Detection of Long Term Adverse Drug Reactions in Electronic Health Care Data

A substantial group of Adverse Drug Reactions (ADRs) is likely to have a much longer time between first exposure and adverse event occurrence, and the risk could actually be increasing with cumulative exposure. The availability of large amounts of longitudinal observational health care data with long follow-up now puts us in a unique position where we would be able to pick up these types of ADRs, but appropriate signal detection screening methods are lacking.

The objective of this research is to develop methods that can detect long term ADRs. The first step will be to develop test data that can be used to evaluate the methods on their ability to pick up long-term ADRs in observational data. Two types of data will be used: simulated data and real data. Methods will be developed to detect ADRs that are correlated with cumulative exposure, and to automatically identify groups of patients that can be used as comparison to eliminate selection bias as much as possible.

Deliverables of this research will be posted below as available.