Large Language Models in Surgical Informed Consent for Reconstructive Plastic Surgery: A Systematic Scoping Review of Accuracy, Patient Comprehension, Medicolegal Risk, and Readiness for Clinical Deployment
Abstract: Introduction: Surgical informed consent is a high-stakes medicolegal process underpinning patient autonomy and surgical safety. Large language models (LLMs)- including GPT-4, Gemini, and Claude- are increasingly used by patients to research procedures, interpret consent documents, and formulate pre-operative questions, often without clinician awareness. No systematic synthesis has evaluated the accuracy, readability, patient comprehension, or medicolegal implications of LLM-generated surgical consent information in reconstructive plastic surgery. Methods: A PRISMA-ScR compliant systematic scoping review was conducted across MEDLINE, EMBASE, Cochrane Library, and PsycINFO (January 2020 - May 2026). Search terms included "large language model", "ChatGPT", "GPT-4", "generative AI", "surgical consent", "informed consent", "patient information", "plastic surgery", and "reconstructive surgery". Studies reporting quantitative outcomes related to accuracy, readability, patient comprehension, or medicolegal risk were included. Results: Forty-seven studies comprising 14,200 participants (patients and clinicians) met inclusion criteria. LLMs generated consent-relevant information at a mean Flesch-Kincaid reading grade level of 13.2- substantially above the recommended maximum of Grade 8 for health information. Factual accuracy for complex reconstructive procedures averaged 71-78% across GPT-4, Gemini, and Claude 3, with complication rates, individualised risk, and centre-specific outcomes most frequently inaccurate or absent. Patient-reported comprehension of LLM-generated consent information was significantly lower than clinician-generated content (mean 58% vs 81%). LLMs failed to disclose material risks in 34% of simulated consent scenarios. No model adequately addressed individualised surgical risk- the primary medicolegal standard under Montgomery v Lanarkshire (2015). Conclusion: LLMs currently provide consent-relevant information at an inaccessible reading level with incomplete factual accuracy, particularly for individualised risk- the standard against which surgical consent is legally judged in the United Kingdom. Reconstructive surgeons must be aware of LLM limitations, proactively address AI-sourced patient misinformation, and engage with the development of clinician-validated AI consent tools.
Keywords: Large language models, Artificial intelligence, Surgical informed consent, Reconstructive plastic surgery, Patient comprehension, Health literacy, Medicolegal risk, Montgomery v Lanarkshire
How to Cite?: Dr. Daniel Gravino, "Large Language Models in Surgical Informed Consent for Reconstructive Plastic Surgery: A Systematic Scoping Review of Accuracy, Patient Comprehension, Medicolegal Risk, and Readiness for Clinical Deployment", Volume 15 Issue 8, August 2026, International Journal of Science and Research (IJSR), Pages: 987-991, https://www.ijsr.net/getabstract.php?paperid=SR26807183311, DOI: https://dx.doi.org/10.21275/SR26807183311