Methods: We used data from the Demographic and Health Survey Nepal 2006 (N=10,793), the national study of population and health containing information on PSU level geographic variables. We used multilevel logistic regression to obtain probabilities of child deaths among indigenous women across 260 geographic communities. We used these probabilities to construct a GIS map to graphically represent changes in, and the statistical significance of, probability of child deaths conditional upon mothers' characteristics, household characteristics, and geographic characteristics.
Results: Controlling for demographics and other covariates, survey participants in our sample displayed statistically significant variation in probabilities of child deaths, ranging between 0% (in PSU 2705) and 64% (in PSU 6802). Controlling for individual level poverty and other characteristics did not bridge the gap that existed between different geographic communities.
Implications: People in Nepal display over 60-folds variations in probabilities of health problems. Such variation can be explained neither by individual characteristics, nor by local development indicators. Attention to the intrinsic development practice is necessary to determine if these variations are reflective of the institutional characteristics of the communities in which the vulnerable population reside.
Words: 279