Using Machine Learning and Web Forensics to Combat Digital Opioid Access
It is estimated that 90 Americans die daily by overdosing on opioids, resulting in 300,000 deaths since 2000. To combat the US opioid epidemic, the US Department of Health and Human Services (HHS) hosted a Code-a-Thon event aimed at leveraging technology and data-driven solutions.
ACCO expert Tim Mackey and his team participated in the event, developing a machine learning tool to accurately detect the marketing and sale of opioids on Twitter. Over the course of the event, they collected 213,041 tweets containing the keywords codeine, percocet, vicodin, oxycontin, oxycodone, fentanyl and hydrocodone. Of these, 692 tweets were identified as the illegal online marketing and sale of prescription opioids. Fifteen of the posts directed buyers to illegal online pharmacies, 11 to individual drug dealers and 7 by marketing affiliates. Many of these vendors also sold other controlled substances and illicit drugs.
The authors of the study were selected as finalists of the Code-a-Thon to further develop and implement their solutions.
Authors:
Tim K Mackay
Janani Kalyanam
Josh Klugman
Ella Kuzmenko
Rashmi Gupta