Effect of ChatGPT-Assisted Pre-Procedural Education on Bowel Preparation Quality and Anxiety in First-Time Colonoscopy Patients

Bowel Preparation Quality, Colonoscopy, Patient Education, Preoperative Anxiety

ChatGPT, Patient education, Bowel preparation, Anxiety, Colonoscopy, Artificial intelligence

Background and rationale: Colonoscopy is central to colorectal cancer screening and prevention, but its diagnostic yield depends heavily on adequate bowel preparation; inadequate cleansing lowers adenoma detection and often necessitates repeat procedures. Pre-procedural anxiety is also common, particularly among patients undergoing colonoscopy for the first time. Conventional education is constrained by time and does not always address individual questions. Large language model-based tools such as ChatGPT can generate individualized, real-time responses and may serve as an adjunct to standard counseling. Objective: To evaluate whether ChatGPT-assisted pre-procedural patient education, added to standard information, improves bowel preparation quality and reduces pre-procedural anxiety compared with standard information alone in patients undergoing their first colonoscopy. Design and setting: Single-center, parallel-group, two-arm randomized controlled trial conducted in the Department of Gastroenterology, Gazi University Faculty of Medicine, Ankara, Türkiye. The study was approved by the Gazi University Clinical Research Ethics Committee (approval no. 2025-1016). Written informed consent was obtained from all participants. Participants: Adults (aged ≥18 years) presenting for their first colonoscopy were eligible. \[Buraya kalan dışlama kriterlerini ekle - ör. önceki kolonoskopi öyküsü, tablet kullanamama, ciddi bilişsel/işitsel engel vb.\] Randomization and allocation: Eligible participants were randomly assigned in a 1:1 ratio to the standard-information (control) arm or the ChatGPT-assisted (intervention) arm, with allocation concealment. Interventions: Control arm: standard pre-procedural verbal and written bowel-preparation instructions per routine clinical practice. Intervention arm: the same standard instructions plus supervised access to a ChatGPT interface (GPT-4o model, ChatGPT Plus) on a clinic-provided tablet before the procedure. Participants could ask questions in their preferred language, up to a maximum of ten. Chatbot responses were reviewed by the study team for clinical appropriateness. Outcome measures: Primary: (1) bowel preparation quality assessed with the Boston Bowel Preparation Scale (BBPS); (2) pre-procedural anxiety assessed with the State-Trait Anxiety Inventory (STAI). Secondary: procedural tolerance (5-point Likert scale); adenoma detection rate; and completion of colonoscopy without sedation. Sample size: Based on an a priori power analysis assuming a moderate effect size (Cohen's d ≈ 0.50) for the primary outcomes, with two-sided alpha of 0.05 and 80% power. Statistical analysis: Outcomes were analyzed following the intention-to-treat principle. Between-group comparisons used t-test or Mann-Whitney U for continuous variables and chi-square/Fisher exact tests for categorical variables; multivariable logistic regression was used for binary outcomes.

Colorectal cancer screening by colonoscopy is effective only when bowel preparation is adequate; inadequate cleansing reduces adenoma detection and often necessitates repeat procedures. Pre-procedural anxiety is common, particularly before a first colonoscopy. Routine counseling is frequently constrained by time and cannot always address each patient's individual questions. Large language model-based conversational tools such as ChatGPT (GPT-4o) can generate individualized, real-time responses and may complement standard physician-provided counseling. This trial was designed to test that hypothesis in routine clinical practice. In the intervention arm, ChatGPT access was provided under supervision on a clinic-provided tablet immediately before the procedure, so that participants without a personal device or subscription could use it on equal terms. Participants asked their questions in their preferred language to reflect real-world use. All chatbot responses were reviewed by the study team for clinical appropriateness, and the tool was positioned as an adjunct to, rather than a replacement for, standard pre-procedural education.

No related papers found yet.

Turkey (Türkiye)