Mining Electronic Healthcare Record Databases to Augment Drug Safety Surveillance

It is perhaps a fundamental truth in medicine that there is no intervention – be it a drug, a medical device or a procedure – that is without risks. Even with the most rigorous eff orts in drug approval and regulation, there is not a drug out there that is 100% safe under all conditions. Randomized controlled trials (RCTs) are considered to be the most stringent approach to determining cause-and-eff ect relationship between an intervention and an outcome, but such trials are rarely designed or powered to detect uncommon or unexpected adverse events.1-4 Once drugs are marketed, they are used in a more diverse group of people, oft en for much longer periods, and sometimes with a wider range of therapeutic indications. While monitoring the risks associated with drug use has come a long way since the thalidomide disaster in the 1960s and the institution of spontaneous reporting systems (SRS), it has become evident that adverse eff ects of drugs may be detected - and acted upon – too late, when millions of persons have already been exposed.5-6 Th ere has been a growing clamor for improving the current passive-reactive paradigm of drug safety surveillance. Prominent issues in the last few years have emphasized the importance of a life-cycle approach to drug safety monitoring and the need to explore new methods to improve surveillance of drugs post-marketing.7-8 It has been posited that electronic healthcare record (EHR) databases represent an important resource for proactive surveillance and can augment existing pharmacovigilance systems. Various public-private initiatives worldwide have been launched, and great investments have been made, to explore the secondary use of EHR for this purpose.9-10 In this thesis, we draw on the experience of the EU-ADR network (Exploring and Understanding Adverse Drug Reactions by Integrative Mining of Clinical Records and Biomedical Knowledge, http://www.euadr-project.org/), a federation of eight EHR databases in four countries in Europe, to demonstrate the feasibility of combining diverse and diff erently structured data and pave the way for large-scale drug safety monitoring. We describe the opportunities and challenges that come with heterogeneity in database structure, with diff erences in language and coding of both drugs and diseases, and with the diversity in the organization of European healthcare systems.

• View at Worldcat

Free Full Text ( Final Version , 20mb )