Spatio-Temporal Trends of Diarrheal Mortality of Children in Association with Hydrographic Regions of Brazil
Leyk, Stefan, Phillips, Thomas P., Smith, Jeremy M., Nuckols, John R., Cartography and Geographic Information Science
The importance of an improved understanding of environmental factors associated with enteric diseases such as diarrhea has been widely recognized and has particular relevance in spatial epidemiology (Torok et al. 1997; Kelly-Hope et al. 2007). There is strong demand to investigate dynamics of exposure to and transmission of water-borne pathogens, and to establish improved surveillance systems that can lead to effective preventive measures of such diseases (Eyles et al. 2002; Rushton 2003; Pande et al. 2008; Chaikaew et al. 2009). Such research is of particular priority in less developed regions where the burden from enteric diseases is still very high (Fewtrell et al. 2005).
An important factor in the analysis of environmental risk factors is the geographic extent and framework of the unit of analysis. Recent efforts analyzed spatio-temporal patterns of diarrhea by exploring the influence of regional dynamics of risk factors such as climate and socio-economic conditions (Kelly-Hope et al. 2008; Jepsen et al. 2004). Often the unit of analysis for such studies are administrative reporting units (states or larger) used in disease reporting, resulting in highly aggregated outcomes with limited representation of the underlying environmental phenomenon that might be more realistically reflected by analytical units defined by natural barriers, ecological systems, and other important factors in pathogen occurrence, exposure, and transmission (Curriero et al. 2001). Such disconnect can result in a higher likelihood of error due to ecological fallacy, reduce heterogeneity of observed data, and thus reduce statistical power needed to discern association between the risk factors and incidence of disease (Peters et al. 2004; Wakefield and Shaddick 2005; Beale et al. 2008).
In this article we demonstrate a geospatial framework for identifying spatio-temporal patterns of mortality peak timing from pediatric diarrhea based on relative location of the disease reporting unit in the hydrologic regime of major river basins in Brazil. Such an approach could be useful for a better understanding of the associations between waterborne diseases and environmental processes in general, thus allowing for derivation of predictive models for exposure and transmission within a natural, rather than geopolitical, spatial unit of analysis.
The geographic extent of our study area is the geopolitical boundary of Brazil. We obtained spatially registered GIS data layers of watershed boundaries and stream hydrography for eight (8) Hydrographic Regions from Agencia Nacional de Aguas (ANA; http://www2.ana.gov.br/). We used 1 km spatial resolution topographic data from the Shuttle Radar Topography Mission (SRTM; http://www2.jpl. nasa.gov/srtm/) to carry out hydrologic modeling.
We obtained Brazilian mortality data classified as intestinal infectious disease (ICD-9 codes) and selected all deaths of children younger than 5 years of age for the period 1979-1989 from the Ministerio da Saude do Brasil (http://www2.datasus. gov.br/ DATASUS/index.php?area=02). We used this time period rather than more current data because mortality rates and thus geographic heterogeneity of number of deaths were higher, allowing a more robust data set for the purpose of our study. In the dataset, monthly mortality is reported by Municipality, which is a census reporting unit of the Brazilian government with an approximate median area of 420 Km2. The number of Municipalities in Brazil changed from 3991 in 1979 to 4491 in 1989. Because of large proportions of missing data and changes of political boundaries for Municipalities over our study period, we aggregated the number of deaths per month to Census Micro Regions (CMRs) each of which contain multiple Municipalities. The number of CMRs increased from 542 in 1979 to 558 in 1989 during the study period, with a median area of 5480 Km2 in 1989. We applied screening criteria for accepting CMR-level MPT data: the presence of only one maximum value, average maximum deaths greater than 3 or if less than 3, at least 5 years of recorded data, resulting in a final dataset of 507 CMRs for use in our study, or 91% of all CMRs that reported in 1989. Due to the criteria of at least 5 years of data, newly created CMR regions could be included only if they were established no later than 1985, and had annual deaths reported for each year or had higher numbers of deaths in less than 5 years. We obtained spatial layers of polygons from the Instituto Brasileiro de Geografia e Estatistica (IBGE; http://www.ibge.gov. br/) representing administrative boundaries of CMRs and Municipalities, as well as point data (latitude, longitude) representing Municipal seats locations (cities).
In a first step we calculated the mortality peak timing variable (MPT) for each Census Micro Region (CMR) based on aggregated monthly Municipal mortality data. MPT is defined as the average digital month on a continuous time scale (between 1.0 for May 1st and 12.99 for April 30th) when maximum mortality occurred in each CMR over the study period (1979-1989). Next, we tested the mortality peak time variable for global spatial autocorrelation and local spatial clustering in order to justify the use of geostatistical models. In a third step we assigned the CMR-based MPT to the point location of the Municipality with the highest number of mean annual deaths, termed Maximum Mortality Municipality (MMM), within each CMR. These point locations were used as input to two different geostatistical models in order to create validated continuous surfaces of mortality peak timing across each Hydrographic Region. Finally we tested for trends of mortality peak timing along the major stream channels of the river network within each Hydrographic Region. We validated our modeled trends by comparing them to inspected trends based on MPT values of discrete Municipality locations in close proximity to major streams.
Extracting Mortality Peak Timing
We extracted the number of deaths occurring in Municipalities for each CMR of Brazil (Figure 1) for each month of the years 1979-89. We normalized these numbers for each year to proportional values on a continuous scale in the range [0,1] such that the annual maximum number of deaths was equal to unity, and the proportional number in each month a value between 0 and 1. These proportional values represent measures of similarity between numbers of deaths in each month and the number of deaths in the peak month. We …
Questia, a part of Gale, Cengage Learning. www.questia.com
Publication information: Article title: Spatio-Temporal Trends of Diarrheal Mortality of Children in Association with Hydrographic Regions of Brazil. Contributors: Leyk, Stefan - Author, Phillips, Thomas P. - Author, Smith, Jeremy M. - Author, Nuckols, John R. - Author. Journal title: Cartography and Geographic Information Science. Volume: 38. Issue: 2 Publication date: April 2011. Page number: 223+. © 2008 American Congress on Surveying & Mapping. COPYRIGHT 2011 Gale Group.
This material is protected by copyright and, with the exception of fair use, may not be further copied, distributed or transmitted in any form or by any means.