John Wiederspan, a nurse anesthetist at UW Medicine in Seattle, is well aware of how errors can occur in the high-pressure operating room environment, especially during emergencies when adrenaline and urgency lead to rushed administration of emergency drugs. Despite ongoing patient safety efforts, medication errors remain common, affecting at least 1 in 20 patients, with about 1. 3 million injuries and one death daily in the U. S. alone, according to the World Health Organization. Medication errors often involve administering the wrong drug or incorrect dosages. Hospitals have implemented safeguards like color-coded labels and barcode scanners to mitigate mistakes, yet errors persist. Dr. Kelly Michaelsen, an anesthesiologist and engineer at UW Medicine and the University of Washington, noted that 90% of anesthesiologists admit to medication errors at some point in their careers. She envisioned that AI could assist by acting as a second set of eyes to detect errors in real time, given that about 99% of medications used belong to a limited set of 10-20 drugs. Her focus was on vial swap errors, responsible for around 20% of medication mistakes, which occur when the wrong vial or syringe label leads to patients being injected with incorrect drugs. A tragic example is the death of a 75-year-old woman at Vanderbilt University Medical Center due to receiving a paralyzing drug instead of a sedative. To prevent such errors, Michaelsen developed AI-powered “smart eyewear” integrated with a camera into protective glasses worn during operations. This system scans, reads, and compares labels on vials and syringes, alerting clinicians if mismatches are detected. Building and training the AI took over three years, including obtaining approvals for using prerecorded video of medication preparation and simulated error scenarios for training, due to ethical restrictions against deliberately making errors in real patients.

The AI demonstrated 99. 6% accuracy in detecting vial swap errors. Next steps involve determining how best to deliver warnings, with auditory alerts favored, pending FDA clearance. The study was independently funded. Wiederspan, who has tested the device, is optimistic about its potential to enhance patient safety but notes the current GoPro-style headset is bulky, suggesting a smaller form factor would increase adoption. Patient safety advocates like Dr. Dan Cole of UCLA Health and Melissa Sheldrick, who lost her son to a medication error, support AI’s promise while cautioning that technology alone cannot prevent all errors. Sheldrick emphasizes that errors often stem from multiple factors such as poor communication and fragmented hospital systems, and technology should be viewed as a supportive layer rather than a fail-safe solution. Experts highlight that traditional interventions relying on increasing vigilance or checklists often add cognitive burden to already overwhelmed clinicians and have limited effectiveness. Dr. Nicholas Cordella from Boston University points out that AI-enabled cameras provide passive monitoring without adding to clinicians’ workload, potentially improving safety in hectic environments like operating and emergency rooms. Looking ahead, Michaelsen aims to expand the AI’s capability to measure drug volumes in syringes to prevent dosing errors, especially critical in pediatrics where patient sizes vary greatly and medication dilution is common. Wiederspan also envisions the technology being adapted for use in emergency rooms and hospital floors to help with oral medication dispensing, which involves managing numerous drugs per patient. However, broader implementation of AI in hospitals raises important privacy and data security concerns, especially with devices that might capture patient faces or sensitive information. Michaelsen assures that their system only scans syringe labels and does not store personal data. Cordella stresses the need for clear standards, transparency, and monitoring to address privacy risks and warns against clinicians becoming overly dependent on AI, which could lead to reduced vigilance and neglect of traditional safety checks. Overall, AI-powered wearable technology shows promising potential to reduce medication errors in complex clinical settings, complementing existing safety measures and aiding healthcare providers in improving patient outcomes while highlighting the importance of ethical, practical, and privacy considerations.