Online Program

A Latent Growth Modeling Approach to Longitudinal Mediation Analysis of the Causal Path Between Multiple Sclerosis & Depression

*Douglas David Gunzler, Case Western Reserve University 

Keywords: structural equation modeling, mediation analysis, latent growth modeling, electronic health records, patient-reported outcomes

While multiple sclerosis (MS) patients commonly experience depressive symptoms, clinicians cannot reliably distinguish the mediating pathways through which different trajectories of MS leads to depression over time. In this talk, a longitudinal structural equation modeling (SEM)-based approach, latent growth modeling (LGM), will be discussed for examining the hypothesized mechanism, functional disability, through which MS, as defined by type and baseline time since diagnosis, leads to depression using the Knowledge Program (KP) at the Cleveland Clinic's Neurological Institute data base. SEM is a very general technique combining complex path models with latent (unobserved) variables. LGM is a practical application of SEM for longitudinal data to estimate growth trajectory, analogous to mixed effects modeling in more traditional analyses. The KP links patient-reported depression (via the PHQ-9) responses to the EPIC EHR and provides a powerful opportunity to study and improve patient care and clinical research. SEM is a very appropriate approach to handing the latent variables, causality questions and irregular follow-up times in the KP data base.