.png){kind=icon}
.png){kind=icon}
Return
Computer-Aided Colonoscopy Alert Fatigue and Its Effect on Adenoma Detection
.png){kind=icon}
PRE
AI
F
M
D
M
D
R
D
M
K
B
S
D
E
K
R
L
D
DOI：10.14309/ajg.0000000000003955.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
INTRODUCTION: Computer-aided detection (CADe) can improve colonoscopy adenoma detection rates (ADR). However, prolonged exposure to repeated alerts may lead to alert fatigue, potentially diminishing its effectiveness. We sought to determine whether alert fatigue occurs during CADe use and, if so, how it influences ADR throughout the day. METHODS: We conducted a secondary analysis of a prospectively collected cohort to evaluate how ADR varied with and without CADe in a pragmatic, real-world implementation. Elective colonoscopies performed at the University of Montreal Hospital Center were analyzed according to their position within the endoscopist's daily schedule. The primary outcome was ADR across procedure order, adjusted for withdrawal time (WT). Secondary outcomes included adenomas per colonoscopy, ADR by time of day, and ADR across different WT categories. RESULTS: A total of 1,700 colonoscopies were analyzed, including 880 with CADe assistance and 820 without. In the CADe-unassisted group, ADR remained stable throughout the day (1st-3rd procedures: 39.4%, 95% confidence intervals (CI) [35.1-43.8]; 4th-7th procedures: 40.9%, 95% CI [32.9-49.5]). By contrast, ADR in CADe-assisted procedures was markedly higher in the first 3 procedures (49.9%, 95% CI [45.5-54.3]) but dropped significantly to 39.9% (95% CI [33.8-46.3]) in later procedures. This pattern was consistent when ADR was assessed by time of day. Similarly, adenomas per colonoscopy were significantly higher in early CADe-assisted procedures (mean increase +0.27, 95% CI [0.14-0.41], P < 0.001), but this benefit disappeared in later procedures (+0.11, 95% CI [-0.09 to 0.31], P = 0.51). WT analysis further supported this effect: ADR gains per additional minute were higher in unassisted colonoscopies (21.6% early vs 15.3% late) than in CADe-assisted colonoscopies (10.5% early vs 10.1% late). DISCUSSION: CADe assistance significantly improved ADR during early-day colonoscopies, but this benefit disappeared in later-day procedures. This pattern suggests that alert fatigue may limit endoscopists' responsiveness to CADe prompts as the day progresses.
Keywords:
adenoma detection rate
alert fatigue
computer-aided detection
withdrawal time
Journal
IF:
7.6
Papers:
5.1W
Citations:
3.3W
Organization
Cited Papers