Abstract
Ross River virus (RRV), Barmah Forest virus (BFV), and dengue are three common mosquito-borne diseases in Australia that display notable seasonal patterns. Although all three diseases have been modeled on localized scales, no previous study has used harmonic models to compare seasonality of mosquito-borne diseases on a continent-wide scale. We fit Poisson harmonic regression models to surveillance data on RRV, BFV, and dengue (from 1993, 1995 and 1991, respectively, through 2015) incorporating seasonal, trend, and climate (temperature and rainfall) parameters. The models captured an average of 50–65% variability of the data. Disease incidence for all three diseases generally peaked in January or February, but peak timing was most variable for dengue. The most significant predictor parameters were trend and inter-annual periodicity for BFV, intra-annual periodicity for RRV, and trend for dengue. We found that a Temperature Suitability Index (TSI), designed to reclassify climate data relative to optimal conditions for vector establishment, could be applied to this context. Finally, we extrapolated our models to estimate the impact of a false-positive BFV epidemic in 2013. Creating these models and comparing variations in periodicities may provide insight into historical outbreaks as well as future patterns of mosquito-borne diseases.
Highlights
Seasonal patterns of disease, understood as periods of high and low disease incidence that are consistent across years, were recognized as early as 380 BCE1
The dengue models have larger standard deviations and wider ranges of peak timings than the Barmah Forest virus (BFV) and River virus (RRV) models which reveals that the peak timing for dengue was more variable than the peak timings for BFV and RRV
Using parametric harmonic regression models, we were able to estimate the peak timing of disease incidence between each location and disease, revealing that RRV and BFV peaked in January or February, with South Australia (SA) tending to peak earlier and QLD and New South Wales (NSW)/Australian Capital Territory (ACT) tending to peak later
Summary
We were interested in assessing the contribution of trend, seasonality, rainfall/temperature (with and without lag) and inter-annual periodicity to the temporal patterns This was accomplished by measuring the change in variability explained between the nested models (shown in Tables 1 and 2) by comparing the difference in the overall variability and regression-based variability expressed by the per cent variability explained. We estimated the expected monthly values using model parameters and weather data within the specified period of each outbreak. Using the temperature and rainfall data to estimate disease incidence according to the models’ equations, we extrapolated the models for BFV forwards to May 2015 to predict incidence during this period without the influence of potentially false case data.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
