Comparing ChatGPT Informed Consent Documentations for Energy-Based Devices.

Summary

Using standardized prompts, ChatGPT-4 generated consent forms for 8 energy-based devices that experts rated variably: radiofrequency microneedling scored highest, while monopolar RF scored lowest. Documentation of expected pain and post-procedure restrictions was stronger than general descriptions, and inaccuracies occurred, underscoring the need for dermatologist oversight.

Key Findings

• ChatGPT-4 consent forms showed variable quality across 8 devices; radiofrequency microneedling scored highest (2.75/3), monopolar RF lowest (1.85).
• Expected pain (2.68) and restrictions (2.5) were consistently well-documented; general descriptions (2.16) and overall impression (2.05) were least complete.
• Some content was incomplete or inaccurate, indicating that AI outputs require dermatologist oversight before clinical use.

Clinical Implications

Clinics can leverage AI to draft readable consent forms but must implement expert review, device-specific revisions, and quality control to prevent omissions and inaccuracies.

Limitations

• Small expert panel and limited to eight device categories; no patient comprehension outcomes.
• Single AI model/version and prompt set; generalizability to other models or clinical contexts is uncertain.

Future Directions

Prospective, randomized studies comparing AI-assisted vs. standard consents on patient comprehension, decisional conflict, and safety; development of validated, specialty-curated prompt libraries.