Dengue prediction by the web: Tweets are a useful tool for estimating and forecasting Dengue at country and city level

Background: Infectious diseases are a leading threat to public health. Accurate and timely monitoring of disease risk and progress can reduce their impact. Mentioning a disease in social networks is correlated with physician visits by patients, and can be used to estimate disease activity. Dengue is the fastest growing mosquito-borne viral disease, with an estimated annual incidence of 390 million infections, of which 96 million manifest clinically. Dengue burden is likely to increase in the future owing to trends toward increased urbanization, scarce water supplies and, possibly, environmental change. The epidemiological dynamic of Dengue is complex and difficult to predict, partly due to costly and slow surveillance systems. Methodology / Principal findings: In this study, we aimed to quantitatively assess the usefulness of data acquired by Twitter for the early detection and monitoring of Dengue epidemics, both at country and city level at a weekly basis. Here, we evaluated and demonstrated the potential of tweets modeling for Dengue estimation and forecast, in comparison with other available web-based data, Google Trends and Wikipedia access logs. Also, we studied the factors that might influence the goodness-of-fit of the model. We built a simple model based on tweets that was able to ‘nowcast’, i.e. estimate disease numbers in the same week, but also ‘forecast’ disease in future weeks. At the country level, tweets are strongly associated with Dengue cases, and can estimate present and future Dengue cases until 8 weeks in advance. At city level, tweets are also useful for estimating Dengue activity. Our model can be applied successfully to small and less developed cities, suggesting a robust construction, even though it may be influenced by the incidence of the disease, the activity of Twitter locally, and social factors, including human development index and internet access. Conclusions: Tweets association with Dengue cases is valuable to assist traditional Dengue surveillance at real-time and low-cost. Tweets are able to successfully nowcast, i.e. estimate Dengue in the present week, but also forecast, i.e. predict Dengue at until 8 weeks in the future, both at country and city level with high estimation capacity. Author summary: Dengue is a fast-growing mosquito-borne viral disease, with an estimated annual incidence of 390 million infections, of which 96 million manifest clinically. Dengue burden is likely to increase in the future. Mentioning a disease in social networks is correlated with physician visits by patients, and can be used to estimate disease activity. Traditional, biologically-focused monitoring techniques, based on laboratory diagnostics, are accurate but costly and slow. Alternative approaches for surveillance aim to capture health-seeking behavior at earlier stages of disease progression, specially capturing the asymptomatic and mild clinic manifestation population who do not seek medical care formally. Twitter data have potential application for Dengue surveillance, improving the estimation and prediction of the disease, in space and time, being a valuable and low-cost addition to assist traditional surveillance. We show that tweets are strongly associated with Dengue cases. Tweets are a useful tool for estimating and forecasting Dengue cases until 8 weeks in the future, both at country and city level, even in less developed areas.