'ChatGPT Consent Forms for Energy-Based Devices Show Potential, With Oversight'

 In a cross-sectional study, ChatGPT demonstrated variable effectiveness in generating informed consent forms for treatment with energy-based dermatologic devices, highlighting the need for expert oversight.

METHODOLOGY:

• A cross-sectional study was conducted to evaluate informed consent forms generated by ChatGPT-4 for the use of eight commonly used laser and energy-based dermatologic devices in January 2025.
• The devices were ablative fractional laser, nonablative fractional laser, laser hair removal device, pulsed dye laser (PDL), monopolar radiofrequency device, radiofrequency microneedling device, Q-switched laser, and picosecond laser.
• A standardized prompt requesting an informed consent document was used for each device, with separate ChatGPT sessions conducted in incognito browser windows to minimize confounding factors.
• Analyses included assessment of readability through Flesch-Kincaid grade-level calculations and calculation of average ratings for each domain and device; ratings were provided by five board-certified dermatologists from a private practice in Boston using a four-point rubric (0, inaccurate; 1, absent; 2, incomplete; 3, complete).

TAKEAWAY:

• The average reading level of the consent forms was 11.2, with the forms for PDL having the lowest level (score, 9.4) and the forms for radiofrequency microneedling having the highest level (score, 12.6). The average “reading ease” score was 35.43; the PDL forms were the easiest to read (score, 44.9), and those for radiofrequency microneedling were the hardest to read (score, 25.7).
• Over all categories, radiofrequency microneedling consent forms had the highest overall average rating of 2.75, scoring 2.5 or higher across 9 out of 10 domains, whereas monopolar radiofrequency device consent forms had the lowest average rating of 1.85, scoring 2 or lower across 6 out of 10 domains.
• The highest rating was reported for expected pain documentation across all devices (mean, 2.68), followed by expected restrictions (mean, 2.50), whereas the lowest scores were reported for overall impression (mean, 2.05) and general description of procedures (mean, 2.16).
• Scores for other domains, such as risks and descriptions of alternative treatments, varied across the devices; the average scores for benefits and general descriptions were 2.33 and 2.16, respectively, across all devices.

IN PRACTICE:

“ChatGPT demonstrates potential as a tool for generating informed consent documents for energy-based devices, particularly for procedures such as radiofrequency microneedling, where the majority of the consent form components were rated satisfactory,” the authors of the study wrote. “However, the presence of inaccuracies and omissions underscores the need for thorough review and editing by board-certified dermatologists,” they added, suggesting that “future research comparing AI [artificial intelligence]-generated and clinician-authored consent forms, particularly in terms of patient comprehension and satisfaction, would further clarify ChatGPT’s role in dermatologic documentation.”

SOURCE:

The study was led by Lori S. Kim, Department of Dermatology, Boston University School of Medicine, Boston, and was published online on December 3 in Lasers in Surgery and Medicine.

LIMITATIONS:

Potential bias may have been introduced because dermatologists were aware that they were evaluating AI-generated forms. The study was conducted at one private dermatology practice, limiting the generalizability of the findings. Additionally, the study focused solely on written content, excluding crucial aspects of verbal communication and shared decision-making during physician-patient interactions.

DISCLOSURES:

No specific funding was received for the study. The authors declared having no conflicts of interest.

https://www.medscape.com/viewarticle/chatgpt-consent-forms-energy-based-devices-show-potential-2025a1000z91