Video-based Project to Improve Colonoscopy Quality in Lower-Volume Clinicians

Colonoscopy reduces the risk of colorectal cancer (CRC) mortality through the detection and removal of pre-cancerous polyps (adenomas), but the magnitude of risk reduction depends upon the quality of the colonoscopy. Most colonoscopy quality improvement (QI) initiatives have focused on higher-volume gastroenterologists. In rural and peri-urban communities in the United States, most colonoscopies are performed by lower-volume colonoscopists, who are mostly surgeons, and CRC outcomes are worse compared to urban areas. These lower-volume colonoscopists have marked variability in quality and many fail to adhere to best practices or meet national benchmarks. Higher adenoma detection rate (ADR) is the quality metric most associated with reduced CRC mortality. However, ADR is less accurate at lower colonoscopy volumes, suggesting that ADR alone cannot reliably differentiate quality or provide meaningful information to improve quality in lower-volume colonoscopists. Thus, alternative methods to measure and improve quality in lower-volume colonoscopists are needed. Video evaluation of procedures such as colonoscopy is a valuable tool for assessing and improving technical skills in multiple specialties. In prior work, we found that higher-volume gastroenterologists' colonoscopy skill, assessed by expert evaluation of video-recorded colonoscopies using a validated technical skill scoring rubric, is highly correlated with quality metrics requires fewer colonoscopies for accurate assessment, and correlates with ADR. Further, we showed that delivery of focused feedback based on technical skill scores improves colonoscopy quality. While expert video evaluation of colonoscopies is an effective tool to measure and improve quality in lower-volume colonoscopists, it is not feasible on a larger scale given the time and cost of manual video review. We therefore developed and validated an AI-based tool that identifies key steps in colonoscopy videos, for rapid expert review and scoring. This study focuses on the implementation and evaluation of an AI-augmented intervention to improve colonoscopy quality of lower-volume colonoscopists in rural and peri-urban areas to ultimately reduce CRC mortality. The intervention, the "Colonoscopy Quality Improvement Bundle" (ColonQI Bundle), combines (1) AI-augmented expert assessment and scoring of technical skills; (2) AI-augmented automated calculation of quality metrics (e.g., ADR, withdrawal time), and (3) Targeted feedback of specific technical skill deficits based on the scores (audit-feedback, structured video-based didactics, coaching, mentoring). Lower-volume colonoscopists, with a focus on surgeons will be recruited to: (1) Assess the needs and preferences of, and select strategies for implementation of the ColonQI Bundle using an implementation science framework; (2) Implement the ColonQI Bundle, using tailored implementation strategies tailored to the needs and preferences of lower-volume colonoscopists; and (3) Assess the effectiveness of the ColonQI Bundle to improve colonoscopy quality and technical skills among lower-volume colonoscopists.