Optimizing Hip and Knee Arthroplasty Clinic Flow: A Prospective Evaluation of Artificial Intelligence Scribe Technology

Background The increasing burden of clinical documentation contributes to physician inefficiency and burnout, with electronic health record implementation significantly increasing documentation time from 16% to 28% of clinical time. While human medical scribes have shown benefits, artificial intelligence (AI) scribes represent a promising solution. This study evaluated the impact of AI scribe technology on clinical encounter duration and physician workload in a high-volume hip and knee arthroplasty clinic. Methods This prospective quality-improvement cohort study was conducted by a single surgeon at one institution from November 2024 to February 2025. Hip and knee arthroplasty patients were divided into 2 cohorts: a control group using traditional physician and medical assistant documentation within the electronic medical record and an intervention group using AI-powered clinical documentation. Clinical encounter duration was assessed by 3 independent reviewers using direct observation and stopwatch timing. The National Aeronautics and Space Administration Task Load Index (NASA-TLX) was completed after each clinic day to quantify subjective physician workload across 6 domains on a 20-point scale. Results A total of 112 patient encounters were analyzed: 55 with AI scribes and 57 with traditional documentation. Encounter time was significantly shorter in the AI group, with a mean of 12.03 minutes (range, 4.01-22.13) compared to 15.06 minutes (range, 6.20-32.27) for traditional documentation ( P < .001). Overall, physician workload measured by NASA-TLX was lower with AI assistance, with a mean total score of 30.0 (range, 26-40) vs 40.2 (range, 11-73) for traditional documentation ( P = .485). The AI group showed consistently lower mean scores across all NASA-TLX subdomains, including temporal demand (5.8 vs 10.4), effort (6.8 vs 10.0), and frustration (3.4 vs 5.4). Conclusions AI scribe implementation significantly reduced clinical encounter duration by approximately 20% and demonstrated promising improvements in physician workload metrics.