Using Climate to Explain and Predict West Nile Virus Risk in Nebraska

Abstract

We used monthly precipitation and temperature data to give early warning of years with higher West Nile Virus (WNV) risk in Nebraska. We used generalized additive models with a negative binomial distribution and smoothing curves to identify combinations of extremes and timing that had the most influence, experimenting with all combinations of temperature and drought data, lagged by 12, 18, 24, 30, and 36 months. We fit models on data from 2002 through 2011, used Akaike's Information Criterion (AIC) to select the best-fitting model, and used 2012 as out-of-sample data for prediction, and repeated this process for each successive year, ending with fitting models on 2002-2017 data and using 2018 for out-of-sample prediction. We found that warm temperatures and a dry year preceded by a wet year were the strongest predictors of cases of WNV. Our models did significantly better than random chance and better than an annual persistence naïve model at predicting which counties would have cases. Exploring different scenarios, the model predicted that without drought, there would have been 26% fewer cases of WNV in Nebraska through 2018; without warm temperatures, 29% fewer; and with neither drought nor warmth, 45% fewer. This method for assessing the influence of different combinations of extremes at different time intervals is likely applicable to diseases other than West Nile, and to other annual outcome variables such as crop yield. Plain Language Summary We wanted to see whether we could predict years with higher risk of West Nile Virus in Nebraska using publicly available data on temperature, precipitation, human cases, and population. We used a type of model that lets the data speak for itself, identifying which intervals of time leading up to an infection season are most important. We found that a dry year following a wet year, often in combination with warm temperatures, increased the likelihood of infection. Drought accounted for about 26% of the number of cases from 2002 to 2018.

West Nile Virus in the State of Ceará, Northeast Brazil

“…Such water areas might be attractive for several bird species also, which might increase the bird–mosquito interaction [ 80 , 81 ]. In a study conducted in the USA, a model predicted that without drought nor warmth there would have been 43% fewer cases of WNV in 2018 [ 82 ]. Based on climate model predictions for climate change and potentially greater drought occurrence in the future, the frequency and relative risk of WNV outbreaks could increase [ 76 ].…”

Better null models for assessing predictive accuracy of disease models

“…The uniform null model performed quite poorly in this study. The Prior Year null (used in Smith et al 2020) and the Mean Value (used in Keyel et al 2019 in a non-probabilistic context) to be poor choices of null models for WNV. The AR1 model also performed poorly.…”

“…The results from the prior year are used to predict the current year (Smith et al 2020). This null has a weakness in that avian seroprevalence often increases following a widespread WNV outbreak, and high avian immunity decreases WNV activity (Kwan et al 2012).…”

A proposed framework for the development and qualitative evaluation of West Nile virus models and their application to local public health decision-making

“…This in-depth analysis would not have been possible with in a traditional review format. The models selected here are broadly representative of the WNV models that have been developed (e.g., statistical [ 20 – 24 ], data assimilation [ 25 , 26 ], mathematical trait–based [ 27 ], machine learning [ 28 – 30 ], threshold-based risk [ 31 – 33 ], and distributed lag approaches [ 34 – 36 ]). We also include a probabilistic historical null model in our comparison [ 37 ].…”

“… 2 R 2 pred : predictive R 2 , i.e., an R 2 calculated on data outside the sample, R s : Spearman correlation coefficient, AUC : area under the curve, Threshold-based accuracy : +/−25% of peak week, human cases, total infections over the season; +/−25% or 1 human case, RMSE : Root Mean Squared Error, CRPS : Continuous Ranked Probability Score. 3 Results for 2018 reported here, validation was also performed separately for 2012–2017, see [ 34 ] for details. 4 Three analyses presented: short-term: AUC = 0.856, annual made on July 5: AUC = 0.836, annual made on July 39: AUC = 0.855.…”