Dynamic models of stress-smoking responses based on high-frequency sensor data

Area of Science:

• Behavioral Science
• Digital Health
• Biomedical Engineering

Background:

• Self-reports link stress to increased smoking risk.
• Sensor data offers a more detailed understanding of stress-smoking dynamics.
• Personalized models of stress-smoking responses are crucial for effective interventions.

Purpose of the Study:

• To identify person-specific dynamic models of stress and smoking behavior.
• To leverage wearable sensing technology for real-time stress and smoking detection.
• To analyze stress patterns immediately before, during, and after smoking events.

Main Methods:

• Utilized wearable sensors (AutoSense chestband, MotionSense wristbands) on 45 adult smokers.
• Collected intensive longitudinal data on physiological stress markers and smoking events.
• Employed regression analysis to model minute-level smoking event probabilities based on stress probabilities.

Main Results:

• Identified consistent patterns of increased smoking odds following stress onset, either immediate or delayed, across all participants.
• Observed a subsequent dramatic reduction in smoking probability for approximately 49% of participants.
• Sensor-detected stress probabilities highlighted individual vulnerabilities to smoking.

Conclusions:

• Personalized stress-smoking dynamics can be modeled using wearable sensor data.
• These models can inform the development of tailored, just-in-time interventions for smoking cessation.
• Objective stress data provides a valuable variable for supporting quit attempts.