- The adenoma detection rate (ADR) in more than 1,400 non-AI assisted colonoscopies dropped 6 percentage points, from 28.4% to 22.4%, after endoscopists routinely used the technology in their practice.
- AI-assisted colonoscopy was not significantly associated with a higher ADR when compared to the detection rates before the technology was introduced.
- The findings raise concerns about "deskilling" in non-AI colonoscopies performed by endoscopists who routinely use the technology.
Endoscopists who got used to using artificial intelligence (AI) during colonoscopies detected significantly fewer precancerous growths when the technology was taken away, a study from Poland found.
In the observational study of more than 1,400 non-AI-assisted colonoscopies, the adenoma detection rate (ADR) dropped from 28.4% in the 3 months before AI was introduced into routine practice to 22.4% in the 3 months after (P=0.0089), reported Krzysztof Budzyn, MD, of the Medical University of Silesia in Katowice, Poland, and co-authors in The Lancet Gastroenterology & Hepatology.
The findings suggest that continuous exposure to AI may lead to "deskilling" when it is not being used, said co-author Marcin Romańczyk, MD, also of the Medical University of Silesia, likening it to drivers being over-reliant on GPS.
"Imagine that you want to travel anywhere, and you're unable to use Google Maps," Romańczyk, told MedPage Today. "We call it the Google Maps effect. We try to get somewhere, and it's impossible to use a regular map. It works very similarly."
While AI-assisted colonoscopy has been shown in randomized trials to boost ADR, the results of this study raise the possibility that the apparent benefit may partly reflect reduced performance without the technology.
Budzyn and colleagues noted that since AI implementation, the overall ADR rose from 22.4% to 25.3%, masking the decline in ADR for non-AI-assisted colonoscopy. AI-assisted colonoscopy was not significantly associated with higher ADR when compared to detection rates before its introduction (OR 0.80, 95% CI 0.63-1.02).
"We assume that continuous exposure to decision support systems such as AI might lead to the natural human tendency to over-rely on their recommendations, leading to clinicians becoming less motivated, less focused, and less responsible when making cognitive decisions without AI assistance," the authors wrote.
This could reflect weakened visual search habits, altered gaze patterns, dulling pattern recognition, and reduced diagnostic confidence, said Omer Ahmad, MBBS, PhD, of University College London, in an accompanying editorial.
He called for performance-monitoring guidelines from professional societies, educational programs to raise awareness of deskilling, and incorporating non-AI sessions into routine practice to keep skills sharp.
However, Ahmad told MedPage Today there are no easy solutions. "I do not think there is a quick way to figure this out," he said. "We need high quality studies, perhaps crossover in design, to study both behavioral metrics and outcome measures in AI and non-AI assisted colonoscopies. This would offer insights into how skill erosion may occur."
The multicenter study evaluated 1,443 non-AI-assisted colonoscopies (out of 2,177 total colonoscopies) performed by 19 experienced endoscopists at four medical centers in Poland between September 2021 and March 2022. Of those, 795 were conducted before the introduction of regular AI use and 648 were performed after AI was implemented.
The study found ADR varied by gender and age, being higher in men (OR 1.78 vs women, 95% CI 1.38-2.30) and individuals 60 and older (OR 3.60 vs <60 years, 95% CI 2.74-4.72).
The authors acknowledged several limitations, including the study's observational design, which carries a risk of selection bias, and the possibility of unmeasured confounding factors. Only one AI system was used, which means results may not apply to other systems or to less experienced endoscopists, who could be more susceptible to skill loss than the clinicians evaluated in the study who had performed at least 2,000 colonoscopies each.
Disclosures
The study was funded by the European Commission, the Japan Society for the Promotion of Science, and the Italian Association for Cancer Research.
Budzyn and Romańczyk declared no competing interests. Other co-authors reported relationships with Alfasigma, Boston Scientific, Casen Recordati, Cybernet System, Erbe, Fujifilm, Mayoly, Medtronic, Microtech, Norgine, Odin, Oslo University Hospital, and Olympus Europe.
Ahmad reported receiving medical consultancy fees from Olympus, Odin Vision, Medtronic, and Norgine.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.