Statistical methods for predicting e-cigarette use events based on beat-to-beat interval (BBI) data collected from...

Area of Science:

• Biomedical Engineering
• Data Science
• Public Health

Background:

• E-cigarette use is prevalent among young adults in the USA (14%).
• Many users intend to quit vaping but lack motivation and support.
• Wearable sensors offer potential for real-time physiological data collection to time interventions.

Purpose of the Study:

• To develop and evaluate statistical methods for de-noising and summarizing physiological data from smartwatches.
• To identify physiological patterns preceding e-cigarette use events.
• To compare proposed methods with conventional heart rate variability (HRV) analyses.

Main Methods:

• Collected 7-day beat-to-beat interval (BBI) data from 12 young adult e-cigarette users using smartwatches.
• Applied singular spectrum analysis (SSA) to de-noise BBI data.
• Developed a second-order polynomial model to summarize de-noised data and compared it with existing HRV methods.

Main Results:

• Singular spectrum analysis (SSA) effectively de-noised variable BBI data.
• The proposed second-order polynomial model achieved the highest Area Under the Curve (AUC) of 0.76, outperforming existing HRV methods.
• An increasing heart rate trend was observed before vaping events.

Conclusions:

• The study demonstrates the efficacy of SSA for de-noising BBI data and a novel polynomial model for HRV analysis.
• Observed physiological changes before vaping suggest a potential window for intervention.
• Findings may inform the development of timely interventions for e-cigarette cessation in young adults.