A0448
Title: Examining directional association between depression and anxiety
Authors: Soumik Purkayastha - University of Pittsburgh (United States) [presenting]
Abstract: Utilizing a novel entropy loss (EL) metric, a causal discovery method is proposed to understand directional effects in the causal relationship between depression and anxiety among medical interns. This method advances existing methods of bivariate causal discovery with theoretical guarantees of causal effect identifiability and statistical inference and enjoys good computational performance. Using data from the intern health study (n=6,858), the proposed method reveals with high statistical confidence that depression scores (PHQ-9) consistently predispose anxiety scores (GAD-7) across four longitudinal visits of the study, controlling for demographic confounders. This finding provides crucial insights into the directional effect useful for mental health intervention strategies for medical interns. Simulation studies demonstrate that EL achieves nearly superior accuracy compared to existing approaches across various conditions with reduced computation time. The EL framework's ability to handle discrete clinical scores while adjusting for confounders makes it particularly valuable for psychiatric epidemiology and broader applications in causal discovery with discrete data.
