Identification of outcome predictors in transcranial direct current stimulation for major depression through explainable data-driven analysis

Accurate and interpretable prediction of treatment outcomes is critical for advancing personalized interventions in major depressive disorder (MDD), particularly in the context of neuromodulatory therapies such as transcranial direct current stimulation (tDCS). Traditional approaches often lack precision and offer limited insight into the clinical profiles associated with treatment response. This study aims to identify key demographic and clinical predictors of tDCS efficacy in patients with depression, using interpretable, data-driven modeling. We analyzed a multi-center dataset comprising 169 patients with depression who underwent tDCS treatment. A supervised learning framework was employed to model treatment response, integrating explainable artificial intelligence techniques to enhance clinical interpretability. Predictive features included age, gender, baseline symptom severity, tDCS protocol, diagnostic subtype, comorbidities, illness duration, and concurrent medications. The model achieved a predictive accuracy of approximately 58% in classifying treatment responders and non-responders. Explainability analyses revealed that the most influential factors included age range (40–49 years), gender (female), medication status (drug-naive), and chronicity of illness (duration > 16 years). These characteristics defined the subgroup most likely to benefit from tDCS intervention. Our findings demonstrate the potential of explainable machine learning to support outcome prediction in tDCS therapy for depression. By highlighting clinically meaningful predictors and patient profiles, this approach may facilitate more personalized and effective treatment strategies in psychiatric care.