JMIR Publications

ABSTRACT

Background:

In order to reduce safety issues, regulatory agencies have to make decisions regarding drugs usage and adverse drug reactions (ADRs). Spontaneous reporting by health professionals is the main pharmacovigilance data source. However, underreporting is one of the major limitations of the current pharmacovigilance system. Other new sources for drug reactions have been explored, such as electronic patient records and social media. Yet, social media have not been fully exploited.

Objective:

We assume that the difficulty for regulatory agencies to use social media is mainly associated with the lack of an appropriate tool to support decision makers. Such a tool should allow to access the information through a graphical user interface (GUI) presenting information in a user-friendly way rather than raw data, allowing users to choose visualization options that better represent the data and leads to making better decisions.

Methods:

We thus developed a pipeline composed of four independently editable modules allowing to study in an efficient way health related French web forums for a pharmacovigilance usage. The four modules were (1) web forums’ posts’ extraction, (2) web forums’ posts’ annotation, (3) statistics and signal detection algorithm, and (4) a user interface.

Results:

We showed that using this pipeline and through the user interface, a quantitative analysis is possible and easy to lead with no need of coding and that results were coherent and could also bring new information about drugs.

Conclusions:

Our hypothesis is only partially confirmed as final users were not included in the evaluation. Further studies focusing on ergonomics and impact for professionals in regulatory agencies are necessary in future work. Our pipeline could be used in health surveillance by regulatory agencies such as the French ANSM or by pharmaceutical companies to help detect issues, and could also be used by research teams to retrospectively analyze events.