A large language model artificial intelligence for patient queries in atopic dermatitis
Journal
Journal of the European Academy of Dermatology and Venereology : JEADV
Date Issued
2024-01-03
Author(s)
Sulejmani, Pranvera
Negris, Olivia
Aoki, Valeria
Eichenfield, Lawrence
Misery, Laurent
Mosca, Ana
Orfali, Raquel Leão
Aroman, Markéta Saint
Stalder, Jean-Francois
Trzeciak, Magdalena
Wollenberg, Andreas
Lio, Peter
Abstract
Dear Editor, Atopic dermatitis (AD) is the most common chronic inflammatory skin condition with a staggering global burden of disease. Despite advances in treatment, patients with AD suffer from lower quality of life and poor treatment satisfaction, with over half of self-identified moderate-to-severe AD patients reporting inadequate disease control.1 Management of AD requires participation by the patient and physician and is increasingly complex and multifaceted. Quality therapeutic patient education, regardless of its source, can improve treatment adherence and patient quality of life.2 Patients rely on physicians for guidance and often utilize electronic communications for answers to medical questions. Though convenient, the number of messages physicians receive continues to rise towards unsustainable levels.3 Large language models like Chat-Generative Pre-Trained Transformer (ChatGPT) have become a novel resource for patients and caregivers. As a form of artificial intelligence (AI), ChatGPT serves as a conversational chatbot trained in a broad range of internet sources, with the ability to use reinforcement learning from human user feedback.4 A recent study evaluated the ability of ChatGPT to provide quality and empathetic responses to patient questions posted to a social media forum as compared to physician responses.5 It found that the ChatGPT responses were rated significantly higher in both quality and empathy compared to physician responses.5 These findings suggest there may be utility for AI in drafting electronic communications to patients, and perhaps more importantly, some reassurance about what to expect when patients inevitably ask their questions directly to such AI. In an open study evaluating the reliability of ChatGPT responses, we submitted 99 common questions from AD patients to the most currently available model, ChatGPT-4. The questions were supplied by an international group of 11 dermatologists in which each dermatologist contributed up to 10 questions that their patients commonly asked. The questions included inquiries regarding the aetiology, management and consequences of AD. The ChatGPT responses were independently evaluated by a group of international dermatologists experts in therapeutic patient education.2 The questions were input into ChatGPT-4 in English and the responses were also evaluated in English. The overall quality and reliability of ChatGPT's responses were assessed using the Likert scale from 1 to 10, 10 being best, and were flagged for harmful information. The graded responses were averaged and ranged from 8.18 to 10, indicating high levels of evaluator satisfaction. Standard deviations ranged from 0 to 1.76. See Table 1 for a list of representative questions and responses generated by ChatGPT-4. Figure 1 demonstrates the graded responses per dermatologist. There was overwhelming reliability of the responses provided by ChatGPT to the wide-ranging questions frequently asked by patients. Importantly, most responses acknowledged that it was not a physician and directed users to a healthcare provider. Though no responses were flagged for dangerous information, lower scores were reported for incomplete answers or where advice deviated from evidence-based medicine. In this paper, we make an initial measurement of the quality and reliability of the answers generated by AI to some frequently asked patient questions. ChatGPT-4 provided thorough, high-quality responses without any special priming or additional information. Given the increasing demands on physicians, AI is poised to become a valuable resource. Patients are likely already utilizing AI for medical information—or soon will be—and knowledge about the quality of responses from different Large Language Models may help inform clinicians and patients alike. Living with eczema can sometimes lead to feelings of isolation, self-consciousness or embarrassment, but there are ways to cope with these challenges and improve your quality of life: Connect with others: seek out support groups or online communities where people share their experiences with eczema. Connecting with others who understand your situation can help you feel less alone and provide valuable insights into managing the condition. Educate yourself and others: learn as much as you can about eczema and share your knowledge with friends and family. This can help them understand your situation better and reduce any misconceptions or stigmas associated with the condition. Maintain open communication: Talk openly with your loved ones about your feelings and experiences. Let them know how they can support you and what you need from them. Focus on self-care: Take care of your mental and emotional well-being by engaging in activities that bring you joy and relaxation. This could include hobbies, exercise or spending time with loved ones. Seek professional help: If your feelings of isolation or other emotional challenges become overwhelming, consider talking to a mental health professional, such as a therapist or counsellor, who can help you develop coping strategies and build resilience. Set realistic expectations: Understand that managing eczema is an ongoing process, and there may be ups and downs along the way. Accepting this reality can help you maintain a balanced perspective and reduce feelings of isolation. Focus on your strengths and accomplishments: Acknowledge your achievements and positive qualities and remind yourself that you are more than your eczema. Build a support network: surround yourself with supportive and understanding people, including friends, family and healthcare professionals, who can provide encouragement and assistance. Advocate for yourself: do not be afraid to speak up about your needs and ask for accommodations when necessary, whether it is at work, school or in social settings None. VA has been an investigator and/or consultant to Abbvie, Eli Lilly and Pfizer. LE has served as a scientific adviser, consultant and/or clinical study investigator for AbbVie, Amgen, Aslan, Arcutis, Dermavant, Eli Lilly, Forté, Galderma, Ichnos, Incyte, LEO Pharma, Novartis, Ortho Dermatologics, Pfizer, Regeneron and Sanofi Genzyme. LM Pfizer, Novartis, Sanofi, Abbvie, Lilly, Galderma, Amgen, Incyte, Leo Pharma, Dermira and AstraZeneca. RLO has been an investigator and/or consultant for Bayer, Eli Lilly, Abbvie, Sanofi and Amgen. MSA is employed by Pierre Fabre. MT has been a speaker and/or consultant and/or investigator and/or participant of the Advisory Board for Abbvie, Bausch Health, Bioderma, Eli Lilly, La Roche posay, Leo pharma, Mead Johnson, Novartis, Pfizer, Pierre Fabre, Pfizer, Mead Johnson and Sanofi Genzyme. PL reported receiving grants from AOBiome, Regeneron/Sanofi Genzyme and AbbVie; personal fees from Regeneron/Sanofi Genzyme, Leo, Eli Lilly, Pfizer, Galderma, L'Oreal, Almirall, ASLAN Pharma Advisory board, Dermavant, Pierre Fabre, Menlo Therapeutics, IntraDerm, Exeltis, AOBiome, Arbonne and Amyris; stock options from Micreos and other royalties from patented product from Theraplex. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. There are limitations to the conclusion in that the answers may differ with differences in the formulation of questions, over time or with different versions of AI chat systems. Not applicable. The data that support the findings of this study are available from the corresponding author upon reasonable request.
SDGs
Type
letter