Artificial intelligence compared with traditional methods of generating dermatology consultation letters: a pilot...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Dermatology Clinical Documentation

Background:

• Traditional clinician-secretary methods for generating dermatology clinic letters are time-consuming and can lack accuracy.
• Emerging artificial intelligence (AI) platforms offer potential for improving efficiency and accuracy in clinical documentation.

Purpose of the Study:

• To compare the accuracy, readability, and efficiency of AI-generated (ChatGPT, Heidi) versus traditional dermatology clinic letters.
• To evaluate AI platforms in a simulated clinical setting against established clinician-secretary methods.

Main Methods:

• Four scripted dermatological cases were used to generate clinic letters via traditional dictation/transcription, ChatGPT, and Heidi.
• Letters were assessed by senior dermatologists for accuracy, relevance, satisfaction, readability (Flesch Reading Ease), and time efficiency.

Main Results:

• AI platforms demonstrated superior accuracy and time efficiency over traditional methods.
• ChatGPT yielded highest accuracy (92.6%) and satisfaction (4.59/5) but had lower readability; Heidi (85% accuracy, 4.34/5 satisfaction) offered structured, efficient output (27 seconds).
• Traditional methods had lowest accuracy (58.81%) and satisfaction (3.52/5) but highest readability (12-15 years).

Conclusions:

• AI platforms significantly enhance accuracy and efficiency in dermatology clinic letter generation.
• Heidi presents a potentially more suitable option for clinical integration due to consistent performance and reliability.
• Further research is needed to address AI readability, data security, and real-world clinical workflow implications.