Detecting adverse drug events: accuracy and generalizability

The incidence and impacts of adverse drug events (ADE) have been extensively studied, but there is an emerging focus on real-time detection systems. These can play an important role, along with systems pharmacology and population-level epidemiology, in a multipronged approach to prevent ADEs and mitigate their harm. Tailoring ADE detection systems to a particular health care setting or ADE type can improve predictive accuracy, but the added complexity reduces its wider applicability. As this approach becomes increasingly used we can imagine detection algorithms of greater complexity but also a set of algorithms encompassing the full range of health care settings and ADE types, thus creating a system that is both accurate and widely applicable.