Meta-Analysis of Model Specifications Assessing Harmful Algal Blooms as a Risk Factor for Amyotrophic Lateral Sclerosis in Northern New England

• Activity Number: 625 - Environmental Epidemiology and Spatial Statistics
• Type: Contributed
• Date/Time: Thursday, August 3, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Beth Ziniti* and Ernst Linder and Nathan Torbick and Angeline Andrew and Elijah W. Stommel
• Companies: Applied Geosolutions and Department of Mathematics and Statistics, University of New Hampshire and Applied Geosolutions and Department of Neurology, Dartmouth Medical School and Department of Neurology, Dartmouth Medical School
• Keywords: spatial epidemiology ; change of support ; Bayesian Inference ; INLA ; remote sensing ; water quality

Abstract

In this study, we use a spatial epidemiological approach to evaluate satellite derived harmful algal bloom (HAB) exposure at various proximity scales as a risk factor for amyotrophic lateral sclerosis (ALS), or Lou Gehrig's disease. Since significance of associations in geographic correlation studies depends on several modeling choices that are fixed at the beginning of an analysis, we use a meta-analysis that compares Deviance Information Criterion (DIC) values and HAB p-values to assess these model choices. Multiple Poisson log-linear models are estimated in a Bayesian framework using Integrated Nested Laplace Approximation (INLA) that vary each of the following components: the use of spatial random effects, the background population dataset, the size of the geographic modeling units for case residence locations and the proximity scales of HAB exposure. The most significant differences in DIC values occurred between models that varied the size of geographic modeling units, although there was little difference between DIC values of different HAB proximity scales. The broader outcomes of this work contribute to an interdisciplinary effort to understand neurological impacts of HAB.