Pharmacovigilance in the era of social media: Discovering adverse drug events cross-relating Twitter and PubMed

Abstract

Abstract In pharmacovigilance, post-marketing surveillance is mainly supported by spontaneous reporting systems (SRS) collecting adverse drug events explicitly submitted by physicians alerted by their patients. Nowadays, this activity could leverage on mining opinions and experiences of individuals from social media by monitoring users’ posts citing symptoms, drugs, etc. The most critical problem is the reliability of the information sources. In order to address this challenge, the proposed method tries to cross-relate heterogeneous data sources with correspondingly different levels of trustworthiness. It filters out assertions quoted on social media on the basis of data validated by official information sources. Method adopts the Fuzzy Formal Concept Analysis (Fuzzy FCA) in order to evaluate the reliability of adverse drug events extracted on Twitter and PubMed. It keeps track of difference between the co-citation frequencies by calculating a residual threshold τ . The main outcome is that with τ in the range [ − 4 , + 4 ] , 91% of drug and side effect correlations extracted from tweets can be considered reliable, according to the official site (we used http://sideeffects.embl.de ).