Predicting Opioid Epidemic by Using Twitter Data

• Published in: Bioinformatics Research and Applications Vol. 10847; pp. 314 - 318
• Main Authors: Wu, Yubao, Skums, Pavel, Zelikovsky, Alex, Rendon, David Campo, Liao, Xueting
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• ISBN: 9783319949673; 3319949675
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-94968-0_30

Opioid crisis was declared as a public health emergency in 2017 by the President of USA. According to the Centers for Disease Control and Prevention, more than 91 Americans die every day from an opioid overdose. Nearly $4B is provided to address the opioid epidemic in the 2018 spending bill and help fulfill the President’s Opioid Initiative. How to monitor and predict the opioid epidemic accurately and in real time? The traditional methods mainly use the hospital data and usually have a lag of several years. Even though they are accurate, the long lag period prevents us from monitoring and predicting the epidemic in real time. We observe that people discuss things related to the epidemic a lot in social media platforms. These user behavior data collected from social media platforms can potentially help us monitor and predict the epidemic in real time. In this paper, we study how to use Twitter to monitor the epidemic. We collect the historic tweets containing the set of keywords related to the epidemic. We count the frequency of the tweets posted at each month and each state. We compare the frequency values with the real-world death rates at each month and each state. We identify high correlation between tweet frequency values and real-world death rates. The statistical significance demonstrates that the Twitter data can be used for predicting the death rate and epidemic in future.