Abstract:
|
Warning systems for environmental exposure such as drought, air pollution, flooding, wildfile, and etc. are a high priority of federal and local agencies, but especially drought presents a complex mechanism and linkages between drought and human health outcomes are not fully understood. We use the data from the Mortality and Population Data System (MPDS) from 1980-2014, however, all death counts between a certain range were suppressed due to the National Center for Health Statistics (NCHS) privacy policy. To address this issue, we present a Bayesian censored zero-inflated negative binomial regression model with incorporating spatial and temporal variabilities to examine an overall impact of annual drought on human health risks (all-cause mortality). An efficient MCMC sampling algorithm is developed to carry out posterior inference. The proposed method is applied to public health research to examine drought-related health impact in Nebraska.
|