Understanding the Impact of Epilepsy and Depression on Sleep Disorder: Beyond Associations

Vasundhara Acharya

Confirmation: I have read and agree with the IEEE BHI 2025 conference submission's policy on behalf of myself and my co-authors.

Keywords: Epilepsy; Sleep Disturbance; Causal Structure Discovery; Directed Acyclic Graph; Average Treatment Effect

TL;DR: Using potential causal inference on NHANES survey data, we show that both epilepsy and major depressive disorder exacerbate sleep disturbance beyond mere associations.

Abstract: Epilepsy, sleep disturbance (SD), and major depressive disorder (MDD) often co-occur and are further complicated by antiseizure medications (ASMs). Association-based methods struggle to untangle their causal links. Using National Health and Nutrition Examination Survey (NHANES) data from 2015–2020, we encoded unordered categorical variables via a variational autoencoder–principal component analysis (VAE-PCA) pipeline, learned a consensus directed acyclic graph (DAG) through an adaptive-bootstrapped ensemble of causal structure discovery (CSD) algorithms, and estimated backdoor-adjusted effects with generalized linear models (GLMs) and refutation tests. We observed direct average treatment effects (ATEs) of epilepsy on SD (0.362), MDD on SD (0.167), and ASMs affecting sleep architecture on SD (0.206), but no significant links between ASMs and MDD or between epilepsy and MDD. VAE embeddings highlighted population heterogeneity. While providing valuable hypotheses, the cross-sectional design limits definitive causal claims. These findings are a foundation for future longitudinal studies incorporating objective measures and advanced methods.

Track: 6. Theoretical Biomedical Informatics

Registration Id: 9ZNCTC567FT

Submission Number: 21