The Impact of Ambient Voice Technology in NHS Outpatient Care: A Replicable Protocol for a Mixed-Methods Service Evaluation (Preprint)

BACKGROUND The National Health Service (NHS) faces unprecedented outpatient backlogs owing to the legacy of COVID-19 and long-standing systemic pressures including workforce shortages and rising healthcare utilisation. Documentation demands within electronic health records (EHRs) are a major contributor to reduced clinical capacity, clinician burnout, and diminished patient-clinician communication. Ambient AI documentation tools, which passively capture consultations and generate structured clinical notes, have shown promise in improving efficiency and satisfaction, but evidence from NHS settings remains limited. OBJECTIVE This service evaluation aims to assess the implementation and impact of an ambient voice technology (AVT) scribe across outpatient specialties at a large NHS university healthcare trust. Assessment is structured across four domains: productivity, clinician experience, patient experience, and documentation quality, with a supplementary evaluation of the primary-secondary care communication interface. METHODS This mixed-methods service evaluation uses a phased rollout of AVT integrated with the Epic EHR across NHS outpatient specialties. The primary outcome is clinician documentation activity time per encounter, derived from Epic Caboodle analytics and compared between AVT-assisted and standard encounters using mixed-effects linear regression, with the first 30 days of AVT use per clinician excluded as a familiarisation period. Secondary productivity outcomes include same-day encounter closure, same-day documentation completion, patients seen per clinician-day, and time to finalised clinic letter; an interrupted time series provides a complementary analysis. Clinician experience is assessed by anonymised surveys using the NASA Task Load Index (NASA-TLX) and Mini-Z 2.0. Patient experience is assessed through Patient-Reported Experience Measures. Documentation quality is assessed by blinded reviewers using the Sheffield Assessment Instrument for Letters, and the primary-secondary care interface is evaluated using a Delphi-derived questionnaire. Qualitative data are analysed thematically. The evaluation will be reported per StaRI guidelines. RESULTS Pre-AVT baseline data collection commenced in September 2025. Baseline data were collected from 121 consultants (mean NASA-TLX score 54.4, SD 12.6; 34.7% reporting burnout), 1054 patients or carers (98.5% satisfied with their consultation), and 390,843 outpatient encounters from 1,278 consultants (mean documentation activity time per encounter 7.6 min, SD 9.0; median 5.0 min). Post-implementation data collection is ongoing. Interim findings are anticipated in September 2026 and final results in September 2027. CONCLUSIONS This evaluation will provide real-world evidence on the impact, acceptability, and scalability of AVT in NHS outpatient care. A key contribution is the development of a reproducible evaluation framework, using routinely available EHR analytics data and standardised instruments, that can be adopted by other NHS trusts, particularly those using Epic, to assess AVT implementation in their own settings.