Page 86 - Read Online

Brochu et al. Art Int Surg 2024;4:411-26 https://dx.doi.org/10.20517/ais.2024.61 Page 413

breast augmentation. All concluded that ChatGPT was consistently correct, comprehensive, and well-
organized. However, it lacked the ability to provide personalized advice [12-14] . Another study using vignette-
style questions, giving a description of a patient before asking a question, found ChatGPT to outperform
[9]
physicians in accuracy, completeness, and overall quality based on ratings from physicians . This suggests
that ChatGPT can provide more individualized advice when given general descriptors of a patient, but it
cannot provide this information without a prompt. Understanding the level of information required to elicit
specific responses from ChatGPT is important when advising patients on using it for education.

Depending on the results of further research, ChatGPT could either supplement patient education and
improve efficiency in initial consultations with surgeons, or could replace the majority of patient education
in these consultations. While previous studies have focused on specific consultations for procedures such as
rhinoplasties or breast augmentations, limited literature exists comparing ChatGPT’s responses across
various plastic surgery procedures. To address this gap, this study aimed to investigate the ability of
ChatGPT to answer common patient questions related to the five most common aesthetic plastic surgeries.
The main objectives of this study were to determine the comprehensiveness, accuracy, and
understandability of ChatGPT’s responses to common patient questions to evaluate its ability to be a source
of patient education. Based on previous literature, we believed ChatGPT’s responses would be organized,
understandable, and generally accurate, with only minor potential inaccuracies. However, they might not be
entirely comprehensive or specific. We employed ChatGPT-3.5 to evaluate its capacity, accuracy,
comprehensiveness, and efficacy in providing perioperative responses to patients.

METHODS
We asked ChatGPT three questions related to the five most common aesthetic procedures as determined by
[15]
the 2022 American Society of Plastic Surgery (ASPS) Procedural Statistics Release . The first two questions
were the same for all five procedures and were based on suggested questions from the same report. These
two questions were chosen because they were applicable to all five procedures, were not specific to patient
or physician, and provided a relatively complete picture of the surgery and its risk profile. The last question
was procedure-specific and was chosen based on common complications and procedure specificity. These
questions were asked at the same time on the same account in the order they are presented below. The
questions were pasted directly from a Word document, where they were assessed for grammatical and
syntactical errors. ChatGPT’s responses were then compared to corresponding blogs or articles on the ASPS
website and assessed for accuracy and comprehensiveness. For comprehensiveness, if ChatGPT’s response
covered the key points of the corresponding ASPS article, it was considered comprehensive. If it omitted
information, it was considered less comprehensive, and if it included information not listed, it was
considered more comprehensive. Comprehensiveness was not assessed when ChatGPT provided a
completely different answer from the ASPS article. The information in ChatGPT’s response was considered
accurate if it was comparable to the information in the ASPS article.

RESULTS
Liposuction
What are the risks and complications associated with liposuction?
This response begins with an explanation of what liposuction is and a description of the procedure’s general
safety. It then lists 12 possible complications [Figure 1] with a sentence or two of explanation. The response
ends with a statement about the importance of following up with a qualified surgeon to discuss the
mitigation of these risks.