|
Abstract:
|
We introduce an interface which is developed to make the LRT a web based tool in OpenFDA. The web based LRT tool will provide the public a better insight of valuable adverse event database. The LRT method and tool are designed to search for signals of adverse events (AEs) associated with a drug of interest or for signals of drugs associated with an AE of interest. This methodology for identifying signals is not solely searching for AE with unusually high report counts, but is model based, using a Poisson model for AE report counts, that compares the reporting rates of AEs to detect potential signals. The advantage of the LRT method for signal detection is that the test controls the family-wise type-I error and false discovery rate, in addition to having other good operating characteristics such as high sensitivity and power. The LRT API extends the use of OpenFDA using counts from OpenFDA queries and provides summarized table and visual display of the signals. We also discuss the adjustments made to the LRT methodology for the OpenFDA query system, and study the performance of the adjusted LRT method.
|
