Grant: Discovery Grants Program - Individual

Self-reported data that describes an individual's negative experiences with a product or service may be biased for statistical analysis, especially when the motivation to report is driven by negative events. Instead, these data can be useful to rapidly signal new potential risks which may help to ensure product safety standards. Existing statistical methods were developed to study how often negative events are reported; however, the reasons individuals choose to report may be related to other things than only their experiences with a product or service, for example, due to media coverage. In addition, those who wish to report may tell a different story, on average, than those who do not, making it difficult to tell whether the signals are valid. This may complicate the decision to launch follow-up studies on possible safety issues which can be a major concern when unrealistically large volumes of signals are flagged. This research proposes to develop new statistical methods that will produce a different approach to detecting safety signals by combining the self-reported information with additional sources of relevant data through mathematical modelling. This will allow the signal detection problem to be one that narrows many possible explanations down to a small set of plausible reasons which will enable experts to better understand the potential for a safety concern before launching follow-up studies. This research is motivated by the problem of detecting side effects for new medications and the methods will be tested out on the freely available FDA Adverse Event Reporting System, which is a collection of self reported drug side effect data. Longer term objectives of this research will be to extend the methodologies to alternate sources of self-reported data in commerce, health and in the natural sciences and engineering.