A0308
Title: Statistical approach to handling measurement issues in self-reported data that are longitudinally collected
Authors: MinJae Lee - University of Texas Southwestern (United States) [presenting]
Abstract: In the era of precision medicine, researchers are increasingly sensitive to the heterogeneity among at-risk individuals. Evaluating the association between disease progressions and the longitudinal pattern of pharmacological therapy has become more important. However, in many longitudinal studies, self-reported medication usage data collected at patients' follow-up visits could be missing and/or inaccurate/untenable information. These patterns may also dramatically differ between individuals and thus complicate determining the trajectory of medication use and its complete effects on patients. Although traditional existing methods can deal with specific types of missing/incomplete data, inappropriate handling of this complex issue can lead to misleading findings, especially when it depends upon multiple sources of variation over time. A latent class-based statistical approach under the Bayesian quantile regression framework is proposed that incorporates a cluster of unobserved heterogeneity for handling medication usage data with various measurement issues. Findings from the simulation study indicate that the proposed method performs better than traditional methods under certain data distribution scenarios. Applications of the proposed method are also illustrated to real data obtained from the longitudinal study.