Summary
Background
Several major theories of health behavior change touch on the idea that behaviors that are perceived as self-relevant are more likely to be enacted. Likewise, the perceived value, or relative importance, of the content in a health relevant message may change its impact on behavior. Supporting these ideas, we have recently found that specific neural indicators of self- and value-related processing during exposure to anti-tobacco messaging predict individual behavior change.
To explore the importance of self- and value-related processing in response to smoking-related information in the media environment, we focused on survey items assessing how self-relevant and valuable individuals find e-cigarette media. Our first goal is to determine whether the perceived self-relevance and value of messaging that participants are exposed to relates to concurrent smoking-related behaviors and antecedents to behavior (e.g., intentions, beliefs, and social norms). In addition, as data collection continues, we plan to track longitudinally whether the self-relevance and value of media can predict changes in psychological (e.g., intentions, beliefs, and social norms) or behavioral (e.g., initiation) outcomes.
Our working idea is that (1) the valence of e-cigarette media will influence social norms around vaping, which in turn influence smoking intentions and behaviors, and (2) the self-relevance of e-cigarette media will influence health and risk beliefs about e-cigarettes, which will influence smoking intentions and behaviors.
Cross-sectional Models
Variables considered:
- Main dependent variables considered are e-cigarette use (“have you ever vaped”), intentions to vape, perceived social norms around vaping, and health/risk beliefs about vaping.
- Predictor variables of interest are the self-relevance (“mostly directed at someone like me” vs “mostly directed at other types of people”) and valence (“mostly positive” vs “mostly negative”) of scanned media about e-cigarettes.
- Most models control for ecigarette smoker status, age, gender, race, grades in school, parental education, and whether someone else in the household vapes; weighting variables provided by TCORS group are used.
Click to view descriptives
Click to view boxplots of main relationships between variables
Summary of findings:
Examined relationships between self-relevance and value of scanned e-cigarette media (removed “a mix” answers for models) and having ever vaped, intentions to vape, social norms around vaping, and risk and health beliefs about e-cigarettes. Some mediation models are presented here, although we understand that strong statements about causality cannot be made in cross-sectional data. N=343.
(1). Young adults are more likely to have ever vaped and to be a current e-cigarette smoker with exposure to positive, self-relevant media.
(2). Intentions to vape are associated with exposure to positive media.
(3). Intentions to vape are associated with having ever vaped.
(4a). Exposure to self-related, positive media is associated with positive social norms around vaping.
(4b). Social norms mediate the relationship between positive value of media and intentions to vape.
(4c). Social norms partially mediate the relationship between self-relevance and positive value of media and having ever vaped.
(5a). Valence of media is associated with beliefs re e-cigs (will become addicted to nicotine if I use e-cigs every day; can quit tobacco cigs by vaping; self-efficacy to quit vaping).
(5b). Beliefs re e-cigs (will become addicted to nicotine if I use e-cigs every day; can quit tobacco cigs by vaping; self-efficacy to quit vaping) are associated with intentions to vape.
Longitudinal Models
These models are currently underpowered, and are intended as an example of how we plan to analyze the data rather than up for firm interpretation. Total N in these models is 79, and only ~5 participants change their vaping behavior or intentions to vape from T1 to T2. (Click for descriptives.)
Summary of findings:
(1). Self-relevance of media predicts having ever vaped at T2, controlling for having ever vaped at T1.
(2a). Exposure to positive, self-relevant media predicts intentions to vape at T2, controlling for intentions at T1.
(2b). Intentions to vape mediate the relationship between self-relevance of media at T1 and having ever vaped at T2, controlling for having ever vaped at T1.
(3). Social norms mediate the relationship between positive valence of media at T1 and intentions to vape at T2, controlling for intentions to vape at T1.
(4). Risk and health beliefs mediate relationship between positive valence of media at T1 and intentions at T2, controlling for intentions at T1.
Issues for discussion
- How to incorporate “a mix” answers for self-relevance and value of media questions
- Clarifying theoretical questions
- Modeling intentions and beliefs - binary vs continuous variables
- Covariates
- Small N’s, particularly in longitudinal data
- Improve mediation models (bootstrapping etc)
Cross-sectional Models
Descriptives
## [1] "Cross-sectional N=343"
## [1] "Ecigselfdirectscan: Self-directedness of scanned information; 1=Mostly like me, 2=Mostly not like me, 3=A mix"
## ecigselfdirectscan_n
## 1 2 3
## 232 372 930
## [1] "Valscan: Valence of scanned information; 1=Mostly positive, 2=Mostly negative, 3=A mix"
## valscan_ecig_n
## 1 2 3
## 596 216 717
## [1] "Most models control for:"
## ecigsmoker
## Not current e-cigarette user Current e-cigarette user
## 292 51
## sex
## age2cat Male Female Refused
## 13-17 105 73 0
## 18-25 92 73 0
## race4cat
## Hispanic White (Non-Hisp) Black or AA (Non-Hisp)
## 82 181 45
## Other/1+
## 35
## goodgrades
## Mostly F's Mostly D's Mostly C's Mostly B's Mostly A's
## 5 11 42 135 147
## paredu
## Less than high school (0-11th grade)
## 11
## High school degree (completed 12th grade, high school diploma or GED)
## 62
## Some college (1-3 years, Associate's Degree)
## 60
## College Degree (4 years, Bachelor's Degree)
## 106
## Completed Graduate or professional school after college (for example: MA or PhD,
## 78
## Don't Know
## 26
## Refused
## 0
## hhuse_vap
## No Yes Lives alone
## 245 52 46
Boxplots
Red dots mark the mean
Ever vape
boxplot(qn15_n~ecigselfdirectscan_n,data=survey_selfanswers,names=c("Like me","Others","A mix"),ylab="No=0 | Yes=1",main="Ever vape x Media Self-relevance")
means<-tapply(survey_selfanswers$qn15_n,survey_selfanswers$ecigselfdirectscan_n,mean)
points(means,col="red",pch=18)
boxplot(qn15_n~valscan_ecig_n,data=survey_valanswers,names=c("Positive","Negative","A mix"),ylab="No=0 | Yes=1",main="Ever vape x Media Valence")
means<-tapply(survey_valanswers$qn15_n,survey_valanswers$valscan_ecig_n,mean)
points(means,col="red",pch=18)
Intentions
boxplot(qn22_n~ecigselfdirectscan_n,data=survey_selfanswers,names=c("Like me","Others","A mix"),main="Intentions x Media Self-relevance")
means<-tapply(survey_selfanswers$qn22_n,survey_selfanswers$ecigselfdirectscan_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
boxplot(qn22_n~valscan_ecig_n,data=survey_valanswers,names=c("Positive","Negative","A mix"),main="Intentions x Media Valence")
means<-tapply(survey_valanswers$qn22_n,survey_valanswers$valscan_ecig_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
Norms
boxplot(norms_ecigs~ecigselfdirectscan_n,data=survey_selfanswers,names=c("Like me","Others","A mix"),main="Social Norms x Media Self-relevance")
means<-tapply(survey_selfanswers$norms_ecigs,survey_selfanswers$ecigselfdirectscan_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
boxplot(norms_ecigs~valscan_ecig_n,data=survey_valanswers,names=c("Positive","Negative","A mix"),main="Social Norms x Media Valence")
means<-tapply(survey_valanswers$norms_ecigs,survey_valanswers$valscan_ecig_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
Beliefs
1: “Will become addicted to nicotine if vape/use e-cigs every day”
boxplot(qn46_n~ecigselfdirectscan_n,data=survey_selfanswers,names=c("Like me","Others","A mix"),main="Addicted x Media Self-relevance")
means<-tapply(survey_selfanswers$qn46_n,survey_selfanswers$ecigselfdirectscan_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
boxplot(qn46_n~valscan_ecig_n,data=survey_valanswers,names=c("Positive","Negative","A mix"),main="Addicted x Media Valence")
means<-tapply(survey_valanswers$qn46_n,survey_valanswers$valscan_ecig_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
2: “Can quit tobacco cigs by vaping/using ecigs”
boxplot(qn47_n~ecigselfdirectscan_n,data=survey_selfanswers,names=c("Like me","Others","A mix"),main="Quit x Media Self-relevance")
means<-tapply(survey_selfanswers$qn47_n,survey_selfanswers$ecigselfdirectscan_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
boxplot(qn47_n~valscan_ecig_n,data=survey_valanswers,names=c("Positive","Negative","A mix"),main="Quit x Media Valence")
means<-tapply(survey_valanswers$qn47_n,survey_valanswers$valscan_ecig_n,mean,na.rm=TRUE)
points(means,col="red",pch=18)
Behavior
# Ecigselfdirectscan: Self-directedness of scanned information; 1=Mostly like me, 2=Mostly not like me
xtabs(~ecigselfdirectscan_n + qn15_r, survey_reduced)
## qn15_r
## ecigselfdirectscan_n No Yes
## 1 75 58
## 2 165 45
# Valscan: Valence of scanned information; 1=Mostly positive, 2=Mostly negative
xtabs(~valscan_ecig_n + qn15_r, survey_reduced)
## qn15_r
## valscan_ecig_n No Yes
## 1 150 90
## 2 90 13
Intentions
# Binary intentions
xtabs(~ecigint, survey_reduced)
## ecigint
## Definitely no intention to vape Some intention to vape
## 230 112
# Ecigselfdirectscan: Self-directedness of scanned information; 1=Mostly like me, 2=Mostly not like me
xtabs(~ecigselfdirectscan_n + ecigint, survey_reduced)
## ecigint
## ecigselfdirectscan_n Definitely no intention to vape
## 1 72
## 2 158
## ecigint
## ecigselfdirectscan_n Some intention to vape
## 1 61
## 2 51
# Valscan: Valence of scanned information; 1=Mostly positive, 2=Mostly negative
xtabs(~valscan_ecig_n +ecigint, survey_reduced)
## ecigint
## valscan_ecig_n Definitely no intention to vape Some intention to vape
## 1 142 98
## 2 88 14
# Continuous intentions
xtabs(~qn22_n, survey_reduced)
## qn22_n
## 1 2 3 4
## 230 55 34 23
xtabs(~ecigselfdirectscan_n + qn22_n, survey_reduced)
## qn22_n
## ecigselfdirectscan_n 1 2 3 4
## 1 72 24 22 15
## 2 158 31 12 8
xtabs(~valscan_ecig_n + qn22_n, survey_reduced)
## qn22_n
## valscan_ecig_n 1 2 3 4
## 1 142 44 32 22
## 2 88 11 2 1
# Original measure of intentions to vape is continuous (1=Definitely do not intend to vape, 4=Definitely do intend to vape); binary variable was created by TCORS group (1=Definitely no intention to vape, 2=all other answers). Analyses below are conducted with both measures, produce similar results.
Norms
Average of “how many of your close friends vape”; “how many people your age do you think vape”; “friends’ approval if you vaped”
Beliefs
Main effects - ecig vs tobacco beliefs
Beliefs predict intentions to vape
##
## Call:
## glm(formula = as.factor(survey_reduced$ecigint) ~ qn46_n + as.factor(age2cat) +
## as.factor(sex) + as.factor(race4cat_n) + paredu_n + goodgrades_n +
## as.factor(hhuse_vap_n), family = binomial, data = survey_reduced,
## weights = weight1_main)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.2769 -0.8405 -0.5375 0.9051 2.8178
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 1.3066 0.9490 1.377 0.16857
## qn46_n -0.3685 0.1785 -2.065 0.03894 *
## as.factor(age2cat)18-25 0.2872 0.2859 1.004 0.31520
## as.factor(sex)Female -0.2021 0.2748 -0.736 0.46198
## as.factor(race4cat_n)1 0.7711 0.3516 2.193 0.02827 *
## as.factor(race4cat_n)2 -0.3025 0.4294 -0.704 0.48112
## as.factor(race4cat_n)3 0.3507 0.4233 0.829 0.40738
## paredu_n 0.1644 0.1244 1.321 0.18648
## goodgrades_n -0.4103 0.1584 -2.591 0.00957 **
## as.factor(hhuse_vap_n)1 0.8732 0.3537 2.468 0.01357 *
## as.factor(hhuse_vap_n)2 -0.3734 0.4213 -0.886 0.37540
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 365.41 on 305 degrees of freedom
## Residual deviance: 329.05 on 295 degrees of freedom
## (37 observations deleted due to missingness)
## AIC: 350.67
##
## Number of Fisher Scoring iterations: 4
##
## Call:
## glm(formula = as.factor(survey_reduced$ecigint) ~ qn47_n + as.factor(age2cat) +
## as.factor(sex) + as.factor(race4cat_n) + paredu_n + goodgrades_n +
## as.factor(hhuse_vap_n), family = binomial, data = survey_reduced,
## weights = weight1_main)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.3899 -0.8214 -0.4964 0.9079 2.5825
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.02518 0.88386 -2.291 0.02195 *
## qn47_n 0.66727 0.17817 3.745 0.00018 ***
## as.factor(age2cat)18-25 0.34944 0.28590 1.222 0.22161
## as.factor(sex)Female -0.08014 0.27888 -0.287 0.77384
## as.factor(race4cat_n)1 0.77178 0.34959 2.208 0.02727 *
## as.factor(race4cat_n)2 -0.36841 0.43710 -0.843 0.39931
## as.factor(race4cat_n)3 0.53273 0.43728 1.218 0.22312
## paredu_n 0.16153 0.12099 1.335 0.18183
## goodgrades_n -0.32333 0.14836 -2.179 0.02931 *
## as.factor(hhuse_vap_n)1 0.94867 0.35647 2.661 0.00778 **
## as.factor(hhuse_vap_n)2 -0.38628 0.42301 -0.913 0.36115
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 375.58 on 309 degrees of freedom
## Residual deviance: 331.79 on 299 degrees of freedom
## (33 observations deleted due to missingness)
## AIC: 359.48
##
## Number of Fisher Scoring iterations: 4
##
## Call:
## glm(formula = as.factor(survey_reduced$ecigint) ~ se_vape_all_n +
## as.factor(age2cat) + as.factor(sex) + as.factor(race4cat_n) +
## paredu_n + goodgrades_n + as.factor(hhuse_vap_n), family = binomial,
## data = survey_reduced, weights = weight1_main)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.3620 -0.8129 -0.5005 0.9154 2.4983
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -1.68861 1.04161 -1.621 0.104984
## se_vape_all_n 0.71570 0.19535 3.664 0.000249 ***
## as.factor(age2cat)18-25 0.34668 0.29003 1.195 0.231956
## as.factor(sex)Female -0.02074 0.28150 -0.074 0.941267
## as.factor(race4cat_n)1 0.68348 0.35552 1.922 0.054547 .
## as.factor(race4cat_n)2 -0.43037 0.43533 -0.989 0.322854
## as.factor(race4cat_n)3 0.35569 0.43388 0.820 0.412337
## paredu_n 0.16788 0.12428 1.351 0.176755
## goodgrades_n -0.37171 0.16520 -2.250 0.024445 *
## as.factor(hhuse_vap_n)1 0.70835 0.36239 1.955 0.050624 .
## as.factor(hhuse_vap_n)2 -0.51571 0.42246 -1.221 0.222187
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 368.46 on 302 degrees of freedom
## Residual deviance: 322.43 on 292 degrees of freedom
## (40 observations deleted due to missingness)
## AIC: 344.7
##
## Number of Fisher Scoring iterations: 4
Longitudinal Models
## Descriptives
## [1] "Longitudinal N=79"
Behavior
## [1] "Ever vape breakdown: T1"
## qn15_r
## No Yes
## 55 24
## [1] "Ever vape breakdown: T2"
## qn15_rx
## No Yes
## 51 28
## [1] "Ever vape @ T2 by self & value T1"
## [1] "Ecigselfdirectscan: Self-directedness of scanned information; 1=Mostly like me, 2=Mostly not like me; Valscan: Valence of scanned information; 1=Mostly positive, 2=Mostly negative"
## qn15_rx
## ecigselfdirectscan_n No Yes
## 1 12 16
## 2 39 12
## qn15_rx
## valscan_ecig_n No Yes
## 1 31 27
## 2 20 1
Intentions
## [1] "Continuous intentions, T1"
## qn22_n
## 1 2 3 4
## 53 13 7 6
## [1] "Continuous intentions, T2"
## qn22_nx
## 1 2 3 4
## 50 16 6 7
## [1] "Binary intentions, T1"
## ecigint
## Definitely no intention to vape Some intention to vape
## 53 26
## [1] "Binary intentions, T2"
## ecigintx
## Definitely no intention to vape Some intention to vape
## 50 29
## [1] "Binary intentions T2 by self & value T1"
## [1] "Ecigselfdirectscan: Self-directedness of scanned information; 1=Mostly like me, 2=Mostly not like me; Valscan: Valence of scanned information; 1=Mostly positive, 2=Mostly negative"
## ecigintx
## ecigselfdirectscan_n Definitely no intention to vape
## 1 10
## 2 40
## ecigintx
## ecigselfdirectscan_n Some intention to vape
## 1 18
## 2 11
## ecigintx
## valscan_ecig_n Definitely no intention to vape Some intention to vape
## 1 32 26
## 2 18 3
Norms
Average of “how many of your close friends vape”; “how many people your age do you think vape”; “friends’ approval if you vaped”
Beliefs
Mediations
Risk belief - “Will become addicted to nicotine if I use e-cigs” - mediates relationship between valence of media and intentions
(a) Postively valenced media exposure (T1) predicts increased intentions (T2, controlling T1)
## [1] "Continuous intention"
##
## Call:
## lm(formula = survey_ecigs$qn22_nx ~ qn22_n + as.factor(ecigselfdirectscan_n) +
## as.factor(valscan_ecig_n) + as.factor(ecigsmoker) + as.factor(age2cat) +
## as.factor(sex) + as.factor(race4cat_n) + paredu_n + goodgrades_n +
## as.factor(hhuse_vap_n), data = survey_ecigs, weights = weight2_main)
##
## Weighted Residuals:
## Min 1Q Median 3Q Max
## -1.3613 -0.5499 -0.1588 0.1996 1.6347
##
## Coefficients:
## Estimate Std. Error t value
## (Intercept) 4.64110 0.87149 5.325
## qn22_n 0.19603 0.23529 0.833
## as.factor(ecigselfdirectscan_n)2 -0.88575 0.21363 -4.146
## as.factor(valscan_ecig_n)2 -0.63768 0.26465 -2.410
## as.factor(ecigsmoker)Current e-cigarette user 0.56559 0.52179 1.084
## as.factor(age2cat)18-25 -0.07786 0.26586 -0.293
## as.factor(sex)Female -0.43939 0.23394 -1.878
## as.factor(race4cat_n)1 -0.23871 0.32389 -0.737
## as.factor(race4cat_n)2 -0.75318 0.32694 -2.304
## as.factor(race4cat_n)3 -0.82458 0.51719 -1.594
## paredu_n 0.01443 0.10153 0.142
## goodgrades_n -0.48217 0.13489 -3.575
## as.factor(hhuse_vap_n)1 -0.18593 0.31127 -0.597
## as.factor(hhuse_vap_n)2 -0.52985 0.33824 -1.566
## Pr(>|t|)
## (Intercept) 1.92e-06 ***
## qn22_n 0.408382
## as.factor(ecigselfdirectscan_n)2 0.000118 ***
## as.factor(valscan_ecig_n)2 0.019348 *
## as.factor(ecigsmoker)Current e-cigarette user 0.283121
## as.factor(age2cat)18-25 0.770737
## as.factor(sex)Female 0.065656 .
## as.factor(race4cat_n)1 0.464239
## as.factor(race4cat_n)2 0.025043 *
## as.factor(race4cat_n)3 0.116589
## paredu_n 0.887527
## goodgrades_n 0.000739 ***
## as.factor(hhuse_vap_n)1 0.552745
## as.factor(hhuse_vap_n)2 0.122971
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.7122 on 55 degrees of freedom
## (10 observations deleted due to missingness)
## Multiple R-squared: 0.6479, Adjusted R-squared: 0.5646
## F-statistic: 7.783 on 13 and 55 DF, p-value: 1.844e-08
## [1] "Binary intention - marginal"
##
## Call:
## glm(formula = as.factor(survey_ecigs$ecigintx) ~ ecigint + as.factor(ecigsmoker) +
## as.factor(ecigselfdirectscan_n) + as.factor(valscan_ecig_n) +
## as.factor(age2cat) + as.factor(sex) + as.factor(race4cat_n) +
## paredu_n + goodgrades_n, family = binomial, data = survey_ecigs,
## weights = weight2_main)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.63685 -0.43928 -0.16144 0.00009 2.21660
##
## Coefficients:
## Estimate Std. Error
## (Intercept) 2.082e+00 3.320e+00
## ecigintSome intention to vape 1.574e+00 1.381e+00
## as.factor(ecigsmoker)Current e-cigarette user 1.859e+01 2.672e+03
## as.factor(ecigselfdirectscan_n)2 -1.710e+00 9.473e-01
## as.factor(valscan_ecig_n)2 -2.297e+00 1.305e+00
## as.factor(age2cat)18-25 5.052e-01 9.700e-01
## as.factor(sex)Female -2.419e-03 9.912e-01
## as.factor(race4cat_n)1 6.620e-01 1.551e+00
## as.factor(race4cat_n)2 -1.355e+00 1.473e+00
## as.factor(race4cat_n)3 -4.934e-01 1.843e+00
## paredu_n 9.292e-03 4.302e-01
## goodgrades_n -3.600e-01 6.893e-01
## z value Pr(>|z|)
## (Intercept) 0.627 0.5306
## ecigintSome intention to vape 1.140 0.2544
## as.factor(ecigsmoker)Current e-cigarette user 0.007 0.9944
## as.factor(ecigselfdirectscan_n)2 -1.805 0.0710 .
## as.factor(valscan_ecig_n)2 -1.760 0.0784 .
## as.factor(age2cat)18-25 0.521 0.6025
## as.factor(sex)Female -0.002 0.9981
## as.factor(race4cat_n)1 0.427 0.6695
## as.factor(race4cat_n)2 -0.920 0.3578
## as.factor(race4cat_n)3 -0.268 0.7889
## paredu_n 0.022 0.9828
## goodgrades_n -0.522 0.6015
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 71.35 on 68 degrees of freedom
## Residual deviance: 34.19 on 57 degrees of freedom
## (10 observations deleted due to missingness)
## AIC: 58.112
##
## Number of Fisher Scoring iterations: 18
Belief “Can quit tobacco cigs using e-cigs” (T1) mediates effect of positive media exposure (T1) on intentions (T2, controlling T1)
(1) Social norms at T1 mediate the relationship between positive valence of media at T1 and changes in intentions at T2 (consistent main effect of self-relevance)