Purpose

Ptosis is a common oculoplastic condition with surgical management ranging from routine to complex. Patients with Ptosis frequently seek information online prior to and following consultation. This study evaluated the quality and readability of responses generated by leading AI LLMs to frequently asked ptosis patient questions, using validated assessment instruments.

---

Methods

Twelve representative patient questions on ptosis were identified from published sources and clinical practice, covering aetiology, functional impact, surgical options, and postoperative expectations. Questions were submitted to GPT-4o, Gemini 1.5 Pro, and Claude 3.5 Sonnet. All responses were scored using the DISCERN instrument and assessed for readability via Flesch-Kincaid Grade Level. Two raters independently scored responses; inter-rater agreement was quantified using intraclass correlation coefficient (ICC).

---

Results

Results pending at time of submission. Analysis will compare overall and subscale DISCERN scores across models, alongside readability metrics. Statistical comparisons will use Kruskal-Wallis testing. Preliminary scoring is in progress.

---

Conclusion

As AI chatbot use grows among surgical patients, understanding the reliability of AI-generated information for eyelid conditions is of increasing clinical relevance. This study provides a methodologically rigorous, validated assessment of AI performance in the ptosis context. Results will highlight platform-specific strengths and deficiencies, supporting evidence-based recommendations for patients and clinicians.

---

Additional Authors

First name	Last name	Base Hospital / Institution
Murtaza	Sheikh	South Tees NHS Trust

↑ Back to top