Strand organisers: Paula Griffiths, Loughborough University, and Nicola Shelton, University College London
Comparing health inequalities across time and place: an analysis of Demographic and Health Survey data
Kath Moser, David A Leon
London School of Hygiene & Tropical Medicine
Describing health inequalities in one setting at one time is relatively straightforward. However, establishing whether inequalities are a) larger in country A or B and b) increasing or decreasing is not straightforward and raises significant methodological challenges which to date have not been adequately addressed. We investigate how the choice of measure of inequality magnitude affects the assessment of differences and trends in inequality.
This study of inequalities in under-five mortality analyses Demographic and Health Survey data from 22 low- and middle-income countries (11 in Africa, 5 in Latin America and Caribbean, 6 in Asia), each with 2 surveys 1991-2001. Rate ratios and rate differences are used to describe inequalities between household wealth quintiles.
The ranking of countries by inequality in under-five mortality depends on whether rate ratios or differences are used. The association between the inequality ranks produced by the two measures is poor. In 14 countries rate ratios and differences indicate the same trend in inequality. However the rate difference indicates decreasing inequality in nine countries, the rate ratio in only five. There is a positive relationship between rate differences and average under-five mortality, an inverse relationship between rate ratios and average under-five mortality.
Assessing differences and trends in inequality depends in part on how they are measured. This is problematic when making international comparisons, monitoring trends, and informing policy. We clearly need to use both absolute and relative measures, and compare the findings. In describing inequalities we need to be explicit about how they have been measured.
Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT.
Email: kath.moser@ons.gov.uk
Which aspects of socio-economic status at birth and nine years are most associated with child nutritional status at the end of childhood?
Paula Griffiths1, Noel Cameron1, Shane Norris2, and John Peffifor2
1Loughborough University, 2University of the Witwatersrand
Background
In the last decade the number of deaths from non-communicable diseases in developing countries has exceeded those observed in developed nations. Social class or socio-economic status (SES) is known to be associated with non-communicable diseases including cardiovascular disease, diabetes, some cancers, metabolic syndromes, arthritis, gastrointestinal disease, obesity, as well as adverse birth outcomes, malnutrition, stunting, accidental and violent deaths, child growth and infectious diseases of childhood. However, less is known about the relative contribution of SES to health outcomes throughout the lifecourse.
Aim
To determine which aspects of household socio-economic status measured at both birth and nine years are most important in predicting body composition outcomes at nine years.
Methods
This paper uses data from Birth to Twenty (BTT); the largest and longest running cohort study of child health and development in Africa. Using data from a sub-sample of children with both complete anthropometric and socio-economic measures collected at birth and nine years (n=242), regression models are used to estimate the influence of measures of SES at birth and age nine years on anthropometric outcomes at the end of childhood.
Results
Findings reveal that measures of SES at nine years are stronger predictors of anthropometric outcomes at the end of childhood than SES measures assessed at birth.
Discussion
These findings are not consistent with other studies that have assessed adult health and body composition outcomes. Adult studies show that measures of SES at birth/early childhood are stronger predictors of adult health outcomes than those assessed in adulthood. However, adult studies commonly do not have any SES measurements recorded between early life and adulthood. This study therefore suggests that measures of SES taken in late childhood may be more important in explaining body composition outcomes at age 9 than measures taken at birth.
1Centre for Human Development and Ageing, Department of Human Sciences, Loughborough University, Loughborough, Leicestershire, LE11 3TU
2University of the Witwatersrand Medical School, 7 York Road,
Parktown 2193, South Africa
Spatial inequality in delivery care uptake in Ghana: a multilevel analysis
Fiifi Amoako Johnson, James Brown, Sabu Padmadas
University of Southampton
Background: Poor utilization of maternal health services has been identified as one of the determinants of high maternal and infant mortality in Sub Saharan Africa. The recent Ghana Demographic and Health Survey (GDHS) showed that about 54% of deliveries take place at home under unhygienic surroundings in the presence of untrained birth attendants.
Objective: This study aimed to investigate the spatial variations within and across communities on the utilization of delivery care services in Ghana.
Methods: Data from the 1998 and 2003 GDHS and the 2000 Ghana Population and Housing Census have been used to identify the individual and contextual level factors that explain the levels of unobserved heterogeneity associated with choice of delivery care services. Multilevel multinomial logistic regression techniques were used for the analysis that considered 2,342 and 2,757 mothers who had a birth in the last 5 years preceding the 1998 and 2003 surveys respectively.
Findings: The heterogeneity in delivery care utilization in public sectors was found to be persistently higher between rural communities in the 1998 and 2003 periods when compared to those in urban areas. In urban areas, public sector delivery care deteriorated in particularly those communities where service provisions were already poorer when compared to communities with better care [mij=0.41 (SE: 0.23) in 1998 and mij=1.47 (SE:0.24) in 2003]. The variations in the community effects were found to be highly significant (p<0.001) when statistically controlled for individual characteristics.
Conclusions: Clustering effects were highly significant with regard to utilization of delivery care services especially within rural communities.
Fiifi Amoako Johnson, Research Student, University of Southampton, School of Social Science, Division of Social Statistics, Southampton SO17 1BJ
Email: faj100@socsci.soton.ac.uk
Where not to live: a geo-demographic classification of mortality
Nicola Shelton1, Mark Birkin2, Danny Dorling3
1University College London, 2University of Leeds, 3University of Sheffield
The aim of this paper is to pilot a method for geo-demographic classification for mortality patterns in Britain. Age and sex directly standardized mortality ratios (DSMRs) for 100 grouped International Classification of Disease series 9 causes of death (ICD-9) were calculated. The 84 European Parliamentary (EP) constituencies as defined in 1999 were used as the spatial basis of this study to allow regional comparisons to be made while comparing units of similar population sizes. Scotland was excluded from the final analysis, leaving 76 regions. This paper is a preliminary investigation of the patterns of the causes of death over time and space in England and Wales using cluster analysis to summarise some of the structure in the data. Seven major and three minor cluster profiles were developed.
1Dr. Nicola Shelton, Department of Epidemiology & Public Health, University College London, 1-19 Torrington Place, London WC1E 6BT
Email: n.shellton@ucl.ac.uk
Amos Channon
University of Southampton
Retrospective large scale surveys, such as DHS, have enabled birth weight information to be collected for a representative sample of a population. However, many infants are still not weighed at birth and thus there is much missing birth weight information. The research question addressed is whether a mothers' perception of her child's size can be used as a proxy for birth weight in the calculation of the proportion of infants with LBW. Data are taken from 14 DHS countries with varying levels of missing birth weight information.
It is seen that mothers' who report a birth weight for their children differ markedly from those who do not in many characteristics, many of which are also predictors of birth weight. Furthermore, more children who are not weighed are classified as 'small' or 'very small' by their mothers than those who are weighed, and on an aggregate level the size categories correspond with mean birth weight. However, it is also seen that many weighed children are placed incorrectly into size categories, with many who have LBW being classified as being of average size or above.
Calculating the proportion of infants with LBW utilising the mothers' perception variable increases the percentage with LBW by an average of 13% across all countries. However, this is dependant on whether infants heaped on 2500g are treated as having LBW, as estimates differ greatly if the treatment of these infants is changed. Using the mothers' perceptions as a direct predictor of LBW is seen not to be accurate due to the extent of misclassification of infants into inappropriate size categories.
A. Channon, Department of Social Statistics, School of Social Sciences, University of Southampton, Highfield, Southampton SO17 1BJ
Email: arc102@soton.ac.uk
Neonatal mortality in developing countries: What can we learn from DHS data?
Sarah Hall, Division of Social Statistics, University of Southampton
While the last few decades have seen a significant decrease in child mortality within developing countries, the decline in neonatal mortality (NMR) has been less marked. Each year more than 4 million infants die in the first 28 days of life, mostly from causes that could be prevented through basic health care.
This paper explores the potential contribution of DHS data in improving knowledge of trends in neonatal mortality in developing countries. It briefly outlines the causes and possible consequences of sampling and non-sampling error in survey data of this nature, before using DHS and World Fertility Survey estimates to describe apparent trends in neonatal mortality over the last few decades. It also examines the association between neonatal mortality and per capita gross domestic product (GDP) at national level. The study draws out how both patterns of progress and relationship with GDP differ markedly in the neonatal period than in post-neonatal infancy and early childhood. It will conclude by summarising the potential limitations in using DHS estimates for NMR, as well as outlining potential factors underlying the relatively poor progress being made in reducing neonatal deaths.
Email: sarahhall@londone1.freeserve.co.uk