Deployment of an ambient AI scribe in emergency care: A 12-month evaluation in a large Spanish hospital network.

Alcázar-Peral, Jose María; Álvaro-de la Parra, Juan Antonio; Ciardo, Paolo; Marcano, Henry José; Calderón, Miguel; Villar, Carlos; Blanco, Daniel; Cristóbal, Ion et al. · Int J Med Inform · 2026

retrospective_cohort · Level III

Where this comes from

Abstract

Ambient artificial intelligence (AI) scribes have the potential to reduce documentation burden and improve efficiency, clinician experience, and care quality. However, evidence from large-scale implementations in emergency departments (EDs), particularly regarding patient experience and documentation quality, remains limited. We aimed here to evaluate the implementation of an ambient AI-powered scribe in EDs, assessing adoption, efficiency, transcription accuracy, clinician experience, patient experience, and medical report quality. We conducted a 12-month multicenter retrospective observational study (February 2025-January 2026) across five emergency specialties in 48 Spanish hospitals. All level 4 and 5 emergency consultations were included, with approximately 2.27 million emergency visits during the study period. We analyzed monthly adoption rates, consultation duration, transcription accuracy, patient experience measured by Net Promoter Score, clinician experience assessed through cross-sectional surveys, and documentation quality evaluated via structured audits. Scribe was used in 1,032,558 consultations (45.3 %), with monthly adoption increasing from 7.7 % to 57.8 %. A total of 2,097 physicians used Scribe at least once, including 1,198 high-intensity users. Scribe-assisted consultations were significantly shorter (p < 0.001), with a mean relative time savings of 21.8 %. Transcription accuracy remained high and stable throughout the study (mean: 93.9 %). Audits of clinical reports showed higher documentation quality scores for Scribe-generated reports, with a large effect size and consistent improvements across physicians. Patient experience was significantly higher in Scribe-assisted consultations, and clinician experience was positive overall, particularly regarding workload, stress, and physician-patient interaction. Implementation of an ambient AI scribe in emergency care was feasible, scalable, and associated with improved efficiency, higher patient experience, enhanced documentation quality, and positive clinician experience. These findings support the role of ambient AI scribes as a value-based health care intervention that simultaneously improves care quality, patient experience, and provider well-being.