Time-varying coefficient cumulative gap time models for intensive longitudinal ecological momentary assessment data...

Area of Science:

• Behavioral Science
• Biostatistics
• Public Health

Background:

• Ecological Momentary Assessment (EMA) enhances real-time behavioral studies by minimizing recall bias and maximizing ecological validity.
• Understanding the temporal patterns of behavior, such as smoking bouts, is crucial for developing effective interventions.
• Intermittent smokers (ITS) exhibit unique smoking behaviors that require specialized analytical approaches.

Purpose of the Study:

• To develop and validate a statistical framework for analyzing clustered smoking behavior (smoking bouts) in intermittent smokers (ITS) using EMA data.
• To identify covariates that predict the occurrence and timing of smoking bouts.
• To address data challenges in EMA studies, including missing data and temporal complexities.

Main Methods:

• Introduction of a novel framework using event gap time functions to characterize temporal smoking behavior and distinguish smoking bouts.
• Application of time-varying coefficient models for cumulative log gap time to analyze temporal patterns.
• Incorporation of inverse probability weighting to effectively handle missing data in EMA studies.
• Validation through simulation studies to assess model reliability under various missing data conditions.

Main Results:

• The proposed framework reliably characterizes temporal smoking behavior and identifies smoking bouts in intermittent smokers.
• The time-varying coefficient models successfully adjusted for behavioral covariates influencing smoking patterns.
• Inverse probability weighting demonstrated effectiveness in accommodating missing data, whether missing by design or missing at random.
• Simulation studies confirmed the model's ability to accurately determine prespecified time-varying covariate coefficient forms when within-subject clustering was minimal.

Conclusions:

• The developed framework provides a robust method for analyzing complex temporal smoking behaviors in intermittent smokers using EMA data.
• The approach effectively addresses common challenges in EMA data analysis, such as missingness and temporal dependencies.
• This research offers valuable tools for understanding smoking microprocesses and informing targeted public health interventions for ITS.