Chary M, Genes N, Giraud-Carrier C, Hanson C, Nelson LS, Manini AF. 2017. Epidemiology from tweets: estimating misuse of prescription opioids in the USA from social media. Journal of Medical Toxicology. doi: 10.1007/s13181-017-0625-5
Researchers compared use of language related to misuse of prescription opioids in tweets made between 2012, 2013, and 2014 and results from the National Surveys on Drug Usage and Health (NSDUH) for those years to evaluate Twitter activity as an epidemiological tool for estimating geographical variations in misuse of prescription opioids. Researchers collected two sets of tweets, one of tweets mentioning the use of prescription opioids and one of all tweets containing at least one alphanumeric character, using the Twitter Streaming API, which makes all public tweets anonymous and filterable for use in research. Researchers collected the two sets of tweets to understand whether the geographic distribution of tweets mentioning misuse of prescription opioids was unique compared to the geographic distribution of all tweets. Tweets were analyzed using machine-learning algorithms to quantify basic concepts included in tweets, evaluate the context of a tweet, and estimate users’ locations when geographical metadata were not available on tweets. Researchers collected 3.6 million tweets made over three years and about 70,000 people participated in the NSDUH. Geographical data was collected for 14% of tweets. Geographical estimates of misuse of prescription opioids based on tweets were significantly correlated with geographical data from the NSDUH. These results indicate that Twitter can be an informative epidemiological tool for estimating geographical differences in prescription opioid misuse.