Balancing data limitations and opportunities with geographically linked datasets

Purpose: To demonstrate the benefits and limitations of linking independent datasets using geographic methods. Data: Individual- and household-level socio-demographic characteristics and health data from the 2005 Rwanda Demographic and Health Survey (RDHS) and facility data from Rwanda's 2005 Service Availability Mapping (SAM) census are used. Methods: RDHS clusters and SAM facilities are geographically linked in two simulations. A nationally representative sample of health facilities was selected from the full census and linked to household clusters based on proximity within the same administrative unit. The second simulation linked the census of government facilities to clusters by distance, irrespective of administrative borders. Spatial analytic techniques identify clusters that deviate from the national mean for select reproductive health indicators. Facility data describe the health service environment for these communities. Mapping the clusters alongside their proximate health facilities provides service context for the observed outcomes. Results: 436 facilities were mapped by SAM; 270 (62%) were selected for simulation one. The effects of facility sampling on data linking will be shown. Clusters deviating from the mean will be mapped for a selection of reproductive health outcomes and the linking will generate hypotheses about the influences of services on health outcomes. Additional analyses between individual outcomes and facility services will be explored. Implications: Linking independent datasets supports evaluation of service influences on health outcomes. It's important that linked data is used in a manner consistent with the component datasets' limitations. Understanding the added-value of spatial analyses to address limitations will inform data use and sampling strategies.