| Abstract: | In this paper, we analyze trends in average annual peak timing values (MPT) of pediatric mortality attributed to diarrheal disease in Brazil for the period 1979-1989 using a novel approach for environmental health risk studies, that is using natural boundaries instead of politically derived boundaries to define the unit of analysis (UOA). We evaluate the approach at varying spatial scales: (1) Country-wide based on observed Municipal level mortality data aggregated to Census Micro Regions (CMR); (2) Country-wide based on a grid of 20 Km2 raster cells generated by geostatistical modeling of MPT values; (3) Within eight officially designated Hydrographic Regions of Brazil based on results from the geostatistical models, and (4) Along longitudinal “vectors” of 1 km raster cells defining the stream network (hydrologic regime) within each Hydrographic Region. At the country level, we found evidence of a trend west to east of increasing MPT over an annual cycle (May to April) using the CMR-level estimates. However, when we examined the model results at finer scales i.e., Hydrographic Regions, we discovered greater geographic heterogeneity in MPT across units. At the spatial scale of the stream network within the Hydrographic Regions, we observed consistent trends of increasing MPT from the source areas (upper watersheds) to downstream locations in some Hydrographic Regions, especially those composed of a single river basin. Here, trends were no longer predominantly east to west as at the country level, but oriented in the direction of flow of the major river draining the basin. Our study results indicate substantial spatial variation in peak timing of pediatric mortality attributed to diarrheal disease in Brazil over our study period. This could have important ramifications in studies concerning known or suspected risk factors with significant temporal variation over an annual cycle. We found the geographic orientation of trend in mortality peak timing to be highly dependent on the geographic extent and derivation of the UOA. We demonstrate that a UOA based on natural boundaries, e.g., stream segments or watershed boundaries can result in more consistent and robust prediction of trends in mortality peak timing attributed to diarrhea. |
