Expanding population-based survey indicators through spatial analysis

The Demographic and Health Surveys (DHS) provide a wealth of health information that governments use to inform programming. In most countries, DHS data can not be disaggregated with any statistical significance below a regional/provincial level due to sampling limitations. By combining GIS analysis with traditional data analysis, we can determine whether an indicator is uniformly distributed throughout a region or whether it is concentrated.

The weighted geographic mean center (WGMC) for clusters and an indicator within a region is derived and graphically plotted. A paired t-test of the longitude and latitude axes is used to validate the significance of the difference in mean distance between the cluster locations and two weighted mean points.

Using data from the 2008 Ghana DHS, the WGMC of women in union, ages 15-49, was generated per region, as well as the WGMC of unmet need for family planning in that population. Results of the paired t-test indicate that in Ashanti, unmet need is focused in the north-east of the region (p≤0.01 and p≤0.001) while in Volta and Greater Accra, it is concentrated towards the north of the region (p≤0.05). In Upper West unmet need is concentrated in the east of the region (p≤0.05). In remaining regions, unmet need is likely to be uniformly distributed.

Although this abstract focuses on unmet need in Ghana, this approach can be used globally for any population-based indicator. Regional and lower-level program managers can use it to better target their programming and focus interventions among communities with the greatest need.