2013 Conference of the International Medical Geology Association (25–29 August 2013)

Paper No. 1
Presentation Time: 2:10 PM


MALONE, John B.1, NIETO, Prixia1, MISCHLER, Paula1, MARTINS, Moara1, MCCARROLL, Jennifer1, VOUNATSOU, Penelope2, SCHOLTE, Ronaldo2 and BAVIA, Maria E.3, (1)Pathobiological Sciences, Louisiana State University, Skip Bertman Drive, Veterinary Medicine Bldg, Baton Rouge, LA 70803, (2)Swiss TPH Institute, Basel, Switzerland, (3)Veterinary Epidemiology and Public Health, Universidade Federal da bahia, Escola de Medicina Veterinaria e Zootechnia, Salvador, Brazil, malone@vetmed.lsu.edu

Mapping and modeling methods were used to develop probability surfaces for 5 neglected tropical diseases (NTD) in Colombia, Brazil and Bolivia using climatic, environmental, census and poverty data. Health databases and socioeconomic data on Chagas disease, Leishmaniasis, Schistosomiasis, Leprosy, Lymphatic and Soil-Transmitted Helminths were obtained from government internet sites, literature review, local experts, and PAHO country representatives. Environmental/climatic data and poverty data were accessed via a password-protected FTP data portal in a uniform ASCII format (GCS, WGS 84, 1 km2) for analysis with health data by Maximum Entropy (Maxent) and ArcGIS software.

Ecological risk models were developed by Maxent at a disaggregation level (municipality scale) that allowed identification of the environmental variables and socioeconomic determinants (poverty) of health most associated with presence of the respective neglected tropical disease(s). Significant environmental and socioeconomic variables were identified by standard regression, correlation and variable inflation factor analysis. Maxent probability surfaces were generated to produce risk maps for each disease based on environmental and socioeconomic determinants separately, then as a ‘combined model’ of distribution in the environment weighted by indicators of poverty.

Health data that were initially point data at the municipality level were used to produce Maxent disease probability surfaces at a 1km spatial resolution based on the most significant environmental and social determinants of each NTD of interest. Maxent models based on Worldclim, Bioclim or annual composite MODIS NDVI/LST data alone or in combination revealed similar probability surfaces with acceptable to excellent AUC values. The value of socioeconomic variable analysis for mapping probability of disease was disappointing due to the heterogeneity of municipality level data. Results indicate Maxent extrapolation models can be used to identify hidden/unknown/blind areas possibly existing in each country for NTD’s in the absence of complete data on the selected diseases. A 4-day short course was developed for use by PAHO in training health workers on operational use of ArcGIS and Maxent models for surveillance and control of NTD’s in Latin America.

Previous Abstract | Next Abstract >>