Early Warning of Ross River Virus Epidemics

Abstract

BACKGROUND: Ross River virus disease is spread by mosquitoes, and an average of 5000 people are infected each year in Australia. It is one of the few infectious diseases for which climate-based early warning systems could be developed. The aim of this study was to test whether supplementing routinely collected climate data with mosquito surveillance data could increase the accuracy of disease prediction models. METHODS: We focused on a temperate region of Western Australia between July 1991 and June 1999. We developed "early" and "later" warning logistic regression models to test the sensitivity of data on climate (tide height, rainfall, sea surface temperature) and mosquito counts for predicting epidemics of disease. RESULTS: Climate data on their own were moderately sensitive (64%) for predicting epidemics during the early warning period. Addition of mosquito surveillance data increased the sensitivity of the early warning model to 90%. The later warning model had a sensitivity of 85%. CONCLUSIONS: We found that climate data are inexpensive and easy to collect and allow the prediction of Ross River virus disease epidemics within the time necessary to improve the effectiveness of public health responses. Mosquito surveillance data provide a more expensive early warning but add substantial predictive value.