Mediation Analysis of Co-occurring Conditions for Complex Longitudinal Clinical Data
Douglas David Gunzler, Nathan J. Morris, Adam T. Perzynski, Deborah Miller, Steven Lewis, Robert A. Bermel · SM Journal of Biometrics & Biostatistics · 2017
BackgroundIn mediation, we consider an intermediate variable, called the mediator, that helps explain how or why an independent variable influences an outcome [1][2][3].Gaining an understanding of which particular disabilities act as mechanisms of change (or mediators) between co-occurring conditions would allow for more focused treatment for a patient given changes in severity of both conditions.In many cases of co-occurring conditions, however, temporal precedence for mediation may be unclear due to the unknown origin of symptoms in the conditions (i.e.cognitive impairment in multiple sclerosis and depression symptoms).Thus, in such cases, the mediator and the outcome measured across multiple time points may be viewed as separate parallel processes [4].As a result, the mediational process can be defined as the independent variable influencing the growth of the mediator, which, in turn, affects the growth of the outcome [4].Further, symptoms of the co-occurring conditions may overlap.This can lead to inappropriate clinical decisions (medication selection, escalation, etc.) and to incorrect inferences regarding treatment effectiveness.Scales and diagnoses for patients with co-occurring conditions may be especially problematic as they may suffer from criterion contamination due to the overlap.Criterion contamination occurs when the criterion measure is affected by "construct-irrelevant" [5] factors that are not part of the criterion construct.Methods have been previously proposed to adjust scales to better represent the underlying dimension of the criterion measure for patients with co-occurring conditions using cross-sectional data [6,7].Here we propose methods which should improve our ability to perform mediation analysis and distinguish symptom changes over time in patients with co-occurring conditions.Overlapping symptoms of co-occurring conditions leads to a type of systematic error in scales or diagnoses known as Differential Item Functioning (DIF).DIF can occur when people from different groups (e.g., levels of Multiple Sclerosis (MS)-related fatigue) with the same latent trait (level of depression) have a different probability of giving a certain response on a questionnaire or test (e.g., items for sleep problems and fatigue of the PHQ-9).