Fig 1.
Classification of survey participants as never, former, and current smokers/e-cig users.
Population Assessment of Tobacco and Health (PATH) Study Waves 1–4.5 respondents were classified into never, former, and current smoking/e-cig use. A single participant’s classification could differ between the two products: e.g., former smoker/current e-cig user. In youth survey Waves 1–2 and adult survey Wave 1, regarding electronic products, participants were asked only about “electronic-cigarettes.” In all other survey waves, participants were asked about e-product use, which included "e-cigarettes, vape pens, personal vaporizers and mods, e-cigars, e-pipes, e-hookah or hookah pens." For this analysis, we considered an affirmative response to either of these questions as indicative of e-cig use. *Youth in Wave 1 were asked to attest to ever use instead of ever “fairly regular” use of e-cigs. In our analysis, youth in Wave 1 who attested to never having used e-cigs were classified as never e-cig users. We classified all other youth as “undefined” for Wave 1 e-cig use and excluded them from the analysis in that wave.
Fig 2.
Allowed and disallowed instantaneous transitions in the continuous-time Markov multi-state model of cigarette smoking and e-cig use states.
Continuous time Markov multi-state models do not require exact transition times to be observed and allow multiple transitions to occur between observations. They therefore require allowed instantaneous transitions to be specified. Allowed instantaneous transitions from the indicated pre-transition smoking and e-cig use state (row) to the indicated post-transition state (column) are in green. Disallowed instantaneous transitions are in white. We disallowed transitions that would entail never users going directly to a former use state and those that would entail former users or current users going to a never use state. Cells in gray are those reflecting staying in the same state.
Fig 3.
Annual cumulative transition probabilities according to the continuous-time Markov multi-state model for (a) youth and (b) adults.
Data are presented as the probability, among those in the indicated prior state (row), of being in the indicated subsequent state (column) one year later. In parentheses are the 95% confidence intervals. More likely transitions are pictured in orange, while less likely transitions are pictured in blue. Note that participants may experience more than one instantaneous transition in sequence within the year, meaning that some transitions that cannot occur instantaneously (e.g., NSNE to FSFE) are nonetheless allowed as annual cumulative transitions.
Table 1.
Continuous time Markov model-estimated baseline transition hazard rates and adjusted hazard rate ratios.