AI-enabled wearables help detect medication errors | TechTarget

University of Washington researchers have developed an AI-enabled handheld camera designed to automatically detect potential clinical medication errors before medication is administered. Npj Digital Medicine study.

Drug-related errors — such as syringe and vial swap errors — are a significant driver of preventable patient injury, particularly in operating rooms, intensive care units and emergency rooms. The research team emphasized that drug administration errors are the most common serious events in anesthesia.

These errors often occur during intravenous injections. Before giving the medicine to the patient, the nurses must transfer it from the vial to the syringe. The researchers stressed that about a fifth of errors are related to substitution, where the syringe is labeled incorrectly or the wrong vial is selected. A similar number of errors occur when a medication is correctly prescribed, but incorrectly prescribed.

Safeguards — including barcode systems that read and verify vial contents — are in place, but these devices add an extra step to the physician’s work, making it possible for a provider to forget to check the barcode system. during high pressure stress.

To solve this, the researchers sought to create a device that could seamlessly integrate to prevent medication errors. They started by building a deep learning system that can identify drug labels on syringes and vials during drug preparation events. Then, this method was combined with a GoPro camera in order to create a real-time warning system for possible medication errors.

The system was trained using 4K video data across 2 hospitals, 17 operating rooms and 13 anesthesiology providers over a period of 55 days. syringes and vials of different medicines, which were later written to help the algorithm learn to recognize them in videos and images.

Training the model presented a unique challenge, as lighting and settings varied across operating rooms. Additionally, the system must be able to identify the contents of the vial or syringe may not be clearly displayed on the camera.

“It was very difficult, because someone was there [operating room] it’s holding a syringe and a vial, and you can’t see any of those things in detail. Some letters (on syringe and vial) are covered by hands. And the hands moved quickly. They are the ones who do the work. They’re not posing for the camera,” explained Shyam Gollakota, Ph.D., co-author of the paper and a professor at the University of Washington’s Paul G. Allen School of Computer Science & Engineering, in a press release.

Instead of reading the words on the bottles directly, the app looks at other visual details, such as the print size of the label, the size and shape of each vial or syringe and the color of the vial cap. The model is also able to distinguish between front-end drugs and ignore back-end ones.

After training, the algorithm was tested on 418 drug inhalation videos during normal care. The model achieved 99.6% sensitivity and 98.8% specificity in detecting vial exchange errors.

The research team concluded that these results demonstrate the potential of a handheld camera to reduce medication errors and promote patient safety.

“The idea of ​​being able to help patients in real time or prevent a medication error before it happens is very powerful,” said Kelly Michaelsen, MD, Ph.D, co-author and assistant professor of anesthesiology and medicine. pain. University of Washington School of Medicine. “One can hope for 100% efficiency but even humans cannot achieve that. In a survey of over 100 anesthetists, most wanted the system to be more than 95% accurate, which is a goal we have achieved.”

Shania Kennedy has been reporting on health IT and analytics since 2022.