Detecting Adverse Drug Reaction (ADR) Mentions from Social Media

Abstract

Adverse Drug Reaction (ADR) can be described as undesired consequences
resulting from consumption of medical prescribed drugs. Such reactions are often missed in
clinical trials and are experienced by real world patients. If reported and corrected at early
stages, it could be beneficial for other patients having the same disease. Many government
agencies across different countries are aiming to collect the real time drug reactions from
patients through surveys, reporting tools etc. The direct feedback from patients based on their
experience is more relevant and informative in making decisions on restricting adverse drug
reactions. This self-reported patient feedback in free flowing format is available from
discussions in social media and medical blogs. Such data have more impact than government
induced feedback, reporting and surveys. In this paper our primary focus is to utilize social
media to mine the ADR mentions from patients. We filter tweets that might have adverse drug
reaction mentions. We compare the performance of various classification algorithms in
classifying accurately if a tweet comprises of ADR mention or not.