Health & mortality abstracts

Strand organisers: Amos Channon, University of Southampton; Paula Griffiths, Loughborough University

Mortality and family dynamics - Monday 8 September 1.30pm

Mortality of Jewish population of the United Kingdom
L. Daniel Staetsky, Institute for Jewish Policy Research, London

Mortality of Jewish population of the United Kingdom Mortality of Jews is different from mortality of population surrounding them. This is true of all Diaspora Jewish communities with existing mortality data, including British Jews. Staetsky (2011) showed that British Jewish mortality in 2001 Census year was lower than mortality of the total British population and lower than mortality of the top social classes in this population. This paper presents an up-to-date picture of British Jewish mortality for Census year 2011. Data on age and sex structure of the British Jews in 2011, based on the Census, were made available recently by the national statistical authorities in England and Wales and Scotland. Data on deaths are routinely collected by the Board of Deputies of British Jews. In this paper British Jewish mortality in 2011 is compared to mortality of the total British population, mortality of Jews in Israel and selected low-mortality countries. This exercise is repeated after adjustment for Census undercount and under-registration of deaths. A comparison to Jewish mortality a decade earlier is also made. The findings are interpreted with the reference to the ‘Jewish patterns of mortality’ , recently re-formulated by Staetsky and Hinde (2014). Staetsky, L. 2011. Mortality of British Jews at the turn of the 20th century in comparative perspective, European Journal of Population, 27 (3): 361-385. Staetsky, L. and Hinde, A. 2014. Jewish mortality reconsidered, Journal of Biosocial Science (forthcoming).


Low immigrant mortality in England and Wales: selection or data artefact? Findings from the ONS Longitudinal Study 1971-2001.
Matthew Wallace, Hill Kulu, University of Liverpool

Introduction: Low mortality rates for international migrants have been documented in many industrialised countries. This mortality advantage is often attributed to selection processes in immigration, emigration and re-migration, and to protective health behaviours among immigrants. Data issues are also prevalent in migrant-mortality research. If return migration from the host country is under-recorded, a numerator-denominator bias emerges leading to an under-estimation of migrant mortality. Low mortality due to poor data quality is known as a ‘data artefact’. Aims: The aim of this study is to investigate mortality differences by country of birth in England and Wales and to determine whether variation in mortality rates can be attributed to selection or data issues. Registration issues are intrinsic to register data; controlling for both entry and exit uncertainty in data setup will allow us to determine whether low immigrant mortality observed in many industrialised countries is due to positive mortality selection or data artefact. Data and Methods: We use survival analysis to study mortality rates of immigrants relative to those of the ‘native’ population of England and Wales in the ONS Longitudinal Study (ONS LS). Findings: We find that first-generation, international migrants have lower mortality than the England and Wales population. While registration issues may generate some denominator bias; calculation error cannot account for any migrant mortality advantage. Controlling for the socioeconomic characteristics of individuals plays a significant role in emphasising and uncovering low mortality. We also find that immigrants from neighbouring countries have higher mortality than the England and Wales population.


Urban/rural variation in the influence of widowhood on mortality risk
David M Wright, Dermot O’Reilly Centre for Public Health, Queen’s University Belfast, Michael Rosato, Bamford Centre for Mental Health and Wellbeing, University of Ulster, Londonderry

Death of a spouse is associated with increased mortality risk for the surviving partner (the widowhood effect), although the mechanisms driving the effect are poorly understood. We investigated whether social support at both the household and community levels moderated the influence of spousal bereavement on mortality risk. We assembled death records spanning almost nine years for a prospective cohort of 296,125 married couples resident within Northern Ireland. Presence of other adults within the household and urban/rural residence were used as measures of support at the household and community levels, with support perceived to be strongest in rural areas. We used Cox proportional hazards models to estimate the effects of widowhood, sex, household composition and urban/intermediate/rural residence on all-cause mortality. Elevated mortality risk during the first six months of widowhood was found in all areas and for both sexes (range of hazard ratios 1.24, 1.57). After more than six months the effect among men was attenuated in rural but not urban areas (HRs and 95%CIs 1.09 [0.99, 1.21] and 1.35 [1.26, 1.44] respectively). Among women the effect was attenuated in both rural and urban areas (HRs 1.06 [0.96, 1.17] and 1.09 [1.01, 1.17]). The widowhood effect was not associated with household composition. We found some support for the hypothesis that informal social support reduces the impacts of spousal loss. Rural residence had a positive effect but presence of other adults in the household had no effect. We identify men in urban areas as being at greatest long-term risk.


A longitudinal population-based analysis of couple dynamics and mortality in KwaZulu-Natal, South Africa 2001-2011
Melanie Frost, Oxford Institute of Population Ageing, University of Oxford, Victoria Hosegood 1,3, Nuala McGrath 1,2,3, 1 Department of Social Statistics and Demography, University of Southampton, 2 Academic Unit of Primary Care and Population Sciences, University of Southampton, 3 Africa Centre for Health and Population Studies, University of KwaZulu-Natal, Somkhele, South Africa

Background: While multiple studies have shown that marriage has a protective effect in higher income settings, research is lacking in the African context. This study looks at the effect of different relationship states on mortality, in the context of HIV and the introduction of ART in rural South Africa. Method: The Africa Centre Demographic Information System (ACDIS) was used, which is an ongoing demographic surveillance system in KwaZulu-Natal. Marriage rates are very low in rural South Africa. Relationship status was defined as unpartnered, non-conjugal or conjugal relationship where conjugal relationships are defined as those where both partners belong to the same household. Age-adjusted Cox proportional hazards models were used to estimate the association between relationship status and mortality; models were produced separately for males and females aged 20+, and time varying coefficients were used to allow the relationship to differ before and after the introduction locally of ART in 2005. Results: Before 2005 (and the introduction of ART), being in a conjugal relationship was associated with a significantly lower hazard of death for females in all age groups and all but the youngest males. Being in a non-conjugal relationship was associated with a higher hazard of death amongst both males and females. After 2005 the protective effect of being in a conjugal relationship was substantially weaker. Conclusion: We see a strong association between relationship status and mortality only before ART became available. Additional analyses will explore the association of relationship status with specific causes of death (AIDS vs non-AIDS deaths).


Health: the geography of health - Monday 8 September 4.45pm

Individual level socio-economic factors explaining health inequality at older ages in England and Wales
Pia Wohland, Carol Jagger, Institute for Ageing and Health, Newcastle University, Vikki O'Neill, Fiona Matthews, MRC Biostatistics Unit, Institute of Public Health, Cambridge

Research questions: What are the regional differences in health expectancy at older ages within England and Wales and what role play individual-level socio-economic factors? Methods: Three health expectancies – disability free life expectancy, healthy life expectancy and cognitive impairment free life expectancy were calculated by age and gender from longitudinal data using increment-decrement life-table method as implemented in the IMaCh software. Data: MRC CFAS 1 (Cognitive function and ageing study) is a population based longitudinal study in England and Wales of individuals aged 65 years and over living in the community and in institutions with over-sampling of those aged 75 years and over. Data from five centres-Cambridgeshire, Gwynedd, Newcastle, Nottingham and Oxford - were used for the estimation of health expectancies by region, educational attainment, social class and disease burden. Results: Overall, disability free life expectancy (DFLE) for women aged 65 varied by about 2 years across centres; Gwynedd had the lowest DFLE of about 12 years and Cambridgeshire the highest with about 14. However, across educational attainment groups, inequality increased significantly to almost 4.5 years. Our presentation will investigate further inequalities for different health measures (good health and cognitive impairment) and other ages as well as inequality by social class and the policy implications our findings have for an ageing population.


Towards a New Index of Multiple Area-Level Deprivation for Auckland, New Zealand
Daniel J Exeter, Jinfeng Zhao, Michael Browne, Epidemiology & Biostatistics, School of
Population Health, The University of Auckland

For the past 20 years, the New Zealand Index of Deprivation (NZDep), has been the universal measure of area-based social circumstances for New Zealand and often the key social determinant used in population health and social research. NZDep2006 is a composite measure of deprivation derived from Census data, representing nine aspects of material deprivation, ranging from household income to telephone access. Technological developments over the past decade have seen significant improvements in the completeness of routine data in New Zealand, with more than 97% of the population included in the key health databases since 2007. These enriched individual-level datasets facilitate the creation of social indicators that may be regularly updated relatively easily, without the need for an expensive national survey such as the Census. This research presents the results of a feasibility study conducted in Auckland, New Zealand, to determine whether an alternative measure of deprivation could be created using routine administrative data. Using the Scottish Index of Multiple Deprivation as a template, we first developed a geographical boundary file by aggregating small census areas (Meshblocks) into areas called “Lower Zones” with populations ranging between 500 and 999 (mean=750). Next, data from national health, social development, education, and police databases were used to create over 25 indicators categorised into income, employment, education, crime, health, housing, and access domains, and an overall index of multiple deprivation. We demonstrate the geographical variations in the 7 domains and the overall index, and describe their association with health outcomes.


The pits may be closed but adverse effects on health may still be evident: exploring the links between coal mine locations, socioeconomic deprivation and self-reported health in Great Britain
Paul Norman, School of Geography, University of Leeds

Various studies in the UK have found associations between living in a coalfield area and poor health at both population and individual levels. A database of information on GB coal mines has recently become available and this enables the investigation of a more spatially disaggregate study than has previously been possible. In a context in which ‘traditional’ coal mining in GB no longer exists and the locations are invariably characterised by deep socioeconomic deprivation, the current study seeks to find: • Whether including information on the location of coal mines in and near electoral wards (and equivalents) provides an improved explanation of small area variations in health; • Whether time since pit closure has any observable influence; • Whether regeneration schemes have any apparent effect on community health. In 2001, for Census Area Statistics Wards in England and Wales and Postal Sectors in Scotland, the initial findings are that (controlling for socioeconomic deprivation) self-reported long-term limiting illness health is significantly worse at both small area and district level if a coal mine has been in that locality and especially if there are multiple pits. Poor health decreases with increasing distance from a coal mine. In a case study focussing on North-East England, the longer a local pit has been closed, the better the long-term health of the community. However, regeneration grants are associated with relatively poor health, perhaps because these areas are the most deprived and / or because these locations have lost their relatively population by out-migration.


A comparison of hypertension health care outcomes among older people in the US and England
Alan Marshall, J Nazroo, K Feeney, J Lee, B Vanhoutte, N Pendleton, University of Manchester

This paper describes differences in hypertension health care outcomes among older people (50+) living in the US and England, including stratification by wealth and (US) health insurance (public, private or no health insurance). After controlling for hypertension risk factors the proportion of hypertensives with controlled hypertension is higher in the US compared to England a difference driven by the relatively lower levels of undiagnosed hypertension in the US. The prevalence of uncontrolled hypertension is very similar in each country, however, in the US (and not England) uncontrolled hypertension declines with rising wealth. Hypertension health care outcomes are comparable across public (Medicare/Medicaid) and private insurance categories in the US but,as might be expected, those with no insurance appear to receive poorer care. Surprisingly, in both countries, undiagnosed hypertension is positively correlated with wealth (ages 50-64). It is possible that different practices around the diagnosis/treatment of hypertension drive the observed country differences in undiagnosed hypertension. It is unclear whether the lower levels of undiagnosed hypertension confer cardiovascular health advantages to US older people. The increasing levels of undiagnosed hypertension with wealth in each country may reflect less frequent contact with health professionals amongst the most affluent and diminished opportunity for blood pressure tests. Our results on uncontrolled hypertension suggest that Government health systems perform at least as well as private health care and are more equitable in the distribution of care outcomes across wealth quintiles.


Health: maternal and child health inequalities - Tuesday 9 September 9.00am

Inequities in Maternal Health in Cambodia: The Effect of Health Equity Funds in Urban and Rural Areas
Kristine Nilsen, Andrew Amos Channon, Department of Social Statistics and Demography, University of Southampton

New public health interventions are often associated with increasing health inequities at first, as they tend to reach the economically disadvantaged and peripheral groups last – the inverse equity hypothesis. This includes health interventions in the context of universal health coverage (UHC), unless provisions are explicitly made for vulnerable groups. This paper assesses the impact of one innovative UHC strategy in Cambodia, health equity funds (HEF), on health inequity at different stages of implementation. Representative of many countries, Cambodia has transitioned from a country with limited health service provision to a dual system with private and public providers. Predominantly implemented in rural areas, HEFs are a social health protection scheme that provides care free of charge at public health facilities to those considered too poor to pay. Using DHS data from 2000, 2005 and 2010, methods adapted from income inequality metrics are used to assess the relationship between HEFs and maternal health care utilisation by comparing health inequities across the socioeconomic distribution by residency and over time. Contrary to expectations, findings show that health inequalities have declined since 2000. However, the largest declines took place between 2005 and 2010, coinciding with a substantial increase in HEF coverage. The decline in health inequalities is higher in urban compared to rural areas, suggesting that while HEFs have mitigated growing health inequities in rural areas, urban areas without HEFs are managing to reduce inequities faster. This may be due to rapid wealth increases among the urban population and increased availability of private health care.


High Infant and Child Mortality among Twins in Sub-Saharan Africa: Trends and Explanations
Christiaan Monden, University of Oxford, Jeroen Smits, University of Nijmegen

Sub-Saharan Africa has the highest twin rates in the world and the highest level of infant and child mortality. Twins are particularly vulnerable in this region: 30% do not survive only age 5. About a quarter of million infants and children who die each year are part of a multiple birth. Using DHS surveys we document patterns and trends in twin mortality (vs singletons) and we assess demographic, material and cultural explanations for variations in twin mortality. We answer four questions: What has happened over time to under-5 mortality among twins (vs. singleton)? Can the ‘usual suspects’ of poverty, maternal education, maternal age, and parity explain excess mortality among twins? Are twins more vulnerable to known determinants of under-5 mortality? Is excess twin mortality affected by regional levels of development, regional twin rates, or cultural views of twins? Twin mortality has decreased but less rapidly than singleton mortality. Most of the excess mortality seems driven by biological and epidemiological factors in the first month of life. Lack of medical assistance seems to explain part of the twin-singleton difference too. Socio-economic regional factors do not seem to play an important role. We are currently adding information of cultural/ethnic differences in the acceptance of twins.


How feasible is the 80/40 equity target for skilled attendance at birth? An analysis of 50 Demographic and Health Surveys.
Andrew Channon, Sarah Neal, University of Southampton

The post-2015 development agenda has a strong equity focus, while the drive towards Universal Health Coverage (UHC) also is based around the reduction of inequality. The World Health Organisation and the World Bank have proposed a framework for monitoring global progress towards UHC - by 2030 at least 80% of the poorest 40% of the population will have access to essential health services (the “80/40” target). This paper examines the progress towards this target using the presence of a skilled attendant at birth as a maternal health indicator. Using Demographic and Health Survey data from over 50 low and middle income countries the equity gap between the poorest 40% and the rest of the population is calculated. Countries are categorised into five groups based on overall skilled attendance levels. Results indicate that there are marked inequities between the bottom 40% and the rest in all groups except when country coverage levels are already high. The average annual rates of change needed for each group to reach the goal highlight that as the 80/40 target is an absolute one, countries with currently very low coverage are required to make rapid and sustained progress. Average annual rates of change for the poorest 40% are compared with historical progress trends. The results indicate that the target is achievable mainly in countries where there is already a lower level of inequality, while many countries, at current levels of progress, will not achieve the goal of reducing inequality to the proposed level by 2030.


Do Girls Have a Nutritional Disadvantage Compared with Boys? Breastfeeding, Food Consumption, and Mortality among Indian Siblings
Jasmine Fledderjohann 1, Sutapa Agrawal 2, Sukumar Vellakkal 1,2, Sanjay Basu 3, Oona Campbell 4, Pat Doyle 4, Shah Ebrahim 2,4, David Stuckler 1,2, 1 Department of Sociology, Oxford University, 2 South Asia Chronic Disease Network, Public Health Foundation of India, New Delhi, 3 Stanford Prevention Research Center, Stanford University School of Medicine, 4 Department of Non-Communicable Disease Epidemiology, London School of Hygiene & Tropical Medicine

Background: India is the only nation where young girls have greater risks of under-5 mortality than boys. We test whether female disadvantage in breastfeeding and food allocation can account for India’s gender disparities in child mortality rates. Methods: Multivariate regression and Cox models were performed using Round 3 of India’s National Family and Health Survey (2005-2006). Models were disaggregated by birth order and sibling gender, and adjusted for maternal age, education, and fixed effects, urban residence, household deprivation, and other sociodemographics. Mothers’ reported practices of WHO/UNICEF recommendations for breastfeeding initiation, exclusivity, and total duration (ages 0-59 months), children’s consumption of 24 individual food items (6-59 months), and child survival (0-59 months) were examined for first- and secondborns (n = 20,395). Findings: Young girls were breastfed on average for 0.45 fewer months than boys. There were no gender differences in breastfeeding initiation (OR = 1.04) or exclusivity (OR = 1.06). Compared with boys, girls had lower consumption of fresh milk by 14% and breast milk by 21%. Each additional month of breastfeeding was associated with a 24% lower risk of mortality. Girls’ shorter breastfeeding duration accounted for about an 11% increased probability of dying before age 5, accounting for about 50% of their survival disadvantage compared with other low-income countries. Conclusions: Young Indian girls are breastfed for shorter periods than boys and consume less fresh milk. Future research should investigate the role of additional factors driving India’s female survival disadvantage.


Health: Emerging issues in transitional societies - Tuesday 9 September 1.30pm

Sociodemographic and Health Characteristics over the Life Course and Cognitive Health among Older Adults in Brazil
Flavia Cristina Drumond Andrade, Department of Kinesiology and Community Health University of Illinois at Urbana-Champaign, Ligiana Pires Corona, Department of Epidemiology University of Sao Paulo, Sa Shen, College of Applied Health Sciences University of Illinois at Urbana-Champaign, Yeda Aparecida de Oliveira Duarte, Department of Medical Surgical Nursing School of Nursing, University of Sao Paulo, Maria Lúcia Lebrão, Department of Epidemiology School of Public Health, University of Sao Paulo

A major health problem for a growing elderly population in Latin America is cognitive decline. The aims of this paper are to identify the trajectories of cognitive decline and to investigate a how demographic, social, and health conditions influence cognitive transitions among older adults in the largest city in Brazil. Data from the longitudinal survey SABE (Health, Well-being, and Aging), conducted in 2000, 2006 and 2010, were analyzed. At baseline, 2,143 people aged ≥ 60 years were interviewed in São Paulo. Cognitive health was assessed based on a modified version of the Mini-Mental State Examination. A fully conditional model was applied to model the trajectory of MMSE. Mixed-effect pattern-mixture model was selected to fit the data based on the survival. Results indicate that cognitive levels decline age increases, with a significant quadratic effect for age. Those who survived had higher levels of MMSE across age than those who died during the follow up. Older age (coef =-0.054, p<0.0001), stroke (coef =-0.716, p<0.0001) and having depressive symptoms (coef =-0.495, p<0.01) were associated with declines in cognitive health. Higher educational attainment (coef=0.154, p<0.0001) was associated with better cognitive health. Among women, higher BMI (coef=0.025, p<0.05) was associated with better cognitive health. Identifying risk and protective factors associated with cognitive health is critical for the development of strategies for early identification and the promotion of effective public policies.


Geographic variation in diabetes and detection across China: multilevel evidence from 98,058 participants in the China Chronic Disease and Risk Factor Surveillance System 2010
Maigeng Zhou 1, Thomas Astell-Burt 2, 3, Yufang Bi 4, Xiaoqi Feng 5, Yong Jiang 1, Yichong Li 1, Andrew Page 2, Limin Wang 1, Linhong Wang 1, Wenhua Zhao 6, Guang Ning 4, 1 National Center for Chronic and Noncommunicable Disease Control and Prevention, Chinese Center for Disease Control and Prevention, Beijing, 2 School of Science and Health, University of Western Sydney, 3 School of Geography and Geosciences, University of St Andrews, 4 Key Laboratory for Endocrine and Metabolic Diseases of Ministry of Health, State Key Laboratory of Medical Genomics, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, E-Institute of Shanghai Universities; Shanghai Clinical Center for Endocrine and Metabolic Diseases, Shanghai Institute of Endocrine and Metabolic Diseases, Department of Endocrine and Metabolic Diseases, Rui-Jin Hospital, Shanghai Jiao-Tong University School of Medicine, Shanghai, 5 Centre for Health Research, School of Medicine, University of Western Sydney, 6 Chinese Center for Disease Control and Prevention, Beijing

Background: The spatial patterning of diabetes across China has received little attention. Method: A sample of 98,058 adults aged 18 years and older was extracted from the China Chronic Disease and Risk Factor Surveillance System 2010. Socio-demographics, health and lifestyle-related variables were augmented with blood samples collected from all participants following an overnight fast of at least 10 hours. Diabetes status was defined using American Diabetes Association criteria. Diabetes detection was defined as positive responses to self-reported previous diagnosis of diabetes among the sub-sample identified as having diabetes. Geographic variations in each outcome were assessed using multilevel logistic regression and median odds ratios (MOR). Differences between urban and rural areas and by socioeconomic circumstances (SEC) were explored, adjusting for age, gender and behavioural variables. Results: MORs indicated a two-fold level of geographic variation in each outcome across China. Compared with the south, diabetes was more common in the north, east, and southwest. Within the northern region, diabetes prevalence ranged from 9.1% in the northwest and 9.9% in the northeast up to 15.7% in the north. A clear negative gradient in diabetes prevalence was observed from urban high SEC to rural low SEC. Adjusting for health literacy and other person-level characteristics attenuated this gradient but geographic variation remained substantial. Detection were less common in the southwest and northwest than the south, but higher in the north. The odds of being diagnosed were substantively lower in rural low SEC areas. Conclusion: There is substantial geographic patterning of diabetes and detection across China.


Effectiveness of the Cardiovascular project on glycemia control in Davao City, Philippines
Sophie Pilleron, Aude Brus, Davide Olchini, Handicap International, Lyon, France, Richard Erick Caballero, CVD project, Handicap International, Davao City, Philippines

Objective: To assess the effectiveness of the Cardiovascular (CVD) Project in Davao City, Philippines regarding HbA1c, body mass index (BMI), waist circumference (WC), blood pressure (BP), physical activity level and diabetes knowledge.

Methods: A non-randomized cross-sectional survey was conducted in five intervention and five comparison villages of Davao City in 2013. Diabetics aged ≥18 and registered in village health centers were included. Subjects from intervention area had to have participated in at least 2 Diabetes and Heart Days (DHD) in the past 6 months and those from comparison area in no DHD. Evaluation indicators included: BMI <23 kg/m2, WC<80 cm in females and 90cm in males, and BP <130/80 mmHg, HbA1c<6.5%, moderate physical activity according to International Physical activity questionnaire. The Diabetes Knowledge using the Diabetes, Hypertension and Hyperlipidemia knowledge instrument were used to assess diabetes knowledge level. Multivariate logistic regression models were used to compare the two areas.

Results: No evaluation indicators differed between the intervention (n=123) and the comparison areas (n=160). However, Hba1c mean was significantly lower in the intervention area (8.6±2.1 vs 10.0±2.6, p<0.001) and the percentage of patients with HbA1c<7% was significantly higher in the intervention area (OR=1.98 95%CI: 1.01-3.69; p=0.048). Besides, patients from intervention area had a better diabetes management regarding number of medical visits, the percentage of patients who consulted a nutritionist or attended education session.

Conclusion: Despite limitations, our results provide with arguments of effectiveness of the CVD project on glycemic control but not on control of other cardiovascular risk factors.


Understanding health and mortality - Tuesday 9 September 4.45pm

Is Equality a Double Edged Sword? Gender Equity and Gender Differences in Life Expectancy in OECD Countries over Time.
Ross Macmillan, Adrian Dearriba, Federica Querin, Dondena Centre for Research on Social Dynamics, Bocconi University

Virtually every indicator of gender equality in the social sciences indicate that women have gained ground, socially, politically, economically, and culturally, over the past 30 years. We know little about the implications of this for long standing differences in mortality risk and life expectancy. Moreover, theoretical expectations offer differing ideas about what gender equality might mean. From one perspective, if women's day to day lives start to resemble men's, we would expect that gender differences would converge. Alternatively, if gender equality decreases discrimination and increases access to socioeconomic positions that diminish mortality risk, women should gain life expectancy relative to men. We test these these with time and country fixed effects models and find little evidence that gender equality has altered life expectancies. Moreover, further investigation suggests that gender differences are rooted in cohort specific behaviors (e.g., smoking) and that contemporary changes in the socio-politics of gender will play little role in altering dynamics of life expectancy. While this does not change the importance of social programs promoting gender equality, it does suggest that population health dynamics need to be attuned to both the past and the present in developing understandings of the complexities of gender and health.


Planning for demographic change in the design, evaluation and transportability of large-scale interventions.
G.J. Melendez-Torres, Centre for Evidence-Based Intervention, Department of Social Policy and Intervention, University of Oxford, Stuart Basten, Oxford Centre of Population Research, Department of Social Policy and Intervention, University of Oxford

While demographers have been long concerned with the effects of social policy on demographic change, there is an important and apparently unfulfilled role for them in the design, evaluation and transport of large-scale interventions. Large-scale interventions address structural or social determinants of health, or are interventions at the community level to address individual health and well-being. Based on an inductive analysis of 63 evaluations from 39 studies in four systematic reviews published by the Cochrane Collaboration Public Health Group in 2012 and 2013, we develop a framework for the role of planning for demographic change in the design and transportability of large-scale interventions. Demographers have a key role to play in theory-based and process evaluation of interventions, especially as evaluations, including randomised trials, move towards theorising the relationships between intervention context, implementation, and outcomes. Previous evaluations have focused on influences of social, cultural and resource contexts on intervention implementation and effect. However, four key demographic considerations underlie the intersection of demographic characteristics and change in design and transport of large-scale interventions. These are: • baseline demographic characteristics that may interact with the intervention, • demographic processes (including through modelling and projections) that may alter long-run sustainability and effectiveness of the intervention, and • similarity between populations where the intervention was previously evaluated and populations to which the intervention will be transported on both baseline characteristics and demographic processes that may impact whether similar effects in both short-term and long-term horizons are to be expected.


Deriving local level maternal mortality estimates from census data an illustration from Ghana.
Philomena Nyarko, Ghana Statistical Service, Zoe Matthews, Claire Bailey, Allan G. Hill, University of Southampton

Reducing maternal mortality is one of the key Millennium Development Goals. However, the estimation of maternal mortality levels for countries with incomplete vital registration data and poor quality cause of death attribution remains an important challenge.. In addition to the well-known problems of measuring adult mortality in general, the uncertainty surrounding the identification of maternal deaths (however defined) leads to wide confidence limits around any national estimate. There is thus an important trade-off between the use of large surveys and census data which have the advantage of a large number of cases and more detailed epidemiological enquiries that allow the collection of possibly higher quality and certainly more detailed information on the circumstances and possible causes of each adult death. Both direct (based on recent household deaths) and indirect (sibling survival methods) are used to bridge this gap. Even with good quality data and appropriate estimation methods, the results can often produce widely varying estimates. In this paper, we first present national level estimates of maternal mortality for Ghana using the 2010 census data which recorded overall and maternal deaths in the 12 months prior to the census. We illustrate the details of the adjustments for a) all adult male and female deaths; b) maternal deaths; c) recent births; and compare the final estimates with external sources – the Ghana Maternal Mortality Survey, the recent DHS surveys and other official figures. Using the same methodology, we illustrate how some sub-national estimates may also be derived when using census rather than survey data. From the service provision standpoint, however, the ideal remains the production of local level or even facility-based estimates of maternal death rates since investment in improved maternity care and services is necessarily a local decision. In the second part of the paper, we show how the correlates of a maternal death can be used to estimate maternal mortality rates for much smaller sub-populations and geographical units. The method produces estimates of maternal mortality both by socio-economic class (household wealth, maternal education and employment) and for lower geographical units. The two methods illustrated here provide a template for other parallel analysis for other countries and provide the basis for socio-economic and geographical targeting of high maternal mortality sub-groups in future interventions.


Exploring factors associated with completeness of parental survival data in a longitudinal surveillance system
Gabriela Mejia-Pailles1,3, Victoria Hosegood1,2,3, Ann Berrington1,2, 1ESRC Centre for Population Change, University of Southampton, 2Department of Social Statistics and Demography, School of Social Sciences, University of Southampton, 3Africa Centre for Health & Population Studies, University of KwaZulu-Natal

In the context of HIV treatment in sub-Saharan African, reliable data on the levels and trends of orphaning are needed to understand the impact of improved survival of HIV-infected parents and changes in HIV incidence and prevalence. In addition, these data are also needed to monitor the effectiveness of intervention programmes and policies directly or indirectly targeting orphans (Madhavan et al., 2014). Orphaning during childhood has been found to be disadvantageously associated with numerous health and welfare outcomes including education (Guo & Sherr, 2012), mental health (Cluver et al, 2012; Anwine et al, 2005), and early sexual debut (Operario et al., 2011). The most widely used parental survival data is that available from cross-sectional surveys or censuses. However, in communities in southern Africa where orphaning rates are high and many children are not co-resident with living parents, there are many reasons to anticipate that intentional and unintentional misreporting in cross-sectional surveys may bias the accuracy of orphanhood estimates. The circumstances in which people misreport a child’s parental survival status are not well understood. In this paper, we examine the consistency of parental survival status over time using longitudinal data from a demographic surveillance system in rural KwaZulu-Natal, South Africa, to update previously published data on orphanhood prevalence before and after public HIV treatment was scaled-up in the surveillance area. We explore some of the main factors associated with completeness of parental survival status and propose a series of alternative methodological approaches to deal with the issue.


Inequality and mortality increase at older ages: a paradox?
Jon Anson, Department of Social Work, Ben-Gurion University of Negev, Israel

As living standards improve, mortality declines, and the age-pattern of human mortality changes. Yet, while much of human mortality decline can be related, directly and indirectly, to changes in the material standards of living, differences in human mortality, and in the shape of the mortality curve, cannot be explained in terms of material standards of living alone. Different social conditions, levels of technology and patterns of social relations, between social groups and between men and women, will result in a different age structuring of mortality. How, then, can we describe and explain these differences in the shape of the mortality curve?

The past 50 years has seen the emergence of a new phase in the mortality transition. Survivorship to middle age is still increasing, but slowly. With 95 percent and more of those born surviving to age 50, the locus of change is moving to older ages: the rate at which mortality increases in late middle age; and its corollary, the pushing back of the median age at death. At these ages the mortality curve is Gompertzian (log-mortality is a linearly increasing function of age) and the critical question becomes: what sets the slope of the curve, that is, the rate of ages? In part, this is determined by the level of survivorship to age 50 and the mortality rate at that ages. Given that the variation in mortality at very high ages (100 and above) is far less than at younger ages, the lower mortality is at age 50, the greater must be the rate of increase past that age. However, social effects are also important in setting the rate of ageing, adjusted for the mortality rate at age 50. Using a cross section of current national life tables, we show that mortality at middle age, and the adjusted rate of ageing past that age, are lower in wealthier countries and in countries with a greater degree of democracy. The effects of inequality, however, do not show the same consistency. Inequality is associated with greater mortality at younger ages, but also with a lower adjusted rate of ageing at older ages. We discuss reasons for this apparent contradiction in terms of selective mortality at younger ages.


Health: Obesity and nutrition - Wednesday 10 September 11.00am

The main determinants of the healthy life style at the age of 60 and older micro and macro level indicators?
Vladimir Kozlov, National Research University - Higher School of Economics

The main goal of this research is to find how healthy life style indicators are influenced by the individual social economic and demographic characteristics of the respondent and the macrolevel data showing the development of the countries. I observe the respondents on the age of 60 and older, using as an empirical database for the research “ISSP – 2011. Health Survey”. The healthy life style is based in my research in 3 pillars: anti-goods like tobacco and alcohol consumption; physical activity and healthy food consumption. The determinants for it could be found on both micro and macrolevels. On the microlevel we observe the social-economic and demographic variables, while on the macrolevel we can see the spending on health care, demographic indicators (mainly life expectancy) and characteristics of social economic development of the country, as well as their combinations like HDI. For testing the hypothesis we use the linear multilevel models. According to our preliminary results for the regression with country fixed effects the on the microlevel the education level is negatively correlated with unhealthy habits like smoking, but positively correlated with alcohol consumption (apart from regular drinking), physical activity, healthy food consumption. Respondents without partners tend to smoke and drink more, eat less healthy food and prefer less physical activity. Female respondents though doing less physical activity

have healthier habits in food and anti-goods consumption.


Physical inactivity among older adults: implications for life expectancy among non-overweight and overweight or obese individuals
Ngaire Coombs1,2,3, E Stamatakis1,2,4,5,6, I-M Lee7, 1 Department of Epidemiology and Public Health, University College London, 2 PARG (Physical Activity Research Group), Population Health Domain, University College London, 3 Department of Social Statistics and Demography, University of Southampton, 4 Prevention Research Collaboration, School of Public Health, University of Sydney, 5 Charles Perkins Centre, University of Sydney, 6 Exercise Science Discipline, Faculty of Health Sciences, University of Sydney, 7 Division of Preventive Medicine, Brigham & Women’s Hospital and Harvard Medical School

RESEARCH QUESTION To what extent is physical inactivity in older age is associated with reduced life expectancy among non-overweight (BMI of <25 kg/m2) and overweight/obese (BMI ≥ 25 kg/m2) individuals? METHODS Data from 20,203 respondents aged ≥60 years (2,671 deaths) were drawn from the Health Survey for England and Scottish Health Surveys baseline examinations (1994 to 2004) and 5-year all-cause mortality linkage. Physical inactivity was defined as not meeting the current UK physical activity guidelines for older adults (150 minutes/week moderate or 75 minutes/week vigorous intensity or combinations expending equivalent energy). Cox proportional hazard ratios were used to calculate the sex-specific fractions of mortality attributable to physical inactivity in the two BMI categories. These fractions were applied to sex-specific life tables for Great Britain (1998-2000) to estimate years of life lost due to physical inactivity. Results were stratified by BMI status, and adjusted for age, self-rated health, longstanding illness, ethnicity, and smoking status. RESULTS Not meeting physical activity guidelines accounted for 4.9 (95% CI 3.1 to 6.5) years of life lost among non-overweight females and 2.8 (1.1 to 4.3) years of life lost among non-overweight males at age 60. For the overweight/obese group years of life lost were 1.5 (-0.33 to 3.13) and 1.3 (-0.09 to 2.57) for females and males, respectively CONCLUSION Not meeting the UK physical activity guidelines in older age was associated with reduced life expectancy that is larger among non-overweight individuals than among overweight/obese persons.


Who are the obese? A cluster analysis exploring subgroups of the obese
Mark Green School of Health and Related Research (ScHARR), University of Sheffield

Background: Identifying individuals as obese classifies them based upon their weight, however such a distinction fails to account for the variation within this group across other factors such as health, socio-economic and behavioural characteristics. Current research is therefore constrained in its understanding of obesity, ignoring how factors are inter-related within the group. The study aims to examine the existence of subgroups of obese individuals. Methods: Data were collected from the South Yorkshire Cohort (2010-2012), a longitudinal observational study. Information on demographic characteristics, health status, wellbeing, health- and obesity-related behaviours were selected as inputs for individuals aged 16 to 85. Only individuals with a body mass index of greater than or equal to 30 were considered (n=4144). A two-step cluster analysis was used to explore the existence of individuals with similar characteristics within the data. Results: The cluster analysis showed six distinct groups of individuals within those classified as obese. These subgroups included a cluster of males with high alcohol consumption, two groups of healthy individuals which varied by age, two poor health clusters that differed by the types of long-term conditions found in each and a final cluster which displayed the worst health, exercise and socio-economic characteristics. Conclusions: There are clear subgroups of individual types within the obese BMI group, showing that studies which assume the obese to be similar are at risk of committing an ecological fallacy. It is important to account for this variation in experiences when devising policy interventions as each group will require a different approach.


Parental education and child obesity, differentiating the role of mothers and fathers
Nichola Shackleton, Institute of Education, University of London

This paper considers the role of parental education on child obesity, focusing on the independent association between mother’s education and child overweight status and father’s education and child overweight status. Data from the Millennium Cohort Study (approximately age 7) are analysed. A sub-sample of families with two biological parents is used in the main analysis. Child weight status was measured using the IOTF criteria. Sequential logistic regressions are used to assess i) whether the previously reported association between mother’s education and child weight status is confounded by the exclusion of father’s education, ii) whether there are gender specific relationships between parental education and child weight status and iii) whether the association between parental education and child weight status is independent of other socio-economic factors. The results suggest that father’s education has a stronger association to child overweight status than mother’s education. The findings were consistent with different measures of parental education and were replicated in another UK birth cohort.