Detecting Real-Time Substance Use from Wearable Biosensor Data Stream

• Activity Number: 499
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Chanpaul Jin Wang* and Hua Fang and Stephanie Carreiro and Honggang Wang and Edward Boyer
• Companies: University of Massachusetts Medical School and University of Massachusetts Medical School and University of Massachusetts Medical School and University of Massachusetts - Dartmouth and University of Massachusetts Medical School
• Keywords: wearable biosensor ; real time ; data stream ; statistical learning ; substance use ; behavioral intervention

Abstract

Detecting real time substance use is a critical step for near-real-time clinical interventions. Traditional clinical methods based on self-reporting, urine or blood screening are inefficient or intrusive for drug use detection, and inappropriate for timely interventions. For example, patients may not honestly report their drug use or suffer from their recall bias; while urine and blood screening are time-constrained, e.g., effective only within 72 hours. Methods for real-time substance use detection are severely underdeveloped, partly due to the novice of wearable biosensor technology and the lack of real clinical data for evaluation. Few current studies explored the methodology but are still preliminary. We propose a new real-time drug use detection method for such data stream mining. Specifically, this method is built upon the slide window technique, an engineering approach, to process the real-time data stream, and a distance-based outlier detection method to identify substance use events. This method is evaluated using real datasets and simulation. Our numerical analyses showed that this method could have the potential to auto-detect the substance use events.