A Machine Learning-based Approach for Collaborative Non-Adherence Detection during Opioid Abuse Surveillance using a...

Area of Science:

• Biomedical Engineering
• Computer Science
• Public Health

Background:

• The opioid epidemic in the US necessitates effective monitoring strategies.
• Wearable biosensors offer a promising avenue for monitoring opioid use and enabling interventions.
• Biosensor sharing (CNA) poses a significant threat to the reliability of such monitoring systems.

Purpose of the Study:

• To develop and evaluate a machine-learning approach for detecting counterfeit mengakses (CNA) in wearable biosensor data.
• To leverage accelerometer and blood volume pulse (BVP) data for novel CNA detection.
• To enhance the accuracy and trustworthiness of opioid use surveillance systems.

Main Methods:

• Collected accelerometer and BVP data from 11 patients undergoing naloxone treatment in an Emergency Department.
• Developed personalized machine-learning classifiers trained on individual patient's unique biosensor readings.
• Simulated CNA by interchanging biosensor data snippets between patients to evaluate detection accuracy.

Main Results:

• Achieved an average CNA detection accuracy of 90.96% when the collaborator was within the dataset.
• Demonstrated 86.78% accuracy when the collaborator was from an unseen dataset, indicating generalizability.
• The personalized classifiers effectively captured unique physiological signals for robust detection.

Conclusions:

• Machine-learning analysis of wearable biosensor data can effectively detect counterfeit mengakses (CNA).
• This technology can significantly improve the integrity and effectiveness of opioid use surveillance.
• Personalized detection models offer a promising strategy for real-world application in public health monitoring.