Areal Variation in Contraceptive Use in East Africa

Steve Clements, University of Southampton
Angela Baschieri, University of Southampton
Monique Hennink, University of Southampton
Nyovani Madise, University of Southampton

Research has shown that the decision to adopt a modern method of contraception is closely related to a number of individual and household demographic and socio-economic factors, although substantial variation in contraceptive use often exists even when controlling for these factors. Previous research on contraceptive use has focussed on the role of such individual and household factors, and has largely ignored the influence of the social and physical environment in which the individual lives. The application of multilevel modelling techniques to contraceptive use has facilitated the identification of areal level random effects. The presence of significant random effects indicates that contraceptive use varies across spatial units, and that the factors available in large-scale social surveys are often not adequate in explaining all the variation in outcomes between units. The unexplained heterogeneity in contraceptive use has previously been accounted for through regional variations in cultural, linguistic and behavioural practices that influence health seeking behaviour or the location of health services. There has been little in-depth examination of these factors for explaining variations in modern contraceptive use, especially in sub-Saharan Africa. This is primarily due to the relatively recent uptake in modern methods in this region. Consequently, there is a need to examine a range of individual, household and contextual factors that may create areal variations in contraceptive use in this setting. This paper aims to identify areal variations in contraceptive use in three East African countries, and to examine patterns in such variation across and between the countries. The main sources of contraceptive use data for developing countries are the Demographic and Health Surveys (DHS). Recently these surveys, especially in sub-Saharan Africa, have collected a global positional system (GPS) locator for the primary sampling units (PSUs) in the survey. This allows for maps of the modelling results to be plotted in order to better inform researchers and local policymakers, by highlighting those areas with unusually high or low contraceptive use. Using this new approach, the contextual characteristics of these areas can then be investigated as potential determinants of areal variations in contraceptive use. The objectives of this paper are: 1) To quantify, using DHS data, the determinants of the use of modern contraception in three East African countries. 2) To identify communities of unusually high or low contraceptive use within each of these countries after controlling for a range of bio-demographic and socio-economic variables. 3) To map the outlying communities within each country and to examine the variations in contraceptive use both within and across the three countries. Kenya, Malawi and Tanzania were selected for this study as the GPS location identifiers were collected in these countries. It was also of interest to choose neighbouring countries to allow the identification of patterns in areal variation that may transcend political boundaries. Findings from multilevel models will be presented that estimate the determinants of contraceptive use in each of the three countries. The models are estimated separately for each of the three countries, to allow a comparison of determinants between countries. The models take the form of three-level logistic models, with women (level one) nested within primary sampling units (level two) that are nested within districts (level three). The use of multilevel modelling allows for this hierarchical data structure. The PSU and district level residuals are then used to identify areas in which women have unusually high or low odds of contraceptive use after controlling for a range of bio-demographic and socio-economic factors. The outlying communities identified in the modelling of contraceptive use will then be plotted using the ArcView package. This research provides a unique opportunity to advance the methodology used to explain areal variations in contraceptive use in developing countries through combining multilevel modelling and Geographic Information Systems. It also facilitates a comparison of the determinants of modern contraceptive use across three East African countries.