive rounds of the Ghana Demographic and Health Surveys were used to examine the impact of the policies. Multilevel logistic regression was used to examine the effect of the policies on the uptake of skilled birth care, adjusting for relevant predictors and clustering within communities and districts. Uptake of skilled birth care over the policy periods for the poorest women was trivial when compared with their non-poor counterparts. The rich–poor gap in skilled birth care use was highly pronounced during the ‘cash and carry’ and ‘free antenatal care’ policies period. The benefits during the ‘free delivery care’ and NHIS policy periods accrued more for the rich than the poor. The maternal care fee exemption policies specifically targeted towards the poorest women had limited impact on their uptake of skilled birth care. 


Are fathers’ gender attitudes associated with child health outcomes? Evidence on child undernutrition in sub-Saharan Africa using the Demographic and Health Surveys
Jessica Heckert1, Jasmine Fledderjohann2, 1International Food Policy Research Institute 2University of Oxford 

Extant research on child well-being in sub-Saharan Africa, Africa South of the Sahara (SSA), has focused on women’s role in shaping child outcomes. However, men’s roles as fathers and husbands are also likely to influence children’s well-being. In this paper, we explore how men’s gender role attitudes influence children’s nutritional status in 23 countries in SSA, focusing on the association between attitudes towards wife beating and stunting and underweight among children. Applying logistic regression to the Demographic and Health Surveys for these countries, we find that, controlling for socio-demographic confounders, men’s egalitarian gender attitudes are associated with lower odds of stunting in Cote d’Ivoire, Mozambique, Senegal, Swaziland, and Uganda, and with lower odds of underweight in Cote d’Ivoire, Guinea, Mozambique, Rwanda, and Uganda. In models which include maternal gender attitudes as well, we find that, in Namibia and Malawi, egalitarian gender attitudes among both fathers and mothers together is associated with better child health outcomes. 


Rural and urban differentials in adult mortality in sub-Saharan Africa
Ashira Menashe-Oren, Guy Stecklov, Hebrew University of Jerusalem 

Research from sub-Saharan Africa consistently shows that urban mortality levels are substantially lower than those found in rural areas. Rural populations are disadvantaged in household characteristics, educational and economic opportunities and have little access to health services. This translates into higher rural mortality. Yet, what is less recognised is that this research is heavily based on infant and child mortality measures or occasionally on maternal mortality. In fact, despite recognition of the value of lowering mortality in sub-Saharan Africa, there is shockingly little empirical evidence on how mortality levels for adults compare across urban and rural sectors. Furthermore, this question becomes particularly salient as AIDS mortality shifted the burden of disease so dramatically over the past three decades. This research aims to examine whether and to what extent a difference exists in adult mortality between rural and urban populations. The lack of population registration data and mortality records make it exceedingly difficult to estimate mortality rates. However, using an indirect measure of adult mortality, the orphanhood method, it is possible to derive rural and urban life tables separately. Drawing upon data from Demographic and Health Surveys for 40 sub-Saharan African countries between 1990 to 2013, we estimate separate urban and rural adult mortality levels. Preliminary results from five countries indicate that the difference in the probability of dying between ages 15 to 60 amongst men is 23.7%, and amongst women is 19.6%. In general male mortality is higher in urban areas, but amongst women there is no clear difference in survivorship between rural and urban populations. 


Health: General health and wellbeing: Tuesday 8 September – 9.00am 

Inequalities in self-reported health and self-reported illness among women aged 15-54 in the oPt
Katie Bates, Tiziana Leone, Ernestina Coast, Roula Gandour, Rita Giacaman, Sawsan Imseeh, Suzan Mitwalli, Shiraz Nasr 

Background: Whilst much literature focuses on the ability of indicators of self-reported health to predict mortality, the linkage between self-rated health and self-reported illness remains under-studied, especially in low and middle income countries. We investigate the consistency between self-reported health (SRH) and self-reported illness (SRI) by socio-demographic characteristics among women in the occupied Palestinian territory (oPt). Methods: Using logistic regression we analysed data from the PHS, for women aged 15-54, to investigate socioeconomic and demographic inequalities in reporting poor SRH and chronic or acute illness. Findings: Socioeconomic inequalities exist in SRH and SRI among women in Palestine. These inequalities are consistent across both SRH and SRI. Women with completed secondary or tertiary education are less likely to report poor SRH, acute illness or chronic disease than those with lower levels of education. Women from the lowest wealth quintile are more likely to report poor SRH, acute and chronic illness compared to richer women. Regional disparities exist, but are not consistent across SRH and SRI. Women from Gaza are less likely to report poor SRH compared to women from all other regions (North, Central and South West Bank). For SRI, Gazans report less illness than women in North West Bank. Women report better SRH in rural than urban areas. Women in urban areas and refugee camps are more likely to report a chronic illness than in rural areas. The odds of a woman reporting acute illness are higher in refugee camps than rural areas. Interpretation: Clear within- and between- disparities exist in SRH and SRI in Palestine. These disparities are both socioeconomic and cultural; they need to be accounted for when analyzing Palestinian data. Regional disparities are not negligible and need to be accounted for when considering health service interventions in the country. 


Health inequalities in British post-war cohorts: Are later born cohorts less healthy than earlier generations?
Stephen Jivraj1, Alissa Goodman2, George Ploubidis2; 1UCL Department of Epidemiology and Public Health, 2UCL Institute of Education   

Despite improvements in life expectancy in many countries, there is uncertainty whether increase in years of healthy life expectancy has kept pace. There is evidence from a number of high-income countries, including the UK, that outcomes were worse for later born generations at the same age. This paper explores this hypothesis using a range of subjective, self-reported and biomarker health outcomes in England. Repeated cross-sections are taken from the annual Health Survey for England. We derived comparable health outcomes from 1993 to 2013. Synthetic cohorts of people born between 1945 and 1980 are compared at the same age using regression models. The models are fitted separately for men and women and control for age, education, smoking and drinking. We find that later born post-war cohorts have poorer health at the same age. For example, have poorer self-assessed health, report more diabetes, high blood pressure and mental disorders, and have greater risk of heart disease and diabetes. Moreover, the gaps within these cohorts are widening between social groups for subjective health status. This trend points to a greater demand for public healthcare at younger ages from those groups who are least able to help themselves. The differences between cohorts on many health outcomes have narrowed with increasing age, which suggests that these trends are reversible. Later born post-war cohorts are less likely to report a long-term illness, and have clinically measured high blood pressure than those born earlier in the period 1945 to 1980. 


What is the effect of social mobility between 1991 and 2001 Censuses on the
socioeconomic gradient in mortality in 2007-11?
Chris White, Office for National Statistics

Background and research question
Health selection has been postulated as a plausible mechanism for observation of a social gradient in mortality. This mechanism is founded on the idea that a person’s relative health attributes are associated with their direction of social mobility placing people with more favourable health attributes in higher socioeconomic positions and vice versa. Such trajectories create a social gradient in mortality because more healthy people cluster in more advantaged positions and less healthy people in disadvantaged positions. Consequently, such an explanation militates against establishing a cause effect relationship between relative occupational health hazards and the material circumstances associated with them. If health selection causes social mobility, it might be expected that mortality risk would be similar among those people entering a class of destination from a less advantaged class of origin as those people in that same class at origin and destination. Similarly, those people entering a class of destination from a more advantaged origin would have similar risk to those people in that same destination class as at origin.

The ONS Longitudinal (ONS LS) has been used estimate mortality risk by socioeconomic position at origin at the 1991 Census and at destination at 2001 Census using linked census and death records. Socioeconomic position was based on National Statistics Socioeconomic Classification (NS-SEC), derived from occupation and employment status. The NS-SEC is a measure of advantaged construed from employment relations, and aims to reflect the socio-economic structure of 21st century societies. A study population consisting of men and women aged 20-65 in 1991 Census and classifiable to NS-SEC at 1991 Census and 2001 Census was selected for analysis in the period 2007-11. LSMs surviving to the start of this outcome period were aged 36-80. Social mobility was defined through a simple ordering of classes to determine relative advantage in employment conditions. This is contentious in the treatment of transitions between the Intermediate and Small Employers and Own Account Workers classes. However, the lower mortality observed in the former class compared with the latter class during the period 2007-11 following assignment at baseline, supported this approach.

In order to determine the relationship of class membership and direction of social mobility to mortality in the period 2007-11, separate logistic regression models were used to investigate:
1.        presence of a social gradient in mortality during 2007-11 for class of origin, unadjusted
2.        effect on origin class gradient of adjusting for social mobility
3.        presence of a social gradient in mortality during 2007-11 for class of destination,     unadjusted
4.        effect on destination class gradient of adjusting for social mobility
5.        risk patterns associated with the scale of social transitions using expanded class trajectories


The unadjusted odds of death during 2007-2011 of the routine class was twice that of the Higher Managers and Professionals class, the reference class, and all classes other than the Lower Managers and Professional class had statistically significant raised odds of death compared with the reference class. Adjustment for social mobility increased this gradient by almost 50% rising to 2.4 times the odds and all other classes has higher odds than observed in the unadjusted model. Those moving into less advantaged socioeconomic positions in 2001 had on average raised odds of 33% compared with the class they left, while upwardly mobile people had 90% of the odds of the class they left. These estimates suggest those move in an out of the origin class if unaccounted for act to constrain the gradient rather than exaggerate or create it. The destination class unadjusted model contains those remaining socially static and those entering that class from a different class at the 1991 Census. The destination class social gradient increases to 2.2 suggesting class inequalities are greater if assigning class at the 2001 Census compared with at the 1991 Census. This is partly explained by the greater associated risk of downwardly mobile people compared with upwardly mobile people, offset somewhat by the greater rate of upward transitions. Estimating mortality for destination class places all classes at statistically significant raised odds of death compared with the reference class. The final gradient, adjusting for general social mobility is almost 2.5, suggesting the Routine class has 2.5 times the odds of death of the Higher Managers and Professionals class. Those LSMs who were upwardly mobile into a destination class had raised odds of death compared with the socially static members of the class they joined, suggesting health attributes were not comparable with the class they joined and are consistent with the gradient constraint postulation observed in the origin class models. On average upwardly mobile people had a raised odds of death of almost a quarter. Downwardly mobile people into class of destination failed to detect a statistically significant effect suggesting such people had comparable odds of death to the static members of the class they joined. Models exploring transitions between specific classes failed to detect a discernible hazardous or protective pattern in mortality risk by scale of transition although estimates were imprecise and only a small number achieved statistical significance. 

Potential applications
These models can be used to extrapolate mobility effects to existing gradients and thereby give a more valid perspective on health gaps and their trajectory. This work provides evidence in support of socioeconomic inequalities in health arising from relative exposure to occupational hazards and the material circumstances associated with various socioeconomic positions and against selection effects being the principle agent explaining health differentials between advantaged and disadvantaged populations 


If all we knew about women outside the childbearing period health needs in low income countries was coming from surveys what would we know?
Tiziana Leone1, Katie Bates1, Jacques Emina2; 1London School of Economics, 2InDepth Network 

Evidence on access to health care for women outside the childbearing period (e.g.: childless women, sterilized or menopausal) in Low Income Countries (LICs) is limited – both in respect to their general health. To date there is little evidence on how health systems meet their needs due to the lack of interest and above all lack of data. However, it is important to shed more light into this field as women not only continue to engage in sexual activity but have other health care needs. A failure to address these needs is potentially detrimental to health - not only in the short-term, but also in the long-term particularly given the global context of emerging non-communicable diseases. Moreover, for women in LICs in particular, cumulative effects of multiple births or birth injuries can continue to cause health problems for women across their life course. Using Ghana as a study case the aim of this study is to review both the Demographic and health Survey (2008) and the Study on Global Ageing and Adult Health (2012) to understand what we know about women's health needs and how are they met outside the childbearing experience. Preliminary results show a growing number of older women, with mobility issues, caring responsibilities and poverty that hinder access to health care services. Women usually report worse health than men in particular for low economic groups. The review of the data shows there is still a gap in access to health for women above age 40. This research is timely in light of declining fertility and an increasing number of childless women which would need to gather further policy attention. 


Health: Migration, ethnicity and health – Tuesday 8 September – 1.45pm 

How does risk of cardiovascular disease (CVD) vary for movers and stayers in New Zealand by ethnic group? A survival analysis
Fran Darlington1, Daniel Exeter2, Paul Norman1; 1University of Leeds, 2University of Auckland   

Residential mobility is an important determinant of CVD in Auckland. However, it is not known whether this extends to the rest of New Zealand or if the impact moving has on the risk of CVD varies by ethnicity. Using a cohort of approximately 2 million adults aged 30-85 years residing in New Zealand, we explore how the risk of CVD varies between ethnic groups according to mover status (movers and stayers) and deprivation change (moving up, down or within deprivation quintiles) using survival analysis. In particular, we investigate whether movers of minority ethnic groups have a higher risk of CVD relative to stayers than observed for the majority population. Further, we investigate whether risk of CVD for those either churning in the same deprivation quintile or moving up or down quintiles varies by ethnicity. Our results show that whilst all movers have higher hazards of CVD relative to stayers, there are variations within ethnic groups. For example, Indians in New Zealand have some of the lowest rates of mobility but the gap in the hazard of CVD between Indian movers and stayers is markedly wider than for other ethnic groups. Further, despite the contrasting socioeconomic experiences of New Zealand European groups compared to Māori peoples, the associations between deprivation change and residential mobility are very similar. These results illustrate the need to further examine the divergent experiences between ethnic groups in the risk of CVD for movers and stayers and the importance of deprivation change in explaining these differences. 


Infant and fetal mortality levels and patterns among Greek natives and foreign origin populations in Greece: 2010-2012
Cleon Tsimbos1 and Georgia Verropoulou1, 2, 1 University of Piraeus, 2 Institute of Education, University of London 

The aim of the study is to examine infant and late fetal mortality of Greek natives, all immigrants and selected ethnic groups in Greece and to make comparisons with the respective countries of origin. Data. The study uses vital statistics on livebirths, infant deaths and stillbirths by citizenship registered during 2010-2012; stillbirths are classified by age of mother, too. The data are unpublished and have been provided by the Hellenic Statistical Authority upon special request. National data for the countries of origin have been obtained from the Eurostat database and the World Health Statistics. Methods: Conventional infant and fetal mortality rates and respective standard errors are computed. Stillbirth rates for natives, all immigrants and selected ethnic groups (Albanians and Bulgarians) are also calculated by age of mother and, whenever possible, the rates observed in Greece are compared with those in the respective countries of origin. Results: Infant mortality among immigrants (3.7‰) is higher than among natives (3.3‰) but immigrants from Albania exhibit the most favourable rate (3.0‰). Unfavorable rates were estimated for immigrants from Bulgaria, Romania, Georgia, Syria and Pakistan. Nevertheless, all migrant groups in Greece experience significantly lower infant mortality rates than in their countries of origin. Similar findings are obtained concerning fetal mortality. Stillbirth rates by age of mother are higher among immigrants compared to natives across all age groups. Finally, young Bulgarian mothers aged 15-29 experience relatively higher fetal mortality rates than their counterparts in Bulgaria but the reverse is observed for ages 30 and over. 


Ethnic density and mental health: the role of sense of belonging
Natasha Crawford, Institute for Social and Economic Research, University of Essex 

Ethnic differences in mental health outcomes have been documented in the UK, with some groups reporting poorer health outcomes than those who identify themselves as White British. Increasingly studies suggest that residing in an ethnic enclave, that is an area of high co-ethnics, is associated with a protective or buffering effect upon health outcomes and can help to explain why we see differences across ethnic groups. Utilising data from Understanding Society, the UK’s largest panel study, and 2011 Census data, this paper tests the ethnic density hypothesis and whether this association is moderated by ones attachment to the neighbourhood, operationalised via a battery of questions. Multilevel models were used to estimate the association between ethnic density at the Lower Super Output Area (LSOA) and mental health among the five largest ethnic minority groups in the UK. In accordance with other studies, results were statistically significant for some groups only and ones attachment to the neighbourhood exhibited a small, but significant effect, for some ethnic groups. This study contributes to the literature in two ways: (1) by utilising a more contemporaneous measure of ethnic density, specifically data from the 2011 Census, than previous studies and (2) by testing whether ones attachment to the neighbourhood moderates the association between ethnic density and mental health. Potential policy implications of promoting attachment to the local community are considered. 


The mortality of descendants of migrants in England and Wales: does a ‘Healthy Migrant Effect’ persist beyond the first generation?
Matthew Wallace; Dept. of Geography and Planning, University of Liverpool 

Migrants often have low mortality compared to host populations in Western countries. This healthy migrant effect (HME) is thought to wear off across generations. The aim of this paper is to discern whether low mortality found in migrants in England and Wales extends to descendants or whether mortality of descendants attenuates to host population levels. This paper examines mortality of migrants and descendants using a large, longitudinal sample (the Office for National Statistics Longitudinal Study). Event history analysis is used to study the mortality of 490,000 individuals. Analysis finds low mortality in all migrants but variation in the mortality of descendants from 1991-2012. Descendants of Black Caribbean migrants have high mortality which cannot be accounted for by social background; descendants of Pakistanis and Bangladeshis have high mortality which can be accounted for by social background. Descendants of Indians, Black Africans and Other have low mortality which is initially masked by social background. As to why, the loss of selection from migrants to descendants is considered alongside generational differences in cultural norms, attitudes and behaviours in lifestyle and diet. The early life development of migrants and descendants in different countries is also considered in terms of disease profiles and social position. In sum, while a HME is present in migrants irrespective of ethnic group; there is marked variation in descendants’ mortality. A healthy migrant effect persists in some ethnic groups after the first-generation but not others. Social position has a much more influential effect on the mortality of descendants. 


Health: Socio-demographic predictors of health – Wednesday 9 September – 9.00am 

Explaining the growing education gap in U.S. adult life expectancy, 1990-2010
Isaac Sasson, LSE 

As life expectancy at birth in the U.S. approaches eighty years of age, educational differentials in adult mortality are greater than ever. Low-educated Americans have shorter life expectancies than their college-educated counterparts and have recently suffered absolute declines in longevity. Using vital statistics data, this study decomposes those trends by age and cause-of-death for major educational attainment groups in the U.S. from 1990 to 2010. The findings reveal an education gradient in life years lost from all major causes of death. Among low-educated whites, life expectancy declined predominantly due to a rise in external and smoking-related deaths. Mortality also increased among high-school educated whites under age 55, offsetting mortality reductions in old age. Evidently, large segments of the U.S. population are diverging from the classic health transition model and instead are undergoing a series of divergence and convergence sequences resulting from changes in social conditions, health technologies, and emerging mortality risks. 


Partnership status and mortality in England and Wales: All-cause and cause-specific mortality between 2001-2011
Sebastian Franke, Hill Kulu; University of Liverpool 

Married people have lower mortality rates than single, separated, divorced and widowed individuals. Research suggests that these differences, more pronounced for men than for women, persist even when controlling for economic, social, and demographic characteristics of individuals. One aspect that remains understudied in the literature is a re-evaluation of these trends by different living-arrangements. Using data from the ONS Longitudinal Study (LS) and applying hazard regression models, the project investigates mortality by partnership and family status in England and Wales between 2001 and 2011. By analysing all-cause mortality, we show significant mortality differences by marital status with married people exhibiting lower mortality levels than non-married individuals; cohabiting (living with a partner) explains only some differences. We further analyse cause-specific mortality by marital status. Controlling for socio-economic factors and living arrangements, we investigate causes of death by marital status to gain a better understanding of the role of health selection and health protection. (Cause-specific analysis is in progress and will be complete by end of June.) We present findings for men and women in different age-groups between 30 and 85 years. 


Mental wellbeing and support exchanges between older adults and their children in Eastern and Western Europe: what are the impacts of number and gender of children?
Katherine Keenan & Emily Grundy; London School of Economics 

Studies have shown that older people have frequent support exchange with their adult children, but this varies between Northern and Southern Europe. Much less is known about Eastern Europe. Research has also shown that fertility histories, including timing and number of births, are related to aspects of mental well-being in later life. Again much less is known about Eastern Europe where family building patterns have differed from those in other regions. We use data from the Generations and Gender Surveys to investigate the importance of number and gender of children for exchanges between older adults and their children and subsequent associations with older adults’ depression and loneliness. We used longitudinal data on older adults aged 50-79 years from Bulgaria, Georgia and France. The results show that having more children in Bulgaria and Georgia was associated with greater financial support and lower levels of loneliness among older adults. In France the number of children was less important and downward support flows were more common. Gender of children was not associated with support exchange or mental wellbeing. Childlessness and having co-resident adult children moving out was associated with greater risk of depression at follow-up in Bulgaria and Georgia, but not in France. In all countries losing a partner/spouse was associated with increased depression and loneliness. The results suggest differences in the expectations and extent of parent-child support between Eastern and Western European countries, and that the extent of child support received by older adults have a stronger negative effect of parent’s mental wellbeing in Eastern countries. 


The recession effect on Greek age, gender and cause-specific mortality rates
Alexandra Tragaki, Department of Geography, Harokopion University of Athens

The aim of this paper is to examine the immediate effects of the current economic crisis on crude death rates (CDR) and on age and cause-specific mortality rates. Based on macroeconomic data, analysis estimates age and sex specific death rates for the years 2000-2012, mainly focusing on the trends during the recessionary years compared to the pre-recession period. Analysis relies on mortality statistics by age, sex and cause of death annually provided by the Hellenic Statistics Authority (EL.STAT). Causes of death are coded according to the rules specified by the International Classification of Diseases System. Joinpoint regressions investigate changes in time trends and examine their significance. Some of the first findings can be summarized as follows: • All causes age-standardized mortality rates continue their uninterrupted downward trend, despite the crisis outburst in 2008.This is the case for both sexes but not for all age-groups. • This trend is mainly driven by the steady decline registered in major death causes: all cancers, circulatory and respiratory diseases. • Certain causes of death ceased or reversed their declining movement during recessionary years. More specifically, suicides, homicides and violent deaths as well as deaths caused by mental or behavioral disorder have been resurging since 2008. • Men seem more vulnerable to recession related causes of death than women • Age-specific death rate analysis shows that compared to elders young ages seem more vulnerable to the economic downturn. • Infant mortality has been slightly ascending since 2008. 


How much of the variation in mortality across Norwegian municipalities is explained by the socio-demographic characteristics of the population?
Øystein Kravdal, Kari Alvær, Kåre Bævre, Jonas M. Kinge, Jørgen R. Meisfjord, Ólöf A. Steingrímsdóttir, Bjørn Heine Strand; Norwegian Institute of Public Health 

The goal was to find out whether much of the variation in mortality between the 430 Norwegian municipalities could be attributed to socio-demographic characteristics of the population - operating through individual- or aggregate-level mechanisms. Two-level discrete-time hazard models were estimated for women and men at age 60-89 in 2000-2008, using registers covering the entire population. Year, age and a municipality-level random term were included in the first step. When socio-demographic characteristics of the individual and others in the municipality were added, the variance of the random term was reduced by 73%-80% almost exclusively because of aggregate-level effects. Policy implications of these findings are discussed. 


Health: Early-life association with later health and wellbeing – Wednesday 9 September – 11.30am 

Early-life conditions and dementia in late life: Evidence from a quasi-natural experiment in Germany
Sebastian Klüsener1, Gabriele Doblhammer1, 2, 1Max Planck Institute for Demographic Research, 2University of Rostock 

Population ageing and growing demand for elderly care pose challenges to modern welfare states. For those countries that experienced a baby boom after WWII, a key question is to what degree these relatively big cohorts will age differently in comparison to today’s elderly cohorts who were born in the turbulent times before 1945. In terms of dementia prevalence, there is growing evidence that early life conditions affect brain development with potential long-term implications for dementia risks. But it is difficult to study this link due to a dearth of data on elderly individuals which includes next to their current health status also information on early life conditions. In our study we address this challenge by making use of a quasi-natural experiment in Germany. The country was historically characterized by stark regional variation in early life conditions (i.e., infant mortality levels), which was strongly linked to regional breastfeeding practices. However, in the 1920s and 1930s, breastfeeding prevalence swiftly increased to almost 100% in all German regions and infant mortality substantially declined. We use spatial analysis methods to link regional-level dementia prevalence, obtained from routine claims data of the biggest German health insurance, with regional data on infant mortality at birth. For the cohorts born before 1928, we can detect in contrast to the younger cohorts a much stronger statistical association between infant mortality at birth and dementia prevalence. This supports the view that exploring the link between early life conditions and dementia risks can be fruitful to assess future care demand. 


The effects of childhood circumstances on adult mental health
Lewis Anderson, Department of Sociology, University of Oxford. 

Background: The literatures examining the effects on adult mental health of low childhood SES (CSES), and of childhood adversities (CAs) have generally not been considered together. This paper represents an attempt to understand whether low CSES increases risk of mental illness in adulthood through these relatively extreme stressors, or whether the mechanisms are more mundane, such as the generalised transmission of disadvantage. Methods: Data from the NCDS are used in a series of logistic regressions predicting high psychological distress at age 33. CSES (at birth) and CAs are each included separately in models 1 and 2. Model 3 includes both, model 4 then adds adult SES. Analyses are conducted separately for men and women. Results: For men, the effect of low CSES is accounted for by the inclusion of CAs. Being taken into care, parental mental illness and being bullied at school are all significant predictors, with effects that persist even with the inclusion of adult SES. For women, large effects of CSES are only moderately attenuated when CAs and even adult SES are included. The CAs showing significant effects for women are family criminality, domestic tension, and being bullied at school. Conclusions: The relative importance of different types of childhood disadvantage for adult mental health appears to vary considerably between men and women. Being bullied at school emerges as particularly important for both, but has received relatively little attention in this field. The pathways from low CSES to poor adult mental health in women require further study. 


Changes in the association between birth weight and cognitive development in childhood: a cross-cohort comparison in the UK
Alice Goisis, Mikko Myrskylä, Berkay Ozcan; London School of Economics 

Several studies document a negative association between low birth weight and a range of outcomes measured in childhood and adulthood, including poor cognitive development. It is not known whether this association is stable or has changed over cohorts. Given the advances in modern neonatology and prenatal care, it is reasonable to expect that poor neonatal health in the 1950s would be more likely to carry potentially negative consequences for later life outcomes than it does at present. In this study, we aim to address this gap in knowledge by investigating the secular association between birth weight and cognitive performance in childhood. We use data from three UK birth cohort studies: the 1958 NCDS, the 1970 BCS and the Millennium cohort study (birth years 2000-2001). Using OLS and logistic regression models, we compare the association between birth weight/low birth weight and markers of cognitive performance (i.e. verbal ability) at around age 11 across the three cohorts before and after adjustment for family socio-demographic characteristics. Preliminary results support the hypothesis that the consequences of poor neonatal health have changed over time. In particular, they show that a lower birth weight is more strongly and negatively associated with cognitive performance for children born in 1958 and 1970, compared to children born in 2000-2001 


Does prenatal sex bias substitute postnatal sex bias?: Decomposing the fertility and mortality effects of changes in child sex ratios across the world
Ridhi Kashyap, University of Oxford 

Has the spread of prenatal sex selection, as evidenced by sex ratio at birth distortions, been accompanied by a reduction in postnatal discrimination against girls manifested in excess female child mortality? This paper examines the relationship between prenatal and postnatal sex bias in global comparative perspective using lifetable decomposition methods applied to United Nations population data. I decompose changes in the male-to-female child sex ratios (1-5 and 5-9 years) between 1990-95 and 2005-2010 in 196 countries to changes in the fertility component, accounted by changing sex ratios at birth, and mortality component, accounted for by changing sex-differentials in infant and child mortality. The vast majority of countries saw little or no change in their child sex ratios over the period. The South Korean child sex ratio became more feminine between 1990 and 2010 falling 3.5% from 1.14 to 1.10. This improvement was almost entirely explained by improvements in sex ratios at birth, indicative of declining pre-natal sex selection. Child sex ratios became more masculine in twelve countries, concentrated in Africa, Asia and the Caucasus. In the African countries – Benin, Angola, Liberia and Rwanda – the greater masculinity in child sex ratios over the period was related to greater improvements in male survival relative to female survival, with little or no change observed in sex ratios at birth. In the Asian contexts where increases in masculinity were observed, comprising China, India, Nepal, Vietnam and Pakistan, the increases were mostly explained by a rise in pre-natal sex selection. Excess female mortality showed little or no improvements over the period across these Asian contexts. The most significant changes in child sex ratios in the world over the period were observed in the Caucasus in Azerbaijan and Armenia, with a less steep increase in Georgia. In the Caucasus, the increasing masculinity in child sex ratios was explained predominantly by increasing pre-natal sex selection, but also by worsening survival chances for females relative to males. This global analysis suggests that while prenatal sex selection was the more dominant cause of more masculine child sex ratios in the 1990s and 2000s, it did not accompany improvements in postnatal relative survival chances for females.