Abstract
BACKGROUND: The quality of medical records is crucial for effective patient care but is often compromised by the time doctors spend typing during and after consultations, contributing to physician burnout. Voa is an AI-driven tool developed in Brazil to convert audio from medical consultations into optimized clinical documents. This study examines the implementation of Voa in the clinical environment, employing technologies like Whisper and generative AI to enhance documentation quality, reduce medical errors, and improve doctor-patient interactions.
METHODS: The study involved metrics such as the number of anamneses generated and user adoption rates from March to May 2024, related to healthcare professionals who used Voa during their routine consultations. System architecture involves real-time data capture, speech-to-text conversion by Whisper, and refinement of text through a GPT-4-based Large Language Model (LLM).
RESULTS: The study observed an increase in document generation and user adoption over the analysis period. The cumulative number of documents generated reached 6,380 by mid-May 2024. The number of users grew steadily from approximately 100 in early March to nearly 900 by mid-May. The rolling average of daily document generation indicated consistent growth, with noticeable peaks and seasonality patterns. The variation in the weekly activation rate suggests many doctors registered but did not use the platform effectively, indicating a need for improved user retention through targeted onboarding, training, and support.
CONCLUSIONS: Metrics analysis showed an increase in document generation and users, reflecting growing acceptance. As Voa evolves, its adoption is expected to improve operational efficiency and patient care quality. Continuous improvements and user feedback mechanisms are expected to further increase its acceptance and integration into clinical workflows. Physicians who do not adopt such technologies may find themselves at a significant disadvantage in meeting the increasing demands of modern, data-driven healthcare systems.
Reference20 articles.
1. Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022). Robust speech recognition via large-scale weak supervision. Retrieved from https://cdn.openai.com/papers/whisper.pdf
2. - Hong G, Wilcox L, Sattler A, Thomas S, Gonzalez N, Smith M, Hernandez J, Smith M, Lin S, Harrington R. (2020). Clinicians’ Experiences with EHR Documentation and Attitudes Toward AI-Assisted Documentation [White paper]. Retrieved from Stanford University School of Medicine and Google Health: med.stanford.edu/healthcare-ai and health.google.
3. - Sinsky C, Colligan L, Li L, Prgomet M, Reynolds S, Goeders L, Westbrook J, Tutty M, Blike G. Allocation of Physician Time in Ambulatory Practice: A Time and Motion Study in 4 Specialties. Ann Intern Med. 2016 Dec 6;165(11):753-760. doi: 10.7326/M16-0961. Epub 2016 Sep 6. PMID: 27595430.
4. Reimagining Clinical Documentation With Artificial Intelligence
5. A Time-Motion Study of Primary Care Physicians’ Work in the Electronic Health Record Era