Ambulatuvar İleri Uygulama Sağlayıcıları Arasında Yapay Zeka Yazıcı Teknolojisinin Uygulanmasına Yönelik Toplu Pazarlık Engellerini Aşmak: Karma Yöntem Kalite İyileştirme Çalışması

Özet (abstract)

Background Advanced practice providers (APPs) face rising documentation demands driven by increased productivity pressures, contributing to burnout and reduced time with patients. AI scribes may reduce documentation time and improve clinician well-being and patient interaction. Labor unions have expressed concerns about AI in health care affecting job security, safety, ethics, and governance. Little is known about the deployment of AI scribes for APPs in a union environment. Objective This project aimed to describe the process, challenges, and outcomes of implementing an AI scribe for APPs within a unionized ambulatory academic setting, focusing on collective bargaining considerations and using a pilot project to help address APP union concerns. Methods Following formal notices to labor unions and a meet-and-confer process consistent with California public employer obligations, we conducted an 11-week (55-workday) pilot (June 16 to August 31, 2025) of the Abridge AI scribe among 15 primary care APPs (n=12 nurse practitioners and n=3 physician assistants). Training modules covered consent, privacy, and documentation verification. We tracked scribe use, percentage of AI notes kept, time spent on notes, and same-day encounter closures; results were reported to stakeholders. Pilot metrics were compared with those of our organization’s nonpilot APPs, and additional postpilot survey data were obtained from the APPs by the vendor. This information, along with the required training and patient and health care provider protections, was used to validate this technology to the labor unions. Results Across 8100 APP encounters, the AI scribe was used in 5403 (66.7%) notes. Individual use ranged from 30% (54/179) to 89% (501/563; SD 16.9%). The average percentage of AI-generated note content retained by the APP was 78% (SD 18.2%; range 20%-93%). The time spent in the note writer for the pilot group was a mean 6 (SD 2.2; median 6, IQR 5.0-8.0) minutes vs a mean 14 (SD 11.2; median 11, IQR 8.4-18.5) minutes for nonpilot APPs. Same-day encounter closure in the pilot group was 90% (7290/8100; SD 9.3%) vs 72% (23,682/32,983; SD 30.0%) in the nonpilot group. Postpilot feedback from the pilot group showed positive APP sentiment. No postpilot formal or informal concerns were raised, and the labor unions agreed to widespread implementation of AI scribe technology across the organization for our ambulatory APPs. Conclusions AI scribe implementation among unionized APPs was feasible when paired with structured labor engagement, pilot-testing, privacy safeguards, and optional adoption. The pilot yielded descriptive information showing a difference between pilot and nonpilot AI scribe users, with limited generalizability beyond the context of labor union engagement. The pilot’s efficiency differences and documentation timeliness align with existing published data, which gave confidence in the labor unions’ adoption of the technology. Future work will examine variability in use and effort reduction across APPs.

Yazarların özeti; kaynağından alınmıştır. JMIR Formative Research, 2026 · DOI ↗