Department of Epidemiology & Biostatistics Special Seminar presented by
Raul Cruz-Cano, PhD
Associate Research Professor
Epidemiology & Biostatistics
University of Maryland, College Park
Ongoing climate variability and change is having a considerable impact on the burden of climate sensitive bacterial infectious diseases across the globe. Warming temperature promotes bacterial growth while extreme precipitation enhances the spread of these pathogens, increasing fecal-oral route of exposures, particularly in resource limited settings. Given extreme heat and precipitation events are projected to continue increasing in frequency and intensity, there is an urgent need to develop meaningful early warning system to guide public health decision-making activities at present. Here we show that shallow time-series neural network models can be successfully applied with environmental data to infer diarrheal disease risk with sub-seasonal and seasonal lead time. Using historical diarrheal disease data from Nepal, Vietnam and Taiwan, we show that such approach works well in complex settings. Our approach builds on the existing framework by adding sub-seasonal to seasonal lead time for the disease outlook. As expected, the performance of the network varies based on the availability of the disease surveillance and meteorological data as well as their spatiotemporal resolution. The early warning system for diarrheal disease is well suited to provide categorical probability (Low, Medium, High) of disease burden ahead of time that can help guide public health decision-making at a seasonal level.