Understanding Clinician Perceptions of GenAI: A Mixed Methods Analysis of Clinical Documentation Tasks

Area of Science:

• Medical Informatics
• Human-Computer Interaction
• Artificial Intelligence in Healthcare

Background:

• Generative AI (GenAI) offers potential for improving clinical documentation within Electronic Health Records (EHRs).
• Understanding clinician user experience (UX) is crucial for successful GenAI integration into healthcare workflows.

Purpose of the Study:

• To evaluate clinician UX with GenAI for EHR documentation tasks (Information Extraction, Summarization, Speech-to-Text).
• To assess preferred automation levels, workflow improvements, safety perceptions, and overall EHR satisfaction.
• To identify key themes influencing clinician acceptance of GenAI in EHR systems.

Main Methods:

• A mixed-methods study using conceptual prototyping and a usability framework.
• 38 clinicians interacted with a GenAI-integrated EHR prototype across varying automation levels.
• Quantitative (questionnaires, statistical tests) and qualitative (thematic analysis) data were collected and analyzed.

Main Results:

• Clinicians responded positively to GenAI, anticipating workflow streamlining.
• Medium automation was the preferred level, perceived as "safe with caution".
• Key themes included efficiency gains, reliability concerns, safety, medico-legal issues, automation bias, and the need for adjustable settings and oversight.

Conclusions:

• Clinicians welcome GenAI for documentation but favor moderate automation for a balance of efficiency and control.
• Successful integration necessitates addressing safety, conducting real-world trials, and mitigating bias and medico-legal risks.
• A cautious yet optimistic approach is recommended, prioritizing clinician oversight alongside AI benefits.