Development and validation of a deep learning model for detection of allergic reactions using safety event reports across hospitals (16 November 2020)

Summary

A deep learning algorithm accurately identified allergic reactions in hospital patient safety reports, which could help providers avoid medical errors and improve event surveillance, according to a study from Yang et al. published in JAMA Network Open.

Allergic reactions – to medications, foods, and healthcare products – are becoming increasingly common in the US.

Researchers noted that up to 36% of patients report drug allergies, and 4-10% report food allergies. Patients in healthcare settings are at particularly high risk of developing an allergic reaction, and it’s critical that providers are able to quickly detect and monitor these events.

Results of this study suggest that deep learning can improve the accuracy and efficiency of the allergic reaction identification process, which may facilitate future real-time patient safety surveillance and guidance for medical errors and system improvement.