TABLE 4

Ordered Probit Model Assessing Correlates of Harm Perception of Heavy Smoking

OR (95% CI)aAverage Marginal Effectsb
No HarmA Lot of Harm
Prob. (%)P valueProb. (%)P value
Gender
 Male1.00(Ref.)
 Female1.27(1.19–1.34)−1.0<.0014.2<.001
Ethnicity
 Non-Hispanic white1.00(Ref.)
 Non-Hispanic black0.76(0.70–0.82)1.2<.001−5.0<.001
 Hispanic0.75(0.70–0.81)1.3<.001−5.2<.001
 Non-Hispanic Asian0.96(0.78–1.18)0.1.71−0.6.71
 Other/multiple0.93(0.83–1.03)0.3.18−1.2.17
Age, y
 ≤121.00(Ref.)
 130.95(0.86–1.05)0.2.34−0.9.33
 140.99(0.89–1.09)0.0.83−0.2.83
 150.94(0.83–1.06)0.3.32−1.1.32
 161.1(0.99–1.22)−0.4.081.5.07
 ≥171.03(0.91–1.17)−0.1.620.5.62
Smoking pattern
 Nonsmoker1.00(Ref.)
 Intermittent smoker0.62(0.56–0.69)2.7<.001−10.4<.001
 Light smoker0.56(0.46–0.68)3.7<.001−13.5<.001
 Heavier smoker0.31(0.25–0.38)12.9<.001−33.8<.001
Alternative tobacco use
 No1.00(Ref.)
 Yes0.77(0.70–0.84)1.3<.001−5.3<.001
E-cigarette use
 No1.00(Ref.)
 Yes0.95(0.82–1.10)0.2.49−1.0.48
Family use of tobacco
 No1.00(Ref.)
 Yes0.82(0.78–0.87)0.8<.001−3.6<.001
Model coefficients
 Cut 10.09(0.09–0.10)
 Cut 20.14(0.13–0.15)
 Cut 30.24(0.22–0.26)
  • a OR displayed are from the weighted multivariate ordered probit model in which all listed covariates are included, with 95% CI calculated by Taylor series linearization to account for the complex survey design.

  • b Average marginal effects estimate the average change of altering each covariate (eg, being an intermittent smoker compared with being a nonsmoker) among the study population, estimating the likelihood that an individual perceived heavy smoking as causing no harm or a lot of harm.