Calibrating Transition Probabilities for Tobacco Use in Markov Multi-state Modeling

Richards, Breanna

Full Metadata

Overview

Year:: 2023
Contributor:: Richards, Breanna (creator); Chrysanthopoulou, Stavroula (Advisor); Reddy, Krishna (Reader); Brown University. Department of Biostatistics (sponsor)
Genre:: theses
Subject:: Tobacco use; Generalized Linear Mixed Model; Longitudinal studies; Statistical matching; Electronic cigarettes; ENDS; Markov multi-state modeling; Transition Probabilities; Cigarettes
Extent:: xi, 91 p.

Files

Description

Abstract:: Abstract of Calibrating Transition Probabilities for Tobacco Use in Markov Multi-state Modeling, by Breanna Richards, Degree ScM, Brown University, May 2023 Background and Aims: Tobacco use is the leading cause of preventable disease, disability, and death in the United States. This problem space heightened when the U.S. witnessed a dramatic increase in the use of electronic nicotine delivery systems (ENDS) over the last decade, most notably among high school-aged youth. A study published in 2020 detailed the Simulation of Tobacco and Nicotine Outcomes and Policy (STOP) model, a microsimulation model to project the effects of tobacco use and cessation including the key behavior of relapse. This research expanded upon STOP by using data from a nationally representative tobacco use study (PATH) in the estimation/calibration of new behavioral transitions for tobacco use in youth and adults. Additionally, this study explored extensions of generalized linear models for longitudinal data and compared their performance and results to Markov multi-state models. Methods: Utilizing longitudinal survey response data from Waves 1-5 of PATH, we fit a weighted Markov multi-state model with nine cigarette smoking and ENDS use states, 27 allowed transitions, and sex and age as covariates. From this model, we estimated monthly transition rates between use states which were utilized in the validation of the STOP model. Projected prevalences from STOP were compared to empirical PATH data for goodness of fit. Mixed-effects models on longitudinal data for outcomes on current cigarette and ENDS use were assessed and compared. Results: STOP model results accurately reflected an increase in ENDS prevalence among youth over the course of 5 years. The overall root-mean-squared error and weighted root-mean-squared error for STOP-projected vs. empirical PATH prevalences were 1.87% and 0.95% respectively. Overall, smoking and ENDS use tend to be more unstable among youth than adults. Youth are less likely than adults to start smoking but more likely to start using ENDS. Additionally, youth who are current ENDS users are highly likely to quit but equally likely to relapse. Conclusion: The STOP microsimulation model accurately projected PATH-reported prevalences of product use. The model provides a foundation for estimating the behavioral and clinical impacts of tobacco policies and constitutes a useful tool for public health decision-making.
Notes:: Thesis (Sc. M.)--Brown University, 2023

Content

Citation

Richards, Breanna, "Calibrating Transition Probabilities for Tobacco Use in Markov Multi-state Modeling" (2023). Biostatistics Theses and Dissertations. Brown Digital Repository. Brown University Library. https://repository.library.brown.edu/studio/item/bdr:cbnzjz9q/

Relations

Collection:

Biostatistics Theses and Dissertations

Theses and Dissertations for the Biostatistics department.
...