LAUSR

Introduction Estimates of initiation, cessation, and relapse rates of tobacco cigarette smoking and e-cigarette use can facilitate projections of longer-term impact of their use. We aimed to derive transition rates and apply them to validate a microsimulation model of tobacco that newly incorporated e-cigarettes. Methods We fit a Markov multi-state model (MMSM) for participants in Waves 1–4.5 of the Population Assessment of Tobacco and Health (PATH) longitudinal study. The MMSM had nine cigarette smoking and e-cigarette use states (current/former/never use of each), 27 transitions, two sex categories, and four age categories (youth: 12-17y; adults: 18-24y/25-44y/≥45y). We estimated transition hazard rates, including initiation, cessation, and relapse. We then validated the Simulation of Tobacco and Nicotine Outcomes and Policy (STOP) microsimulation model, by: (a) using transition hazard rates derived from PATH Waves 1–4.5 as inputs, and (b) comparing STOP-projected prevalence of smoking and e-cigarette use at 12 and 24 months to empirical data from PATH Waves 3 and 4. We compared the goodness-of-fit of validations with “static relapse” and “time-variant relapse,” wherein relapse rates did not or did depend on abstinence duration. Results Per the MMSM, youth smoking and e-cigarette use was generally more volatile (lower probability of maintaining the same e-cigarette use status over time) than that of adults. Root-mean-squared error (RMSE) for STOP-projected versus empirical prevalence of smoking and e-cigarette use was <0.7% for both static and time-variant relapse simulations, with similar goodness-of-fit (static relapse: RMSE 0.69%, CI 0.38–0.99%; time-variant relapse: RMSE 0.65%, CI 0.42–0.87%). PATH empirical estimates of prevalence of smoking and e-cigarette use were mostly within the margins of error estimated by both simulations. Discussion A microsimulation model incorporating smoking and e-cigarette use transition rates from a MMSM accurately projected downstream prevalence of product use. The microsimulation model structure and parameters provide a foundation for estimating the behavioral and clinical impact of tobacco and e-cigarette policies.