Fine-temporal forecasting of outbreak probability and severity: Ross River virus in Western Australia

Abstract

Health warnings of mosquito-borne disease risk require forecasts that are accurate at fine-temporal resolutions (weekly scales); however, most forecasting is coarse (monthly). We use environmental and Ross River virus (RRV) surveillance to predict weekly outbreak probabilities and incidence spanning tropical, semi-arid, and Mediterranean regions of Western Australia (1991-2014). Hurdle and linear models were used to predict outbreak probabilities and incidence respectively, using time-lagged environmental variables. Forecast accuracy was assessed by model fit and cross-validation. Residual RRV notification data were also examined against mitigation expenditure for one site, Mandurah 2007-2014. Models were predictive of RRV activity, except at one site (Capel). Minimum temperature was an important predictor of RRV outbreaks and incidence at all predicted sites. Precipitation was more likely to cause outbreaks and greater incidence among tropical and semi-arid sites. While variable, mitigation expenditure coincided positively with increased RRV incidence (r 2 = 0·21). Our research demonstrates capacity to accurately predict mosquito-borne disease outbreaks and incidence at fine-temporal resolutions. We apply our findings, developing a user-friendly tool enabling managers to easily adopt this research to forecast region-specific RRV outbreaks and incidence. Approaches here may be of value to fine-scale forecasting of RRV in other areas of Australia, and other mosquito-borne diseases.

Temperature explains broad patterns of Ross River virus transmission

“…The model correctly predicts that RRV disease is year-round endemic in tropical, northern Australia with little seasonal variation due to temperature, and seasonally epidemic in temperate, southern Australia. These results provide a mechanistic explanation for idiosyncrasies in RRV temperature responses observed in previous studies ( Hu et al, 2004 ; Gatton et al, 2005 ; Bi et al, 2009 ; Werner et al, 2012 ; Koolhof et al, 2017 ). A population-weighted version of the model (assuming a two-month lag between temperature and human cases based on mosquito and disease development times) also accurately predicts the seasonality of human cases nationally.…”

“…Moreover, compared to vector-borne diseases in lower-income settings, RRV case diagnosis and reporting are more accurate and consistent, and variation in socioeconomic conditions (and therefore housing and vector control efforts) at regional and continental scales is relatively low. Previous work has shown that in some settings temperature predicts RRV cases ( Gatton et al, 2005 ; Bi et al, 2009 ; Werner et al, 2012 ; Koolhof et al, 2017 ), while in others it does not ( Hu et al, 2004 ; Gatton et al, 2005 ). Understanding RRV transmission ecology is critical because the virus is a candidate for emergence worldwide ( Flies et al, 2018 ), and has caused explosive epidemics where it has emerged in the past (infecting over 500,000 people in a 1979–80 epidemic in Fiji) ( Klapsing et al, 2005 ).…”

“…First, how the impact of temperature on transmission differs across diseases, via what mechanisms, and the types of data needed to characterize these differences all remain uncertain. Second, the impacts of temperature on transmission can appear idiosyncratic—varying in both magnitude and direction—across locations and studies ( Gatton et al, 2005 ; Jacups et al, 2008a ; Stewart-Ibarra and Lowe, 2013 ; Peña-García et al, 2017 ; Koolhof et al, 2017 ). Although inferring causality from field observations and statistical approaches alone remains challenging, nonlinear thermal biology may mechanistically explain this variation.…”

“…As the climate changes, filling these gaps becomes increasingly important for predicting geographic, seasonal, and interannual variation in transmission of mosquito-borne pathogens. Here, we address these gaps by building a model for temperature-dependent transmission of Ross River virus (RRV), the most important mosquito-borne disease in Australia (1500–9500 human cases per year) ( Koolhof et al, 2017 ), and potentially at risk of emerging worldwide ( Flies et al, 2018 ).…”

“…Understanding RRV transmission ecology is critical because the virus is a candidate for emergence worldwide ( Flies et al, 2018 ), and has caused explosive epidemics where it has emerged in the past (infecting over 500,000 people in a 1979–80 epidemic in Fiji) ( Klapsing et al, 2005 ). RRV is a significant public health burden because infection causes joint pain that can become chronic and cause disability ( Harley et al, 2001 ; Koolhof et al, 2017 ). A mechanistic model for temperature-dependent transmission could help explain these disparate results and predict potential expansion.…”