Predictive forecasting of Ross River virus outbreaks and their relative severity in Western Australia

Environmental determinants of Ross River virus (RRV) have long been studied for their role in outbreak epidemics. Studies into environmental drivers in RRV outbreaks provide a potentially critical tool in developing early warning detection systems capable of bolstering mitigation strategies. Here, we expand on previous studies exploring the development of forecasting and severity (incidences) models in five areas in Western Australia which experience high attack rates and outbreak epidemics of RRV. Our sites included three Northern areas; Derby, Broome, and Port Hedland and two Southern areas; Mandurah, and Capel. Hurdle and linear regression models were constructed using weekly observations of the maximum and minimum temperature and tidal heights, and total rainfall precipitation to predict both outbreak probability and severity respectively. Results yielded good predictability in both outbreak probability and severity across all sites with the exception of Capel owing to continuous RRV notifications throughout the study period respective to its population size and regardless of seasonal predictors. The hurdle models provided a robust way to effectively model RRV transmission where mosquito surveillance is unable to be conducted along coastal areas. This has significant benefits to cost reduction in modeling and forecasting RRV transmission for site-specific areas. These results suggest that using hurdle models for RRV forecasting may provide a relatively cheap alternative to public health officers in integrating their use into directing mitigation at appropriate time points to improve mitigation effectiveness in Western Australia.