Eliminating Nonvalue-Added Tasks in Health Care: Artificial Intelligence in the Orthopedic Office

Read full text Summarise with VIRAT
Authors: Thomas Doss, Jomar N. A. Aryee, Diana Vitkovska, Aryan Borole, Connor J. O’Leary, Marissa Viqueira, Daniel Okoh, Brian M. Katt
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Purpose: This study examines the efficiency and proof-of-concept use of a generative artificial intelligence (AI) documentation prototype, evaluating whether it can produce usable notes and estimating its potential time advantage compared with physician-produced documentation in an orthopedic practice. Methods: A comparative study analyzed AI-generated documentation versus physician-generated documentation of 68 new patient encounters with a board-certified hand surgeon, including surgical and nonsurgical patients presenting for evaluation of upper-extremity conditions. AI-generated documentation was created using unedited audiorecordings of encounters containing no identifiable patient information that were processed by standardized ChatGPT prompts. A senior resident (PGY-5) categorized all visits as minimal, low, or moderate complexity using the Centers for Medicare & Medicaid Services guidelines. Variables measured were documentation time (AI and physician), and physician time to edit AI documentation, visit complexity, visit time of day, patient demographics, and a physician-produced note quality assessment using a 5-point Likert scale. Results: Editing AI-generated documentation required a median of 55.1 seconds, compared with 81.2 seconds for physician-generated documentation. AI-generated notes were generally high quality, with 91% rated 4 or 5 on a 5-point Likert scale. Physician documentation time increased with greater visit complexity, whereas AI-documentation time showed minimal variation across complexity levels; complexity-stratified findings should be interpreted as exploratory because of sample size. Critical errors resulting in incomprehensible or erroneous content were identified in six (8.8%) AI-generated documents. Conclusions: This proof-of-concept study demonstrates that AI can produce clinically usable notes across varying encounter complexities and that physician editing of AI-generated documentation is faster than independent documentation. Although human oversight remains essential, AI-assisted documentation has substantial potential to improve workflow efficiency. Level of evidence: IV.

Fields of Study & Subjects

Artificial Intelligence in Healthcare and Education Electronic Health Records Systems Nursing Diagnosis and Documentation

Publication details

Journal / source
Journal of Hand Surgery Global Online
Volume
8
Issue
3
Pages
100992–100992
Publication date
31 March 2026
Language
EN