Use of Generative AI for Digital Triage: Clinical Accuracy, Risks, and Professional Acceptance

This doctoral study employed a sequential explanatory mixed-methods design (QUAN → QUAL) to investigate the integration of generative AI into digital triage systems, analyzing clinical precision, perceived risks, and professional acceptance. Quantitative results from 152 healthcare professionals and 300 standardized cases revealed that AI achieved 78.3\% overall triage accuracy, performing well on critical cases (91.2\%) but with a significantly higher rate of under-triage errors in urgent cases compared to humans. Survey data (TAM-3 model) showed moderate acceptance, heavily mediated by clinical specialty (higher among nurses), perceived risks, and trust, which was found to be conditional and gradual. Qualitative interviews with 30 participants identified transparency, human oversight protocols, and clear accountability as non-negotiable prerequisites for adoption. The triangulated findings conclude that generative AI's potential in digital triage is constrained not primarily by technical accuracy, but by socio-technical and trust-related barriers. Successful integration requires an "augmented intelligence" paradigm where AI acts as a supportive tool under stringent human supervision, embedded within a robust governance framework that prioritizes patient safety and professional agency over full automation.