Table 4.

Association of Baseline E-Cigarette Ever Use and Covariates to Combustible Tobacco Use Outcomes at 6- and 12-Month Follow Ups among Baseline Never-Smokers

	Outcome
	Any Tobacco Producta		Combustible Cigarettesb		Cigarsc		Hookahd		Number of Tobacco Productse
Baseline Regressors and Covariates	OR(95%CI)f	P	OR(95%CI)f	P	OR(95%CI)f	P	OR(95%CI)f	P	OR(95%CI)g	P
Unadjusted Models
E-cigarette ever (vs. never) use	4.27 (3.19, 5.71)	<.001	2.65 (1.73, 4.05)	<.001	4.85 (3.38, 6.96)	<.001	3.25 (2.29, 4.62)	<.001	4.26 (3.16, 5.74)	<.001
Time (12- vs. 6-month follow-up)	1.09 (0.90, 1.32)	.38	1.03 (0.76, 1.39)	.85	1.25 (0.95, 1.64)	.11	0.99 (0.78, 1.25)	.90	1.08 (0.89, 1.30)	.45
Ever e-cigarette use × Timeh	0.64 (0.39, 1.04)	.07	0.73 (0.34, 1.56)	.41	0.68 (0.36, 1.26)	.22	0.60 (0.32, 1.12)	.11	0.64 (.039, 1.03)	.07
Adjusted Models
Categorical covariatesi
Female (vs. male)	0.88 (0.70, 1.11)	.28	1.50 (1.05, 2.15)	.03	1.48 (1.06, 2.06)	.02	0.67 (0.51, 0.90)	.007	0.92 (0.73, 1.17)	.52
Hispanic (vs. other) ethnicity	1.09 (0.82, 1.44)	.55	0.91 (0.59, 1.39)	.65	0.97 (0.66, 1.43)	.87	1.33 (0.97, 1.83)	.08	1.11 (0.83, 1.48)	.48
Lives with both biological parents (vs. other living situation)	0.80 (0.64, 1.01)	.07	0.96 (0.67, 1.37)	.80	0.58 (0.42, 0.80)	<.001	0.82 (0.62, 1.08)	.15	0.80 (0.63, 1.01)	.06
Substance ever (vs. never) use	2.33 (1.81, 2.99)	<.001	1.35 (0.91, 2.00)	.14	2.42 (1.72, 3.42)	<.001	2.25 (1.67, 3.01)	<.001	2.24 (1.73, 2.90)	<.001
Family history of smoking (yes vs. no)	1.11 (0.88, 1.41)	.37	1.34 (0.91, 1.98)	.14	0.95 (0.68, 1.32)	.75	0.99 (0.74, 1.32)	.94	1.11 (0.87, 1.42)	.39
Continuous covariatesj
Age	1.07 (0.96, 1.20)	.23	1.13 (0.96, 1.34)	.15	1.12 (0.95, 1.31)	.17	0.97 (0.84, 1.11)	.64	1.07 (0.95, 1.20)	.27
Parental education	0.83 (0.73, 0.94)	.004	0.81 (0.67, 0.97)	.03	0.88 (0.72, 1.06)	.18	0.86 (0.74, 1.01)	.07	0.83 (0.73, 0.95)	.007
Peer smoking	1.13 (1.00, 1.28)	.05	1.08 (0.90, 1.30)	.41	1.18 (1.00, 1.38)	.05	1.16 (1.01, 1.34)	.03	1.13 (1.00, 1.28)	.05
CESD-Depressive Symptoms	0.95 (0.84, 1.08)	.42	1.07 (0.89, 1.29)	.45	0.97 (0.81, 1.15)	.70	0.99 (0.85, 1.14)	.87	0.97 (0.86, 1.10)	.66
TCI-Impulsivity	1.23 (1.09, 1.38)	<.001	1.11 (0.93, 1.33)	.25	1.14 (0.97, 1.35)	.11	1.28 (1.11, 1.46)	<.001	1.22 (1.09, 1.38)	<.001
Delinquent Behavior	1.38 (1.18, 1.61)	<.001	1.35 (1.10, 1.66)	.004	1.39 (1.14, 1.70)	.001	1.38 (1.14, 1.66)	<.001	1.40 (1.20, 1.63)	<.001
Smoking susceptibility	1.08 (0.92, 1.26)	.37	1.46 (1.18, 1.80)	<.001	0.98 (0.79, 1.21)	.84	0.94 (0.77, 1.14)	.53	1.09 (0.93, 1.29)	.28
Smoking expectancies	1.24 (1.10, 1.41)	<.001	1.19 (0.98, 1.44)	.08	1.43 (1.20, 1.69)	<.001	1.26 (1.08, 1.46)	.003	1.27 (1.12, 1.44)	<.001
Regressors
E-cigarette ever (vs. never) use	2.73 (2.00, 3.73)	<.001	1.75 (1.10, 2.77)	.02	2.96 (2.00, 4.38)	<.001	2.26 (1.56, 3.29)	<.001	2.73 (1.99, 3.75)	<.001
Time (12- vs. 6-month follow-up)	1.09 (0.90, 1.33)	.37	1.03 (0.76, 1.40)	.85	1.26 (0.95, 1.67)	.10	0.98 (0.77, 1.24)	.84	1.08 (0.89, 1.32)	.42
Ever e-cigarette use × Timeh	0.64 (0.38, 1.05)	.08	0.74 (0.34, 1.61)	.44	0.68 (0.36, 1.29)	.23	0.59 (0.31, 1.12)	.11	1.54 (0.94, 2.51)	.08

Note. All analyses include only never-users of combustible tobacco products at baseline. -2 Res Log Pseudo-Likelihood fit index for unadjusted models without the interaction term:

^a24,932.93,

^b28,391.21,

^c27,019.65,

^d26,131.31,

^e70,339.66. -2 Res Log Pseudo-Likelihood fit index for unadjusted models with the interaction term:

^a24,946.6,

^b28,397.11,

^c27,050.36,

^d26,137.37,

^e70,411.89. Range of -2 Res Log Pseudo-Likelihood fit indices for adjusted models without the interaction term across the five imputed data sets:

^a25,701.9 – 25,743.77,

^b29,378.05 – 29,461.92,

^c28,211.45 – 28,286.11,

^d27,193.74 – 27,282.15,

^e73,652.58 – 73,852.15. Range of -2 Res Log Pseudo-Likelihood fit indices for adjusted models with the interaction term across the five imputed data sets:

^a25,717.24 – 25,759.27,

^b29,382.85 – 29,467.24,

^c28,238.77 – 28,312.83,

^d27,209.93 – 27,297.96,

^e73,730.36 – 73,930.47.

^fOR from repeated binary logistic regression model predicting respective outcome from baseline ever e-cigarette use status (yes/no) including school fixed effects.

^gOR from repeated ordinal logistic regression model predicting respective outcome from baseline ever e-cigarette use status (yes/no) including school fixed effects, with the OR reflecting the change in odds of being in a category with use of one or more tobacco products (3 vs. ≤2; ≥2 vs. ≤1, ≥1 vs. 0) associated with the covariate/regressor.

^hInteraction term added in subsequent model; parameter estimates for other regressors/covariates are from model excluding the interaction term.

ⁱPrevalance of combustible tobacco use outcomes at follow-ups by status on the categorical covariates can be found in eTable 2.

^jContinuous covariates rescaled (M=0, SD=1), such that the ORs indicate the change in odds in the outcome associated with an increase in one standard deviation unit on the covariate continuous scale. CESD = Center for Epidemiologic Studies Depression Scale. TCI = Temperament and Character Inventory.