‘Extreme Duplication’ in the US FDA Adverse Events Reporting System Database

Excerpt

We recently encountered an example of extreme duplication in the publicly released version of the US FDA Adverse Events Reporting System (AERS) database, which is available through the Freedom of Information (FOI) Act. We use the term ‘extreme duplication’ because the majority of reports of the drug-event pair under study were found to be duplicate reports, and these duplicates resulted in the occurrence of a large signal of disproportionate reporting (SDR)[1] discovered during a data-mining exercise. This occurred during a recent demonstration within our company of a commercial vendor’s data-mining software containing these data. Data mining includes emerging computer-based quantitative tools for screening spontaneous reporting system (SRS) databases to assist in the identification of potential new drug safety issues. Although duplicate reporting is one of the numerous well recognised forms of data corruption and distortion in SRS databases, the finding surprised us nonetheless, given the duplicate detection algorithms and procedures included in the commercial data-mining software we used. …