Twitter and Facebook as a source of pharmacovigilance data
Social media use has substantially expanded over the last decade, with 321 million monthly users and 126 million daily users on Twitter. This year, Twitter made a few tweaks to its own user metric: instead of daily users, they count monetizable daily active users. This figure might not be directly comparable to other social networks such as Instagram (500 million daily users) and Snapchat (186 million daily users). Because Twitter cracked down on bots, fake profiles, and spam users, its total number of active monthly users seems to be going down. The positive news is that the network generally consists of real users whose identity is verifiable .
Facebook’s growth continues to grow despite all the privacy scandals and ongoing investigations. In Q1 2019, Facebook reported 2.37 billion monthly active users, 55 million more than in the previous quarter. The number of users in the U.S. and Canada remains stable (234 million in Q1 2017 vs. 243 million in Q1 2019) and Europe follows the similar trend (354 million in Q1 2017 vs. 384 million in Q1 2019). Considering the total population of the U.S. (327 million in 2018), Canada (37 million) and Europe (743 million), nearly everyone in North America and half of the population of Europe is on Facebook already. The highest growth was observed in Asia-Pacific (predominantly India) and the Rest of the World. In Q1 2019, 186 million North Americans (U.S. and Canada) and 286 million Europeans were using their Facebook account on a daily basis. Despite expected fines stemming from privacy violations, Facebook’s financial performance remains solid .
Social media holds considerable promise as a source of pharmacovigilance data. From a pharmacovigilance perspective, the key areas that need to be satisfactorily addressed correspond with the minimum reporting criteria for individual case safety reports. To be valid, a report must contain at least an identifiable reporter, identifiable patient, identifiable product, and a suspected adverse event.
The requirement of identifiability and the practicability of follow-up remains the main challenge. Bots, fake profiles and spammers are confounding factors in post-market safety surveillance that only generate noise. Because pharmacovigilance surveillance is based on reporting information to national regulatory bodies, the location of the user defines who and how will be processing the case report.
While in most instances the user who reports an event is also the patient, scenarios that people describe experiences of others also occur. In the case of an indirect relationship, the report typically would not meet the validity criteria for an identifiable patient.
The granularity of the data regarding a suspect product or their combination and the context of its use are critical to the quality and usability of the report. Experiences regarding drug-drug interactions, off-label use, misuse, abuse, quality problems, unanticipated benefits, patterns of use including self-medication or diversion are all valuable in pharmacovigilance context.
Suspected adverse event
Recognizing an event in the sea of creative idiomatic expressions, terms not mentioned in medical lexicons, poor grammar and spelling errors requires a machine learning component that needs resources for appropriate training through human-generated annotation. Moreover, an experience relating to a drug needs to be distinguishing from indications and concomitant conditions. This is an example of the Name Entity Recognition (NER) problem in information retrieval. Addressing the diversity of ADR terms can be solved by using a variety of dictionaries, such as the FDA Adverse Event Reporting System (FAERS), Coding Symbols for a Thesaurus of Adverse Reaction Terms (COSTART), Collaborative Consumer Health Vocabulary (CHV), MedEffect, Unified Medical Language System (UMLS), Medical Dictionary for Regulatory Authorities (MedDRA) and Side Effect Resource (SIDER). 
With minimum case validity criteria met, we can look forward to the next challenge: human annotation in supervised machine learning algorithms.
 Kastrenakes, J. (2019). Twitter keeps losing monthly users, so it’s going to stop sharing how many. Retrieved from https://www.theverge.com/2019/2/7/18213567/twitter-to-stop-sharing-mau-as-users-decline-q4-2018-earnings
 Hutchinson, A. (2019). Facebook Reaches 2.38 Billion Users, Beats Revenue Estimates in Latest Update. Retrieved from https://www.socialmediatoday.com/news/facebook-reaches-238-billion-users-beats-revenue-estimates-in-latest-upda/553403/
 Sloane, R., Osanlou, O., Lewis, D., Bollegala, D., Maskell, S., & Pirmohamed, M. (2015). Social media and pharmacovigilance: A review of the opportunities and challenges. British Journal Of Clinical Pharmacology, 80(4), 910-920. doi: 10.1111/bcp.12717