Clinician EHR Work Burden Review Post Ambient AI Implementation
Kozlowski, Keiran Cangiano, Michelle Maloney, Sean Jacobs, Alicia McEntee, Rachel
Kozlowski, Keiran
Cangiano, Michelle
Maloney, Sean
Jacobs, Alicia
McEntee, Rachel
Citations
Altmetric:
License
Attribution-NoDerivatives 4.0 International
Abstract
Background: University of Vermont Health Network (UVMHN) participates in the AMA Joy in
Medicine Health System Recognition Program. One category of metrics is “Efficiency of Practice.”
UVMHN implemented Abridge Ambient AI scribing to reduce documentation burden in early 2025.
Objective: To evaluate changes in physician EHR use and work burden following early ambient AI
implementation by comparing Epic Signal metrics across Family Medicine (FM), General Internal
Medicine (GIM), Primary Care Pediatrics, and Rheumatology.
Methods: We conducted a quality improvement analysis of Epic Signal audit-log data comparing a
pre-implementation baseline (January 2025) with an early post-implementation period (June 2025).
Metrics included Total EHR time (EHR8), Outside Work time (WOW8), Inbox time (IB-Time8), and
Note documentation time (Note-Time8), normalized to 8 hours of scheduled patient time.
Specialty-level results were aggregated; no identifiable individual data were retained.
Results: Effects varied by specialty. FM showed marked EHR8 reduction (−2.23 h; −38.9%) and
IB-Time8 decrease (−6.9%), with increased Note-Time8 (+21.4%) and stable WOW8 (+3.1%). GIM
remained stable for EHR8 (+0.1%), with slight IB-Time8 reduction (−4.4%) and increases in
Note-Time8 (+15.2%) and WOW8 (+6.1%). Pediatrics showed notable WOW8 improvement (−0.31
h; −36.9%) but small increases in EHR8 (+3.4%) and IB-Time8 (+16.7%), with decreased
Note-Time8 (−5.1%). Rheumatology was stable for EHR8 (−0.2%) and WOW8 (+1.2%), with
reduced IB-Time8 (−12.1%) and slight Note-Time8 increase (+2.5%).
Discussion: 16 statistical analyses were run using independent sample t-test (Welch’s) and effect
size (Cohen’s d) and adjusted with Holm-Bonferroni to screen for false positives. Early trends
suggest specialty-specific impacts, potentially reflecting adoption curves and workflow variation.
Results are unadjusted for case mix or visit type and limited by small sample sizes (<100). More
physicians were audited in June for FM and GIM. Differences may reflect who was measured and
not behavior changes. Overall, FM reduction in EHR8 was statistically significant, while Pediatrics
WOW8 was not significant post-adjustment.
Conclusions/Future Directions: Expansion of Abridge Ambient AI could benefit from targeted
optimization (e.g., inbox triage, training) and continued monitoring of Signal data to ensure tools are
aligned to specific departmental needs.
References:
- 2025 Project Guidelines. Sept. 2024,
www.ama-assn.org/system/files/joy-in-medicine-guidelines.pdf.
- Sinsky CA, Rule A, Cohen G, et al. Metrics for assessing physician activity using electronic health
record log data. Journal of the American Medical Informatics Association. 2020.
Description
Date
2025-08-15