Ross River virus, climate and environment in the Darling Downs, Queensland, Australia

Abstract

Ross River virus (RRV) disease is the most significant mosquito-borne disease in Australia with approximately 5,000 cases notified each year. The disease is characterised by joint pain and lethargy which can persist for several months and that places a significant burden on individual patients, the health care system and economy. Given the lack of effective treatment or vaccine for the disease, it is important to develop early warning systems and predictive RRV models that can then inform population health initiatives and RRV prevention.

The complex ecology of RRV makes it a complicated disease to predict. The virus is unique in its capacity to exist across all environments and climates of Australia based on the number of vector and reservoir host species involved in transmission. Therefore, predictive models need to be developed at local scales in order to produce accurate and useful results. Moreover, climate has been isolated as a factor which influences animals, humans and the environment and therefore may be useful to RRV predictive models. This thesis aimed to develop a predictive model for RRV disease for the inland region of the Darling Downs Hospital and Health Services (HHS) area in Queensland, Australia.

A negative binomial regression model was developed using lagged climate data and RRV notification data from the period of July 2001 to June 2014. Variables were selected using Spearman’s rank correlation and the model was developed using backward elimination. This model was then evaluated through the comparison of observed and model predicted RRV case numbers for the period of July 2014 to June 2019 using Pearson’s correlation and sensitivity and specificity measures.

The final model used vapour pressure, solar radiation, relative humidity and the Southern Oscillation Index as predictor variables. The model was moderately effective at predicting RRV case numbers (Pearson’s correlation = 0.420) and RRV outbreaks (accuracy = 55%, sensitivity = 47%, specificity = 58%).

Ultimately, this thesis found that climate is an important component of the RRV disease ecology cycle, but climate variables alone cannot accurately predict RRV outbreaks in the Darling Downs HHS, Queensland. This finding supports the One Health approach to population health policy and research which emphasises the interconnectedness between humans, animals and the environment. Therefore, there is a need to consider climate variables as well as socioeconomic and political factors in the development of predictive models for RRV.