The main goal of this project is to develop a set of methods and tools for detecting drug safety signals from social media to supplement the traditional methods based on the standard data such as electronic health records and centralized reporting system, MedWatch.
Develop and implement a social media crawler Develop algorithms for mining associations between drugs and ADRs with the support of
consumer health vocabulary (CHV)
Develop a working prototype for drug safety signal detection
A Java prototype is developed and an experiment is conducted. The experimental results show that the proposed technique achieves the sensitivity of 0.805 and the specificity of 0.690.
This project confirms that social media is a valuable resource for detecting drug safety signal. However, a good development of consumer health vocabulary is important because the social media mainly consists of health consumer opinions, in which the vocabulary is very different from the health professional vocabulary.