Performer: Epidemico, LLC
Principal Investigators: John Brownstein, Nabarun Dasgupta
Project Duration: 9/30/12 - 9/29/15
Regulatory Science Challenge
When a drug is being developed, clinical trials test safety and efficacy, but those patients enrolled in the trial may be less diverse than those who eventually use the drug following approval. Sometimes unforeseen risks can surface after a medical product is on the market and is used by many more people than were in the clinical trials or used in new populations. Surveilling the safety of FDA-regulated products once they reach the market to obtain a more complete understanding of a drug’s side effects is part of the Agency’s mission to protect and promote public health.
However, FDA is always looking to identify new sources of data to strengthen its surveillance activities. Acquiring additional safety information from social media on sites like Twitter, Facebook, and Reddit (stripped of identifying information before being used for scientific analysis) may result in a higher number of reports for each event. This would enable FDA to link adverse events reported in social media to reports from other sources – such as drug manufacturers and physicians. This, in turn, could lead to changes to make medication use safer for all.
FDA collaborated with the health informatics company Epidemico to develop the digital data-mining platform MedWatcher Social. MedWatcher Social is a digital listening platform designed to automatically collect and categorize social media information about drug safety events and experiences with medical products. It uses Amazon Web Services cloud technology, which takes in publicly available posts from Twitter and Facebook. First filtered to focus on those that refer to both medical products and adverse events, the data are mapped to standard medical product and adverse event dictionaries, where they are analyzed to identify potential safety adverse events.
It addresses the problem of adverse event underreporting by increasing the amount of safety information available to FDA after a drug has been launched.
MedWatcher Social uses innovative technology – specifically, natural language processing and machine learning algorithms (i.e. artificial intelligence) – to mine public social media data for descriptions of adverse events. First, public posts are collected through social media sites using medical product search terms. Next, the posts are processed to remove duplicates and personally identifiable information like first names and phone numbers. The posts are then processed again by a robot that has been trained to recognize discussions of adverse events in social media, and then processed once more to translate slang terms into a standard set of vocabulary that is used by FDA and pharmaceutical companies to record side effects (the Medical Dictionary for Regulatory Activities, or “MedDRA”). All posts that are classified as adverse event descriptions are then visualized in an interactive dashboard .
To create a social media data-mining tool to help detect negative side effects related to the use of FDA-regulated medical products.
Provides data in near real-time and enable earlier detection of drug safety signals, greatly enhancing the power of the FDA Adverse Event Reporting System (FARES).
Currently, social media analysis has specific limitations in monitoring safety drug concerns.