Towards Identifying the Optimal Timing for Near Real-Time Smoking Interventions Using Commercial Wearable Devices

Tobacco addiction is one of the most challenging behavioral health problems with successful cessation rates remaining in single digits. With increased availability of commercially available mobile and wearable devices we have the opportunity to infer lapse vulnerability and intervene in near real-time. In this work we present findings from a mixed-method study where we collected around 398.2 hours of physiological data from regular smokers in the field (N is 5) using commercial wearable devices along with contextual data using EMAs. We applied quantitative and qualitative analysis techniques to identify key factors contributing to smoking lapse and developed a statistical model capable of inferring lapse vulnerability from physiological and contextual signals collected from the natural environment. Our methodology and findings show promise in designing practical, near real-time smoking intervention systems that can be used by regular smokers in their everyday living situation.