TABLE 3.

Generalized Linear Models Used to Link Absolute Harm Perceptions of Tobacco Products Among Youth at W1 to New Use at W2

	RRa	95% CIa	ARRa,b	95% CIa
Cigarettes
No or little harm	1.5	0.7–3.0	1.5	0.8–3.0
Some harm	2.4c	1.9–3.1	1.9c	1.4–2.4
A lot of harm	1.0	—	1.0	—
E-cigarettes
No or little harm	3.0c	2.4–3.8	2.2c	1.7–2.8
Some harm	1.6c	1.3–2.1	1.3c	1.0–1.7
A lot of harm	1.0	—	1.0	—
Cigars overall
No or little harm	3.7c	2.5–5.5	2.6c	1.8–4.0
Some harm	1.6c	1.2–2.2	1.3	0.9–1.8
A lot of harm	1.0	—	1.0	—
Pipes
No or little harm	4.2c	1.8–9.5	2.8c	1.2–6.6
Some harm	1.4	0.7–2.9	1.1	0.6–2.2
A lot of harm	1.0	—	1.0	—
Hookah
No or little harm	3.6c	2.6–5.0	2.7c	2.0–3.8
Some harm	1.5c	1.0–2.1	1.2	0.8–1.7
A lot of harm	1.0	—	1.0	—
Smokeless tobacco
No or little harm	3.2c	1.9–5.4	2.8c	1.7–4.5
Some harm	1.5	0.9–2.4	1.3	0.8–2.2
A lot of harm	1.0	—	1.0	—

From the US PATH Study (2013–2015). The sample included youth who completed a W2 interview and had never used each product at W1 but had heard of the product, respectively. —, not applicable.

^aEstimates and 95% CIs are weighted to represent the US youth population at W2.

^bARRs are from generalized linear models (binomial family with log link) that are controlled for age, sex, race, region, parental education, ever alcohol use, and ever tobacco use at W1. In adjusted models for e-cigarettes, cigars, and smokeless tobacco, a generalized linear model (Poisson family with log link) was used.

^cStatistically significant at the α = .05 level.