View as “PDF version”
Patient Safety Tip of the Week
Can AI Prevent
Ophthalmological Surgery Errors?
Ophthalmological surgery is generally quite
safe. But occasionally things go wrong. We’ve seen cases where a procedure is performed
on the wrong patient or wrong eye and cases where the wrong intraocular lenses
(IOL’s) are implanted. In fact, it was after a case in the mid-1990’s in which
2 consecutive patients received the wrong intraocular lenses that we actually developed one of the first surgical timeout
protocols. New York State a year later adopted the
protocol almost verbatim and Joint Commission’s Universal Protocol followed
shortly thereafter with most of the same features.
Tabuchi
et al. recently evaluated an AI-based system for potential patient and surgical
material verification system in ophthalmology (Tabuchi 2024). Their AI-based surgery safety system
integrates three main components: facial recognition, surgical laterality verification
and IOL authentication. It uses data from the electronic medical record plus
the patient’s facial photograph for authentication.
Using
an iPad mini, a dedicated circulating nurse manages the AI authentication
process, carrying it throughout the authentication sequence: facial recognition
at the entrance, followed by surgical laterality and IOL authentication inside
the operating room. The user interface is intuitive, with clear instructions and
immediate feedback. Each authentication is initiated with a single tap, and the
system provides real-time guidance for optimal image capture. Devices connect
to the hospital network via secure Wi-Fi, enabling immediate data synchronization
and verification results. The system does not replace the WHO Surgical Safety
Checklist, but rather works in conjunction with it.
They
analyzed 18,767 cases in the pre-implementation period and 18,762 in the post-implementation
period. There were more actual errors in the post-implementation period (5
errors post-implementation vs. 1 pre-implementation). However, far more
near-misses were identified in the post-implementation period (30 near misses post-implementation vs. 9 pre-implementation). Of
those 30 near-misses, there were four cases of incorrect left-right drape
placement and 26 cases of IOL preparation mistakes.
The authors
attribute the discrepancy between pre-implementation and post-implementation
error reporting (1 vs 5 cases) to improved detection rather than an actual
increase in errors.
There
were implementation challenges and also some authentication
failures due to substandard image quality. The authors also did an economic
analysis and found that the AI-based system offers significant economic advantages.
The
authors note that an initial learning curve and resistance highlight the
importance of thorough training and change management strategies when adopting
any new technologies. Having good systems does not guarantee that errors will
be avoided. For example, they noted a case where a physician disregarded both the
timeout procedure and the AI.
Though
this was a single-center study, the authors note that a smaller validation
study at another hospital showed similar authentication rates and perfect post-authentication
accuracy, supporting their findings. This is a promising technology and system.
It will be interesting to see if results can be replicated at other sites.
Given the high volume of ophthalmological surgery cases in the world, an
easy-to-use and economically viable intervention like this could make ophthalmological
surgery even safer.
Some
of our previous patient safety columns involving ophthalmology issues:
June
5, 2007 “Patient
Safety in Ambulatory Surgery”
March 11, 2008 “Lessons
from Ophthalmology”
June
8, 2010 “Surgical
Safety Checklist for Cataract Surgery”
June 2012 “Tailored Timeouts for Ophthalmologists”
May 20, 2014 “Ophthalmology: Blue Dye Mixup”
September 2014 “Another Blue Dye Eye Mixup”
May 17, 2016
“Patient
Safety Issues in Cataract Surgery”
December 5, 2017 “Massachusetts
Initiative on Cataract Surgery”
September 14, 2021 “Wrong Eye Injections”
January 25, 2022 “More on Dental Patient
Safety Issues”
References:
Tabuchi
H, Ishitobi N, Deguchi H, et al. Large-scale
observational study of AI-based patient and surgical material verification
system in ophthalmology: real-world evaluation in 37,529 cases. BMJ Quality
& Safety 2024; Published Online First: 29 November 2024
https://qualitysafety.bmj.com/content/early/2024/11/29/bmjqs-2024-018018
Print “PDF version”
http://www.patientsafetysolutions.com/