Triage Nurses' Perceptions of Artificial Intelligence: A Cross-Sectional Study.

INTRODUCTION: Emergency department crowding and increasing patient complexity challenge traditional triage models. Artificial intelligence may support triage decision making, but nurses' perceptions shape real-world adoption. METHODS: We conducted a cross-sectional online survey (February to April 2024) among triage nurses from 7 emergency departments within the USL Toscana Centro, Italy. The questionnaire assessed 5 artificial intelligence dimensions, affidability (reliability), applicability, training, relational impact, and safety (binary items), and a composite "Total Challenges" score (0-5). Descriptive statistics, chi-square tests with Cramer's V, Mann-Whitney U, and Welch analysis of variance were applied (α = 0.05). RESULTS: Eighty-four nurses participated (73.8% female; largest age group, 35-54 years; 48.8%). Positive perceptions were as follows: relational impact, 63.1%; affidability (reliability), 48.8%; applicability, 47.6%; training, 36.9%; and safety, 36.9%. The mean of total challenges was 2.33 (SD = 1.23; range, 1-5). Age was associated with training (χ2 = 7.122; degrees of freedom = 2; P = .028; V = 0.291), with older nurses reporting lower positive perceptions. Total challenges differed by age group (Welch analysis of variance, F[2, 46.042] = 4.106; P = .023), indicating higher perceived barriers among nurses aged 55 years or older. No significant associations emerged for sex or years of experience. DISCUSSION: Triage nurses showed cautious optimism toward artificial intelligence, valuing potential relational benefits but expressing concerns about training and safety. Older age was linked to greater perceived barriers. Targeted, competency-based education and clear governance are needed to support safe, human-centered artificial intelligence integration in emergency triage.