Data fusion aspects of pharmacovigilance

Safety signal detection problems in pharmacovigilance are examined in the general context of data fusion systems. The general perspective reveals commonalities between seemingly disparate fields of endeavor, and it suggests methods and technologies that may provide a “way forward” to advance to state of the art. Such insights are examined for exploring safety signal detection problems in (passive) pharmacovigilance using cross-functional data such as product complaint, medical inquiries, and adverse event databases. This paper discusses the definition of signal and safety signal, and issues that arise in assertion of signal present in the pharmacovigilance setting. Visualization methods for understanding correlations that suggest the presence of signal are considered. A multivariate correlation statistic is proposed to elucidate possible joint dependencies across multiple data streams.