AI-powered smart glasses catch drug errors

Editor's Note

An AI-powered wearable camera has achieved 99.6% accuracy in detecting potentially deadly drug mix-ups, including in the OR, NBC News reported May 25. Developed by Kelly Michaelsen, MD, at UW Medicine, the smart glasses scan medication labels in real time, alerting anesthesia providers to syringe-vial mismatches before drugs are administered.

The article quotes John Wiederspan, a nurse anesthetist at UW Medicine in Seattle, describing how high-pressure moments in the OR can trigger errors when drugs are administered in haste. Despite systems like barcode scanning and color-coded labels, mistakes—especially vial swaps—remain common. Dr Michaelsen, an anesthesiologist and engineer, saw an opportunity to train AI to serve as a second set of eyes, noting that most medications actually used consist of a relatively a small set of 10 to 20 distinct drugs.

Michaelson’s team reportedly focused on vial swap errors, which account for about 20% of all medication errors. The device, worn like GoPro glasses, uses computer vision to read vial and syringe labels and trigger a warning if they don’t match. The system was trained on video of correct drug preparation and simulated error footage in lab settings. Michaelsen emphasized that eliminating false alarms was a top priority, given widespread concerns over OR alarm fatigue.

While awaiting FDA clearance, Michaelsen is exploring auditory alerts as the next step toward clinical use, NBC News reports. Wiederspan, who tested the device, supports its potential, although he notes the headset must become sleeker to gain wider acceptance.

Various leaders in anesthesiology and patient safety advocacy told NBC News they view this development as part of a broader push to integrate AI into error prevention. Dan Cole, MD, president of the Anesthesia Patient Safety Foundation, likened the technology’s potential to that of self-driving cars. Still, he and others, including grieving parent and advocate Melissa Sheldrick, warned that tech alone isn’t a panacea. Many errors stem from communication failures, fragmented data, or human factors that no AI can fully replace. Nonetheless, some argue that AI may be uniquely suited for fast-paced settings like the OR and ER, where traditional checks and cognitive vigilance often fall short.

Meanwhile, the UW team is reportedly expanding the device’s scope to monitor drug volumes, which could reduce pediatric dosing errors. Wiederspan sees future potential in adapting it for oral medications on hospital floors, where pill mix-ups are also a risk. Privacy concerns and the specter of intrusive monitoring remain live issues, but Michaelsen said the current model only reads syringe labels and stores no patient data.

The need for stringent oversight aside, Wiederspan concludes the article with a statement about AI’s immense potential for the OR. “Time is of the essence in an emergency situation where you’re trying to give blood, lifesaving medications, checking vital signs, and you’re trying to rush through these processes,” he told NBC News. “I think that’s where this kind of wearable technology can really come into play, helping us shave off vital seconds and create more time where we can really focus on the patient.”