Evaluation of Complex Interventions on Drug (Re)Fill Behavior Using Electronic Health Care Data

This study develops generalized per-protocol analyses based on Inverse Probability Weighting and Targeted Learning to control for or evaluate the effect of medication nonadherence in observational studies that use electronic health databases to inform the management of chronic conditions. Unlike current implementations of these analyses that require investigators to make untestable and often arbitrary assumptions regarding patients’ use of drugs from pharmacy dispensing data, the proposed causal methodologies directly evaluate the health effects of drug (re)fill regimens. To develop these analytic approaches, we seek guidance from stakeholder partners representing a broad range of constituencies to illustrate and ensure the practical relevance of the proposed methods from two prior studies with electronic health databases conducted to inform the management of type 2 diabetes.

0