Symptom Distribution Heterogeneity as a Marker for the Course of Major Depression

Major Depressive Disorder (MDD) presents diverse symptom patterns, such as sadness or fatigue, influencing its course and treatment response. Using dynamical systems and network theories, which model depression as interconnected symptoms stabilizing in specific patterns (attractor dynamics), we examined whether symptom structure reflects clinical state stability. We applied four mathematical metrics---Entropy, Variance, Gini-Simpson Index, and Coefficient of Variation---as proxies for symptom structure, independent of overall severity, using cross-sectional data. In two studies (longitudinal cohort, $N=1799$; inpatient sample, $N=255$), we compared Healthy Controls vs.\MDD, acute vs.\remitted, and pre- vs.\post-treatment states. Healthier or improved states showed lower Entropy/Gini-Simpson and higher Variance/Coefficient of Variation, indicating more flexible symptom profiles. These metrics distinguished clinical states with 73\% accuracy and achieved 63\% generalizability across datasets after calibration. Unlike prior network studies focusing on symptom interactions, our severity-independent markers reveal the ''stickiness'' of depressive states, linking rigid profiles to persistent depression and flexible ones to recovery. This approach offers a scalable tool for personalized MDD interventions using routine clinical data.