A large European study found that advanced machine learning models can predict which patients with epilepsy will develop depression — and vice versa — up to a year in advance. Analyzing data from nearly 12 million patients across seven countries, the models flagged key risk factors including low socioeconomic status, alcohol use, and high healthcare utilization. The findings could help clinicians intervene earlier in high-risk patients.
Epilepsy and depression are closely linked, but predicting which patients will develop one condition given the other has been a clinical challenge — until now. A large retrospective study published in BMJ Neurology Open used advanced machine learning (ML) models to identify demographic, socioeconomic, and clinical predictors of future disease onset in both directions: depression in patients with epilepsy (PWE) and epilepsy in patients with depression (PWD). The study drew on longitudinal data from 18 sources across seven European countries.
The models performed best on UK data, and shared risk factors across both patient groups included low socioeconomic status, high psychiatric multimorbidity, and heavy healthcare utilization — suggesting a common vulnerability profile that cuts across both conditions.
By the Numbers:
Why it matters: These ML tools could give clinicians a meaningful head start — identifying high-risk individuals before comorbidity strikes and enabling timely, targeted interventions. The authors call for a multidisciplinary care model that addresses the overlapping risk factors shared between epilepsy and depression.
