Thursday, December 12, 2013; 10:30 – 10:50am
Grand Ballroom D
Rolina van Gaalen, McGill University
Post-marketing surveillance of pharmaceuticals is critically important to ensure patient safety. Traditional study designs and methods for pharmacoepidemiology that have been adapted for prospective pharmacovigilance currently rely on simple metrics to define exposure. The true relationship between past drug use and risk of adverse events is usually unknown and may be complex. We used simulation to assess the timeliness with which simple parametric models and more recently-developed, complex flexible estimation models detect different patterns of drug / adverse event association using hazard models. Our results suggest that parametric models may be sufficient to detect both simple and complex harmful associations.
*NOTE: At the request of the authors, this abstract was not published in the Online Journal of Public Health Special Supplement.