IDENTIFICATION OF VALID CASE REPORTS FROM DIGITAL SOCIAL MEDIA WITH THE HELP OF MACHINE LEARNING

Abstract

Social media becomes a novel source for information about pharmacovigilance (PV) and patient perceptions on adverse events. The main questions remain on the supplement routine of social media to provide PV surveillance. This research work has the objectives to find out whether the social media data assessment can determine the new signals, recognized signals from routine PV, recognized signals earlier and certain problems such as patients perceptions and quality issues. In addition, the task is to obtain the number of ‘posts with similarity to AEs’ (proto- AEs) and the classifications/forms of products that receive the advantage from social media data assessment. Worldwide, adverse drug reactions (ADRs) is the severe menace for public health leads to sickness, frailty or even death. Up to now, few countries create a framework to guide the post market drug safety strategy in place of emergent consent to pharmaceuticals globally and emerging trend to gather information including therapeutic record in the post-market rather than premarket time. Such crisis involves a novel hypothetical concept as ‘pharmacovigilance’. It consists the entire governing structures, policy instruments and institutional authority like the capability to perform, employ and execute the processes, rules, regulations and guidelines. These can handle the entire endorsed social interests related with the patient’s safety and protection from ADRs. A methodical configuration for examining pharmacovigilance is showed in this research work that explains the relationship between the governance and pharmacovigilance. It explains the influential level of pharmacovigilance. Several databases included to cover a broad range of domain as grey literature, computer science, nursing as well as health and medical research. This work also considered the other supplementary approaches that involved searching of internet search engines, contacting experts in this field, reference checking all consisted articles and included articles, newsletters and conference proceedings, hand searching journals, browsing internet blogs and related systematic reviews. This research includes the data collected from the social media sites such as Flickr Vkontakate,Web,Facebook, YouTube, Twitter, Tumblr, DailyMotion and Reddit.The internet searches and database detected 5075 records. Such outcomes were further augmented with the research works acknowledged from peer reviewers suggestions, contacting experts, reference vii checking, hand searching and studies already recognized by the authors. The outcomes from the selected social media sites were further used to calculate the adverse event of posts by using the narrative analysis. Nearly 4% of adverse event measured from the posts of Facebook, while it is in between 2-4% for the Twitter. The pilot study suggests adverse events are recognizable on social media by monitoring the major social sites. Although, it has a substantial heterogeneity in the type as well as frequency of reported events. In addition, the validity or reliability of data has not been comprehensively assessed.