Markers of adverse drug reactions in medication histories. An analysis of inhaled steroid utilization

Abstract

Prescription sequence analysis is based on the observation that a subset of adverse drug reactions are themselves indications for the prescription of another drug. We propose an approach to the analysis of clustering of prescriptions in medication histories that contain records of a presumed side-effect-causing drug A and a side-effect-alleviating drug B. This set of histories with records of both A and B is analysed in a logistic regression model that considers the start of B against exposure and non-exposure of A, stratified on the level of the individual histories. A correction is presented for the periods of B use, during which a new start of B cannot occur. Prescription sequence analysis is demonstrated with a set of histories of use of inhaled steroids (A) and topical oral anticandida drugs (B). A can cause oral candidiasis, B is therapy for this condition. An odds ratio of 1.43 was found for the association of A use and the initiation of therapy with B.