In addition to asking questions around the value and validity of using social media monitoring for pharmacovigilance, we also often get questions around our thoughts on whether social media data are inherently biased. When I brought this question to the Booz Allen Epidemico data team, they educated me on the three dimensions of bias that come from different drug safety datasets: Patient reported outcomes, seriousness, and completeness. The graphic below shows how those three dimensions differ across social media, clinical trials, spontaneous reports, and EHR records.
The next question is if the data skew towards a certain demographic with social media. Looking across Facebook, Twitter, and Instagram, the answer seems to be, "it depends".
So, like most things data related, it's all about knowing limitations and understanding the best way to overcome those limitations. See our primer on using social media monitoring for pharmacovigilance. One of the ways that our clients are doing that is by ensuring that social monitoring is just one source of the truth. By coupling social media data with clinical, spontaneous, and other real-world datasets like EHR's, using the Evidex platform, they are able to see all versions of the truth and answer virtually any safety question, quickly. Feel free to reach out to discuss further!