Identification and Validation of Real-Time Health Events through Social Media

Twitter, the popular microblogging platform, has more than five hundred million registered users (Tweeters). These Tweeters generate a large amount of information every day. A big challenge and opportunity that is explored in this paper is to use this information to analyse health events -- ideally in real-time. Such real time information is essential for outbreaks of disease and identifying where and who might be affected. In this context however it is essential to verify that the information is accurate and can be compared with other data sources. This paper presents a methodology and infrastructure delivering such capabilities. Unlike other approaches that have been on a small scale, this work exploits large-scale Cloud facilities and much larger collections of data. Specifically, we collected and analysed over 46 million tweets from the three most populated cities in Australia (Sydney, Melbourne and Brisbane) to find patterns related to health events. Five diseases were explored: ebola, dengue fever, flu, H1N1 and hayfever, however the platform can be used for other disease areas. We compared and validated the results with Google Trends data as well as data from the Australian Institute of Health and Welfare. We identified a high and measurable correlation between our data and these other sources. Building on these quantifiable degrees of accuracy, we suggest that social media can indeed be a key approach to alert authorities and the population at large of health disease events, e.g. pandemics, and allow them to track disease spread. At present no such infrastructure or capability exists.