Abstract—Medical triage is the key in healthcare systems around the world, as there is a rising demand in healthcare systems against the available resources to accommodate them. The available automated symptom checkers and triage systems lack accuracy, over-triage, and inadequate performance on rare conditions. In order to address these difficulties, we propose TRIage AGent ENhanced Technology (TRIAGENT), a new multi-agent application to optimize medical triage, which evaluates patient symptom reports through a set of hierarchical structure with specific dialogue agent, symptom agent, and decision agent. A Dynamic Symptom Relationship Graph (DSRG) algorithm used by TRIAGENT builds individual symptom networks in real-time during patient interaction and is based on two knowledge graphs: general medical knowledge graph based on Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT) and the International Classification of Diseases, Tenth Revision (ICD-10) and a patient-specific dynamic graph constructed in response to reported symptoms and depends on symptom temporal relationships. The system applies contrastive learning-based rare disease detection and quantifications of uncertainties to risk-aware decision making. The overall analysis frequency of 10,000 clinical cases showcases that TRIAGENT reports the comparable triage classification mean of 89.7%, which is statistically superior by a margin of 17.2% points to the most accurate commercial systems (p < 0.001). The system shows unequivocal performance within different tested population (age: 18–85, diverse ethnicities) and symptom typology, especially in emergency management (F1 = 0.93) and self-care prescriptions (F1 = 0.91) and is capable of generating 2 (2X), 3 (3X), and up to 10 (10X) times less over-triage (7.3% vs. 15.2%) and under-triage (3.0% vs. 8.0%) rates than other leading commercial tool. It is also noteworthy that TRIAGENT does not degrade to the extent of known or rare conditions equal to <0.1%, with baselines as high as 70% accuracy rating, vastly exceeding existing implementations that tend to diminish to <45%, suggesting the potential of the system in increasing the access of healthcare opportunities and making sure that life-threatening medical conditions would not be overlooked.
 
Keywords—intelligent systems, medical triage systems, natural language processing, multi-agent architecture, symptom relationship mapping, healthcare artificial intelligence


Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).