Background: Large language models (LLMs) have a potential role in providing adequate patient information.
Objectives: To compare the quality of LLM responses with established Dutch patient information resources (PIRs) in answering patient questions regarding melanoma.
Methods: Responses from ChatGPT versions 3.5 and 4.0, Gemini, and three leading Dutch melanoma PIRs to 50 melanoma-specific questions were examined at baseline and for LLMs again after 8 months. Outcomes included (medical) accuracy, completeness, personalization, readability and, additionally, reproducibility for LLMs. Comparative analyses were performed within LLMs and PIRs using Friedman's Anova, and between best-performing LLMs and gold-standard (GS) PIRs using the Wilcoxon signed-rank test.
Results: Within LLMs, ChatGPT-3.5 demonstrated the highest accuracy (P = 0.009). Gemini performed best in completeness (P < 0.001), personalization (P = 0.007) and readability (P < 0.001). PIRs were consistent in accuracy and completeness, with the general practitioner's website excelling in personalization (P = 0.013) and readability (P < 0.001). The best-performing LLMs outperformed the GS-PIR on completeness and personalization, yet it was less accurate and less readable. Over time, response reproducibility decreased for all LLMs, showing variability across outcomes.
Conclusions: Although LLMs show potential in providing highly personalized and complete responses to patient questions regarding melanoma, improving and safeguarding accuracy, reproducibility and accessibility is crucial before they can replace or complement conventional PIRs.
Large language models (LLMs) are a type of artificial intelligence that can be used to assess large amounts of information, and could play a role in providing patients with accurate information about melanoma. This study, conducted in the Netherlands, aimed to compare the quality of responses from LLMs and established Dutch patient information resources (PIRs) to questions from patients about melanoma. We evaluated the responses of ChatGPT versions 3.5 and 4.0, Gemini, and three leading Dutch melanoma PIRs to 50 specific melanoma-related questions. We looked at these responses initially and asked the questions again to LLMs after 8 months. We assessed medical accuracy, completeness, personalization, readability and reproducibility for LLMs. We used statistical tests to compare results within LLMs, within PIRs, and between the best-performing LLMs and the gold-standard PIR (leaflet from dermatologists). We found that within the LLMs, ChatGPT-3.5 had the highest accuracy. Gemini performed best in terms of completeness, personalization and readability. The PIRs consistently showed high accuracy and completeness, with the general practitioner’s website excelling in personalization and readability. The top-performing LLMs outperformed the best PIR on completeness and personalization, but was less accurate and less readable. Over time, the reproducibility of LLM responses decreased, showing varied outcomes. In conclusion, although LLMs have the potential to provide highly personalized and complete answers to patient questions about melanoma, their accuracy, reproducibility and accessibility need to be improved before they can be a reliable source of information or complement existing PIRs.
© The Author(s) 2024. Published by Oxford University Press on behalf of British Association of Dermatologists.