Abstract:
|
When a toxic drug is taken, especially for an extended period, numerous adverse reactions (AEs) may be observed in each patient. Frequency of these adverse events will vary. The primary thrust of this paper is to evaluate safety signals by using association rule learning. An association rule learning is a machine learning (ML) method for discovering relations between different items (i.e. AEs) in a large dataset. Confidence and lift are two measures that describe the intensity of the relationship. The methodological aspect will be fully described. In the empirical setting, relevant rules will be picked using the highest lift and the confidence. Graphical techniques such as scatter plot (confidence vs lift), group matrix plot for different associations will be explored. Finally, an R-shiny app will be developed and the association rules will be scrutinized for evaluating safety signal.
|