SmokingOpp: Detecting the Smoking 'Opportunity' Context Using Mobile Sensors

Area of Science:

• Behavioral Science
• Digital Health
• Machine Learning

Background:

• Impulsive behaviors are influenced by environmental contexts, with some contexts triggering rather than preventing them.
• These 'opportunity' contexts can be leveraged for timely, context-sensitive interventions.
• Smoking cessation is a key area where understanding opportunity contexts is crucial.

Purpose of the Study:

• To define and operationalize the concept of 'opportunity' contexts for impulsive behaviors.
• To apply this concept to smoking cessation, identifying specific triggers for smoking.
• To develop and validate a model for automatically detecting smoking opportunity contexts.

Main Methods:

• Defined 'opportunity' contexts based on self-reported smoking allowance and cigarette availability.
• Utilized Granger causality to establish the link between opportunity contexts and smoking occurrences.
• Developed the SmokingOpp model by extracting features from GPS traces, including novel 'smoking spot' locations.

Main Results:

• Demonstrated the clinical utility of opportunity contexts in predicting smoking occurrences.
• Successfully developed the SmokingOpp model for automated detection of smoking opportunity contexts.
• Validated the model using a large dataset of GPS points and self-reports from 90 smokers.

Conclusions:

• 'Opportunity' contexts are significant predictors of impulsive smoking behavior.
• The SmokingOpp model offers a novel, data-driven approach to identify these contexts.
• This research provides a foundation for developing effective, context-aware smoking cessation interventions.