| dc.description.abstract | Medication non-adherence is a widespread problem and has been known to be associated with worse health outcomes and increased healthcare costs. There have been large efforts to improve adherence by finding effective interventions, but an agreement on a good measure for adherence still has not been established. Although many measures of adherence have been developed, their definitions and calculations are not transparent, nor consistent across studies. Furthermore, statistical methods for analyzing adherence measures have not been rigorously evaluated yet. In this thesis, we present a summary of commonly used adherence measures, discuss the strengths and weaknesses of each, and suggest a framework for the generalization of outcome derivations. We then discuss modeling strategies using common derivations of medication adherence as the outcome. Four different statistical methods are considered: logistic, ordinal, negative binomial regressions, and generalized estimation equation (GEE) methodology. We present a case study using medication adherence data from 653 patients and conduct a sensitivity analysis across the four models and seven different outcome generating mechanisms. Finally, we conduct simulation studies to evaluate the performance of the four models using different criteria for patient inclusion. | |
