S537 ANNIE to the Rescue: Improving Bowel Prep Quality Through Integration of AI Generated Reminders at a Large Volume VA Center

Introduction: Colonoscopy remains the gold standard modality for colon cancer screening, yet achieving adequate bowel prep quality rates remains a limitation. Accessible and timely instructions are key to good quality preparations which are a standard metric for endoscopy units. Our primary objective was to improve a single center bowel prep quality rate through optional enrollment into an artificial intelligence (AI) ‘ANNIE’ text reminder program. Reminders were tailored towards the timing and date of a patient’s colonoscopy. Methods: All veterans were initially contacted by GI nurses before their procedure with a choice to opt in for ANNIE text instructions or traditional paper instructions. We divided each year into 4 quarters (Q) with Q1 October-December, Q2 January-March, Q3 April-June, and Q4 July-September. Our study spanned from Q1 2024 to Q1 2025. Enrollment in ANNIE started Q3 2024. A random subset of veterans completed a survey assessing their perceptions of pre-procedure instructions in facilitating better bowel prep and receptiveness to the future use of the same modality of instructions. Results: A total of 3,182 veterans from April to December 2024 agreed to receive ANNIE reminders. A total of 4,685 colonoscopies were completed from Q1 2024-Q1 2025. Prior to ANNIE implementation, inadequate bowel prep was 9.82% in Q1 2024 and 10.04% in Q2 2024. Over 3 cycles of implementing ANNIE text reminders, inadequate bowel prep decreased from 8.92% in Q3 2024 to 8.16% in Q4 2024 to 6.39% in Q1 2025. A 2-proportion z-test comparing bowel prep quality before and after implementation of the new instructions revealed a statistically significant reduction in the rate of poor bowel prep (P < 0.05). This indicates that the intervention was effective in improving bowel prep quality. A subset of 25 veterans completed a survey assessing ease of enrollment and ability to follow prep instructions. Eleven veterans received traditional paper instructions, and fourteen veterans received ANNIE text instructions. Both groups had a median Boston Bowel Preparation Scale of 8, although compliance with instructions was higher in the ANNIE subgroup and more were receptive to using ANNIE texts for repeat colonoscopy. Conclusion: Ultimately, we identified a need to improve bowel prep quality at our center with the opportunity to enroll patients in AI driven text reminders specific to their procedure date and time. The ease of enrollment, clarity of instructions, and prompt delivery of texts significantly improved the quality of bowel prep.