ANALYSIS OF MENTAL WELLNESS OUTCOMES USING SOCIAL MEDIA AND CLINICAL DATA: A DIGITAL EPIDEMIOLOGY APPROACH

Abstract

This work proposes a digital epidemiology approach that aligns social media data and electronic health records (EHRs) for the prediction of mental well-being outcomes through the application of the concept of digital phenotyping, the real-time quantification of human behaviors through digital traces. We utilized a large Twitter post dataset (≈ 500,000 users) and corresponding linked EHRs (with clinically validated depression and anxiety diagnoses derived from PHQ-9 and GAD-7 tools). We used natural language processing (NLP) and machine learning models like XGBoost and random forests to extract linguistic features like sentiment, pronoun frequency, temporal posting patterns, and activity measures predictive of behavioral markers. The prediction was highly accurate with an Area Under the Curve (AUC) of ~0.84 for depression and ~0.80 for anxiety, and was substantially better than EHR-only models by ~13%. Specifically, social media-derived predictors predicted clinical symptom worsening as early as four months prior to official diagnosis. The findings affirm that the combination of social media text and clinical information increases early prediction and anticipation of mental health outcomes. The stakes are high: hybrid monitoring could allow for timely, individualized treatment for vulnerable individuals. But large-scale production will depend on careful attention to algorithmic justice, data confidentiality, and representative inclusion across diverse groups. Our results lean toward a behavior-based, ethical surveillance paradigm that connects online behaviors and clinical expertise to inform novel mental health care.