Knowledge-augmented Agentic AI for Mental Health Medication Information Seeking

A new framework combines patient narratives from Reddit and WebMD with two decades of FDA records to surface psychiatric medication safety signals, researchers report. The system uses a knowledge-graph-based multi-agent architecture to keep regulatory facts distinct from patient experience. The study, posted to arXiv on 24 June 2026, integrated 466,525 Reddit posts, 60,782 WebMD reviews, and twenty years of U.S. FDA Adverse Event Reporting System data covering nine antidepressants [1][2]. A large-language-model entity-recognition pipeline, benchmarked against physician annotations, achieved F1 scores of 0.969 for medications and 0.973 for conditions [1][2]. The two community platforms showed far greater concordance with each other — reaching a Jaccard similarity of 0.905 — than either did with regulatory reports, suggesting patient-generated data constitute a partly independent safety signal [1][2]. For sertraline, many adverse events appeared in community sources hundreds of days before the corresponding FDA date [1][2]. A Neo4j knowledge graph grounded in ATC-N, ICD-10, and MedDRA vocabularies preserves provenance, keeping every claim traceable [1][2]. The authors state that usefulness and patient benefit must be tested prospectively [1][2]. The work arrives as artificial intelligence applications in healthcare face scrutiny over data privacy, algorithmic bias, and patient trust [3]. The researchers note that poorly contextualised psychiatric medication information can amplify fear, nocebo responses, and non-adherence [1][2]. Reddit communities have previously drawn criticism for shaping biased views of evidence-based medicine, with critics arguing that inconsistent enforcement of platform rules can allow misleading health content to persist [7]. The framework does not conflate anecdotal reports with regulatory findings; instead it keeps the two evidentiary streams distinct while making both queryable through a single interface [1][2]. The study examined nine antidepressants; the authors did not release patient-level data but made the knowledge-graph schema and entity-recognition benchmarks publicly available [1][2].