Assessment of medication self-administration using artificial intelligence.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Patient Safety Technology

Background:

• Medication self-administration (MSA) errors are frequent with devices like inhalers and insulin pens, leading to poor adherence, hospitalizations, and increased costs.
• Current monitoring methods for MSA can be intrusive or require direct patient contact, posing challenges for continuous, unobtrusive oversight.

Purpose of the Study:

• To develop and evaluate a contactless, AI-powered framework for detecting and monitoring errors in medication self-administration.
• To assess the system's ability to identify the use of specific medication delivery devices and adherence to correct usage steps.

Main Methods:

• Developed an artificial intelligence (AI) framework analyzing wireless signals within a patient's home environment to detect MSA events.
• Trained the AI model by observing medication self-administration by volunteers.
• Evaluated the system's performance by comparing its predictions against human annotations of medication use.

Main Results:

• The AI framework demonstrated high accuracy in detecting inhaler use (AUC=0.992) and insulin pen use (AUC=0.967).
• The system also accurately assessed whether patients followed the correct steps for device usage (AUC=0.952).
• The approach is contactless and unobtrusive, requiring no physical interaction with the patient or devices.

Conclusions:

• Leveraging AI to analyze ambient wireless signals offers a promising, low-overhead solution for monitoring medication self-administration.
• This technology has the potential to significantly improve medication safety and treatment adherence for patients using devices.
• The contactless nature of the AI framework minimizes patient burden and facilitates integration into home healthcare settings.