- Email: [email protected]
Predictors of the transition to regular smoking during adolescence and young adulthood
JOURNAL OF ADOLESCENT HEALTH 2003;32:314 –324
ORIGINAL ARTICLE
Predictors of the Transition to Regular Smoking During Adolescence and Young Adulthood JOAN S. TUCKER, Ph.D., PHYLLIS L. ELLICKSON, Ph.D., AND DAVID J. KLEIN, M.S.
Purpose: To identify predictors of the transition from experimentation to regular smoking in middle adolescence, late adolescence, and young adulthood. Methods: California and Oregon students completed self-report surveys assessing the following potential predictors of the transition to regular smoking from grades 8 to 10 (n ⴝ 2496), grades 10 to 12 (n ⴝ 2149), and grade 12 to age 23 years (n ⴝ 1534): demographic characteristics; smoking-related attitudes, behaviors and environment; other problem behaviors; academic orientation; parental bonding; and mental health. Huberized regression techniques, which adjust for weighting and clustering of observations, were used to determine the independent associations of the predictor variables on subsequent smoking status. Results: Risk factors for the transition to regular smoking during middle adolescence included being white, prosmoking attitudes, friend smoking, weak academic orientation, and less parental support. During late adolescence, being African-American was protective, whereas risk factors included prosmoking attitudes, drinking, non-intact nuclear family, and less parental support. Risk factors in young adulthood included younger age and prosmoking attitudes. Conclusions: Results point to several smoking-related attitudes, social influences, and behaviors that prevention efforts may target to curb the escalation of smoking. © Society for Adolescent Medicine, 2003 KEY WORDS:
Adolescents Escalation Longitudinal Smoking Young adults
From RAND, Santa Monica, California. Address correspondence to: Joan S. Tucker, Ph.D., RAND, 1700 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138. Manuscript accepted November 1, 2002. 1054-139X/03/$–see front matter doi:10.1016/S1054-139X(02)00709-7
Experimentation with cigarette smoking is common during adolescence, with 78% of high school students having tried smoking by grade 12 [1]. Although most of these adolescents never become committed smokers, a sizable minority of those who experiment with cigarettes progress to regular smoking during the adolescent years. For example, two national surveys found that more than 20% of 12th graders reported smoking daily or almost daily in the past 30 days [1,2]. Regular smoking is much more prevalent among Whites than African-Americans or Hispanics and slightly more prevalent among male than female adolescents [1]. Although research examining the predictors of smoking has focused largely on smoking onset or initiation, it is important to understand the factors contributing to the escalation of smoking for several reasons. Smoking transitions tend toward heavier use; it is rare for adolescents who smoke regularly to revert to only occasional use [3]. Adolescents who are more committed smokers are also less likely to quit smoking than other adolescent smokers [4]. Furthermore, our own work has shown that adolescent smokers at grade 7 are at higher risk than experimenters both concurrently and 5 years later for academic difficulties, alcohol misuse and drug use, and delinquent behavior [5]. Thus, reducing the likelihood that an adolescent smoker will transition from experimentation to regular use is an important goal for prevention programs. Although the literature on smoking initiation and experimentation may provide some leads as to potentially important determinants of regular smoking, the risk factors for initial levels of smoking often differ from those for more advanced smoking [6 –9]. For example, family smoking has been identified as © Society for Adolescent Medicine, 2003 Published by Elsevier, 360 Park Avenue South, New York, NY 10010
April 2003
an important risk factor for smoking initiation in adolescence [10] and more strongly related to initiation than to the transition to regular smoking [11], perhaps because it allows the adolescent easy access to cigarettes and opportunities for experimentation. Once adolescents have experimented with cigarettes, peers may replace family members as the dominant social influence on smoking behavior. Consistent with this idea, several studies have found that “friend smoking” is a stronger predictor of the transition to regular use than “family smoking” [9,12], particularly when the majority of one’s friends smoke [13]. Cognitive models of health behavior such as the Theory of Reasoned Action give a central role to attitudes and beliefs in determining behaviors such as smoking [14]. Given that attitudes and beliefs tend to have a greater impact on behavior when they are based on direct experience [15], they may be particularly relevant to understanding the transition from experimentation to regular smoking. Chassin et al [11] found that the transition to regular smoking was best predicted by attitudes and beliefs about smoking, and behavioral intentions to smoke, among those who were already experimenting with smoking. There is also some evidence that parental and peer smoking influence smoking escalation indirectly through the adolescents’ beliefs about the negative health consequences of smoking and feelings of self-efficacy in refusing cigarette offers [8]. Other cross-sectional analyses have shown that 7thgrade regular smokers are more likely than experimenters to view cigarettes as affordable and smoking as having positive social consequences [9]. Some evidence suggests that deviancy and rebelliousness play a role in the transition to regular smoking. Chassin et al [11] found that a cluster of 17 variables chosen from Jessor and Jessor’s Problem Behavior Theory [16], such as having a devianceprone personality and low expectations for academic success, predicted the transition from experimental to regular smoking over a 1-year period. Other longitudinal research has indicated that having poor grades, experimenting with alcohol, and being inclined toward risk-taking in 7th grade are associated with heavy smoking 5 years later [13]. Rebelliousness has also been found to differentiate 7th-grade experimenters and regular smokers in cross-sectional analyses [9]. Important developmental changes occurring during adolescence and young adulthood may result in age-related differences in the predictors of smoking behavior [6]. Friends become increasingly important
PREDICTORS OF TRANSITION TO REGULAR SMOKING
315
during adolescence and the time spent with family decreases [17]. This suggests that peers may exert a stronger influence on smoking behavior as adolescents get older, although two studies of smoking initiation have yielded conflicting results. Whereas Krosnick and Judd [18] found that peer smoking was a stronger predictor of smoking initiation among adolescents than preadolescents, Chassin et al [19] did not find that peer influences were stronger risk factors with increasing age. Attitudes and beliefs also change during adolescence and young adulthood in ways that may be relevant to smoking transitions [20]. However, it is unclear whether such smoking attitudes become stronger or weaker predictors of the transition to regular smoking as adolescents mature. We know of only one study investigating the antecedents of the transition to regular smoking over multiple time periods in adolescence. Wang et al [21] found that the progression to regular smoking among both 12- to 15-year-olds and 16- to 18-yearolds was associated with being white, positive beliefs about smoking, depression, and not liking school. Additional predictors for those aged 12–15 years included poor academic performance; irregular school attendance; and having parents, friends, and teachers who smoked. Although that study provides a useful initial step in terms of understanding stability and change in the risk factors for regular smoking over time, its limitations include the confounding of smoking initiation with escalation and the possibility that age-related differences may have been owing to cohort effects, rather than developmental changes. The present study uses longitudinal data to identify predictors of the transition to regular smoking among initial experimenters during middle adolescence (grades 8 to 10), late adolescence (grades 10 to 12), and young adulthood (grade 12 to age 23 years). Based on prevalence rates of regular smoking by race/ethnicity and gender [1], Whites were expected to be more likely than other racial/ethnic groups to make the transition to regular smoking during each period, although significant gender differences were not necessarily expected. Prosmoking attitudes were also expected to be risk factors for the transition to regular smoking. Although Chassin et al [20] found age-related differences in smoking attitudes, we did not necessarily expect that their effect on the transition to regular smoking would differ between adolescence and young adulthood. In terms of exposure to a prosmoking environment, we anticipated that family influences (household smoking, parental approval of smoking) would be weaker predictors of
316
TUCKER ET AL
the transition to regular smoking than peer influences (perceived prevalence of peer smoking, friend smoking, and approval of smoking) at all three assessment periods. Based on Jessor and Jessor’s Problem Behavior Theory, we further hypothesized that adolescents who engaged in other problem behaviors (drinking, delinquency) had a weaker academic orientation, and poorer relations with parents would be more likely to transition to regular smoking, although we did not expect these associations to persist into young adulthood. Finally, a number of studies have linked emotional distress with smoking [22–24]. Based on this work, we expected that initial experimenters with poorer mental health would be more likely to transition to regular smoking during both adolescence and young adulthood.
Methods Participants Participants in the original full sample were 6527 males and females in the RAND Adolescent/Young Adult Panel Study, conducted to evaluate the effectiveness of the Project ALERT drug use prevention program for middle school children [25]. Students were recruited from 30 California and Oregon schools in grade 7 (1985) and assessed again in grades 8, 9, 10, and 12 and at age 23 years (1995). The schools were chosen to represent a wide range of community types (urban, suburban, and rural), socioeconomic status (18 schools drew from neighborhoods with household incomes below the median for their state), and racial and ethnic composition (nine schools had minority populations of 50% or more). This study focuses on three transition periods: grades 8 to 10 (1986 to 1988), grades 10 to 12 (1988 to 1990), and grade 12 to age 23 (1990 to 1995). Extensive tracking methods allowed us to retain 85% of the original baseline in 1988 (n ⫽ 5523), 67% of the sample in 1990 (n ⫽ 4390), and 53% of the sample in 1995 (n ⫽ 3466). We restricted our sample to adolescents who reported “ever smoking but less than weekly” at the beginning of each transition period. Additional individuals were dropped from analyses if they were missing smoking or ethnicity information. These exclusions resulted in final sample sizes for each transition period of n ⫽ 2496, n ⫽ 2149, and n ⫽ 1534, respectively. Of the final sample for the grades 10 to 12 transition period, which was weighted to reduce the bias associated with attrition, 37% were self-classified as a minority (13% African-
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
American; 12% Hispanic; 8% Asian; 4% Other), 47% were female, and 16% dropped out of school by the grade 12 assessment (the demographic profiles of the other two subsamples are nearly identical). Participants who could not be located or had withdrawn from the study differed from those who were retained on demographic characteristics, substance use, and grades [26]. Previous work has shown that the use of sample weights is the best method of adjusting for attrition in this school-based sample [27], and weights developed to deal with attrition by 1990 have been found to remove 90% or more of the bias exhibited in the unweighted sample [28]. For the present analyses, the sample was weighted to reduce bias associated with attrition occurring at each assessment. The weighting was conducted by using logistic regression to create predicted probabilities of responding to the 1988, 1990, and 1995 surveys that were derived from 7th grade information about each 7th grade respondent (including race, gender, family structure, delinquency, substance use, and grades). Measures A comprehensive review of previous empirical and theoretical work guided the development of the survey instruments for this project. Almost all of the items and measures have been successfully used in national surveys of adolescent drug use, although some were modified to match the reading level and experience of the sample. Smoking status. Participants were eligible for the analyses if they had “ever smoked but less than weekly” at the beginning of each transition period. Weekly smoking was defined as “having smoked at least 3 to 5 days in the past month.” Participants were considered to have made the transition to regular smoking at each follow-up if they reported weekly or more frequent smoking. Most of the experimenters were not current smokers at the beginning of each transition period, with the weighted percent who had not smoked in the past 30 days steadily increasing from grade 8 (83%) to grade 10 (87%) to grade 12 (92%). Demographics. Participants’ gender, race/ethnicity (white vs. African-American, Hispanic, Asian, and Other), age, and educational attainment of their father and/or mother at the beginning of each transition period were included as control variables in the main analyses.
April 2003
Smoking beliefs. Smoking resistance self-efficacy (RSE) at grade 8 was assessed by asking participants what they would do if they were offered a cigarette on a date and did not want it (“1” ⫽ “I would tell my date no” to “3” ⫽ “I would smoke the cigarette”) and how strongly they agreed with the statement “If all my friends were smoking cigarettes at a party, I’d feel pretty out of it unless I smoked, too” (“1” ⫽ “strongly disagree” to “4” ⫽ “strongly agree”). Two additional items at grade 10 assessed whether they could resist smoking if offered a cigarette when their best friend was smoking and they were bored. Four items at grade 12 indicated whether they could resist smoking when their best friend was smoking, they were bored, they were at a party at which all their friends were smoking, and their date was smoking (“1” ⫽ “very sure I could” to 5 ⫽ “not at all sure I could”) (␣ ⫽ .73 to .96). Participants indicated whether they intended to smoke in the next 6 months (“1” ⫽ “definitely no” to “4” ⫽ “definitely yes”). Single items were used to assess the harmfulness of smoking cigarettes occasionally (“1” ⫽ “a lot” to “4” ⫽ “not at all”) and whether they might get addicted if they smoked cigarettes every weekend (“1” ⫽ “strongly agree” to “4” ⫽ “strongly disagree”). Beliefs about the “nonhealth consequences of smoking” were assessed by four items at grades 8 and 10 and seven items at grade 12 (e.g., it relaxes you; makes you feel more at ease with others). Higher scores indicated more positive beliefs (grade 8: ␣ ⫽ .66; grade 10: ␣ ⫽ .65; grade 12: ␣ ⫽ .77). Smoking environment. We assessed five aspects of the participants’ smoking-related social environment: friend smoking; perceived prevalence of peer smoking; friend and parent approval of smoking; and any household smoking. Friend smoking was assessed using two items: how often they are with people their age who are smoking (“1” ⫽ “never” to “4” ⫽ “often”) and whether (grade 8) or how often (grades 10 and 12; “1” ⫽ “never” to “4” ⫽ “often”) their best friend smokes (␣ ⫽ .69 at grade 8; ␣ ⬎ .70 at grades 10 and 12). Perceived prevalence of peer smoking at grades 10 and 12 was assessed by asking participants to estimate the percentage of their classmates who had smoked in the last month. Parental and friend approval of smoking were rated in terms of how each would feel if they found out that the participant smoked (parent rating: “1” ⫽ “very upset” to “4” ⫽ “not at all upset or would not care”; friend rating: “1” ⫽ “they would disapprove and stop being my friends” to “4” ⫽ “they would approve or would not care”). Any household smoking
PREDICTORS OF TRANSITION TO REGULAR SMOKING
317
at grades 10 and 12 was assessed in terms of whether any of these individuals smoked (“0” ⫽ “no”, “1” ⫽ “yes”): an older sibling, the adult male and female who are most important to them (presumably the parents). Alcohol consumption and deviance. Participants provided information on their history and frequency of alcohol use; this information was recoded on a scale ranging from “0” (“never used”) to “11” (“used 20 or more days in past month”). Engagement in delinquent behavior was assessed by four items at grade 8 and eight items at grade 10 (e.g., school problems, stealing, broken into or damaged property: “0” ⫽ “not at all” to “3” ⫽ “more than three times in past year”; ␣ ⫽ .63 at grade 8 and .78 at grade 10, respectively). At grade 12 participants indicated how often they had engaged in each of 26 problem behaviors (e.g., school problems, stealing, violent behavior, damaging property, selling drugs: “0” ⫽ “not at all” to “5” ⫽ “20 or more times in past year”; ␣ ⫽ .86). Academic orientation. Participants provided information on their grades at grade 10 (“1” ⫽ “mostly A’s” to “5” ⫽ “mostly F’s”) and grade 12 (“1” ⫽ “mostly A’s” to “8” ⫽ “mostly below D”). They also indicated their academic intentions at grades 10 and 12 in terms of how much schooling they anticipated completing (“1” ⫽ “attend graduate or professional school” to “6” ⫽ “some high school.” Because this information was not available at the grade 8 assessment, we used grade 7 information as a reasonable proxy for grade 8 academic orientation (grades: “1” ⫽ “mostly A’s” to “5” ⫽ “mostly F’s”; academic intentions: “1” ⫽ “attend graduate or professional school” to “5” ⫽ “may not finish high school”). Parental bonding. Participants indicated whether they were living with both biological parents (“0” ⫽ “no,” “1” ⫽ “yes”); not having an intact nuclear family would suggest that the child’s bond with one or both biological parents was weakened by diminished contact. At grade 8 parental support was assessed in terms of whether participants indicated that they would be most likely to talk to their parents about a personal problem (“0” ⫽ “yes”; “1” ⫽ “no”). The measure of parental support contained three items at grade 10 (parents listen to my opinions, respect my feelings, don’t trust my judgment) with three items added at grade 12 (parents keep track of how I am doing in school, sense when I am upset, know where I am and what I am doing: “1” ⫽
318
TUCKER ET AL
“strongly agree” to “4” ⫽ “strongly disagree”). Higher scores indicate weaker parental support (␣ ⬎ .70 at grades 10 and 12). Mental health. Perceived mental health status was assessed at grades 10 and 12 by the mental health subscale (MHI-5) of the Medical Outcomes Study Short-Form Health Survey (SF-20) [29]. Items ask how much of the time, during the past month, participants have: “been a very nervous person”; “been a happy person”; “felt calm and peaceful”; “felt downhearted and blue”; “felt so down in the dumps that nothing could cheer them up” (“1” ⫽ “all of the time” to “6” ⫽ “none of the time”). Higher scores on the scale indicated poorer perceived mental health (␣ ⫽ .50 at grade 10; ␣ ⫽ .83 at grade 12).
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
as covariates in each of these predictor models. Trimmed models that contained only variables that were at least marginally significant (p ⬍ .10) predictors of the transition to regular use were then examined. These trimmed models were also run controlling for smoking frequency and the number of years the participant had smoked to determine whether the psychosocial and behavioral variables remained significant predictors of the transition to regular smoking after accounting for these indicators of smoking dependence. We checked for any effect of participation in the drug prevention program in grades 7 and 8 by adding treatment to the relevant models; because it was not significant, we have not included it in the tables of results.
Statistical Methods
Results
Participants at each wave of data collection generally completed the entire survey. However, there was a small amount of missing data for most variables (1% to 6%), which can result in a sizable sample reduction for multivariate analyses that require complete information for all included variables. To circumvent this problem, we used least squares regression imputation on continuous predictors and logistic regression imputation on dichotomous predictors, with the exception of race/ethnicity and smoking status [30]. These imputation procedures generated a predicted value (the criterion) from other variables that were strongly correlated with the criterion value (e.g., demographic characteristics, substance use, academic performance, deviant behaviors) and then substituted the predicted value for the missing value. The clustering of observations within schools and use of sample weights were accounted for by computing Huber standard errors [31,32]. Huberized regression techniques were used to determine the independent associations of the predictor variables on subsequent smoking status. Logistic regression analyses were initially conducted to identify predictors of the transition to regular smoking. Given the large number of predictor variables in the analyses, seven separate models were tested. The initial model included demographic variables only (gender, ethnicity, parental education, and age). The remaining six models each included one of the conceptually related groups of psychosocial and behavioral predictors (smoking beliefs, smoking environment, alcohol consumption and deviance, academic orientation, parental bonding, and mental health). Demographic variables were entered
Prevalence of Transitioning to Regular Smoking The number (and weighted percent) of initial experimenters who became regular smokers by the end of each transition period was 473 (19%) individuals from grades 8 to 10, 329 (16%) individuals from grades 10 to 12, and 259 (17%) individuals from grade 12 to age 23 years. Predictors of the transition to regular smoking are shown in Table 1 for the period from grades 8 to 10 (middle adolescence), Table 2 for grades 10 to 12 (late adolescence), and Table 3 for grade 12 to age 23 years (young adulthood). Trimmed final models are shown in Table 4. Demographics. Ethnicity was a significant predictor of the transition to regular smoking during adolescence but not young adulthood. Results from the trimmed models indicated that African-Americans, Hispanics, and Asians were all at significantly lower risk, compared with Whites, of transitioning to regular smoking during middle adolescence. The lower risk associated with being Hispanic and Asian dissipated during late adolescence, and the lower risk associated with being African-American dissipated during young adulthood. Younger smokers were at higher risk of transitioning to regular smoking but only during young adulthood. Gender and parental education were not significant predictors of the transition to regular smoking. Smoking attitudes. At all ages, a stronger intention to smoke in the future was a risk factor for transitioning to regular smoking. Those with lower RSE were also more likely to escalate their smoking in adolescence, although not in young adulthood. Pro-
April 2003
PREDICTORS OF TRANSITION TO REGULAR SMOKING
319
Table 1. Predictors of the Transition to Regular Smoking From Grade 8 to Grade 10 (n ⫽ 2496) Grade 8 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
1.16 0.36** 0.72 0.55** 1.17 1.14 0.98
[0.90, [0.19, [0.48, [0.40, [0.65, [0.93, [0.86,
1.49] 0.68] 1.07] 0.76] 2.09] 1.41] 1.12]
1.86
(1.05)
0.00 –3.00
1.22* 1.42*** 1.19* 1.19* 1.30*
[1.04, [1.23, [1.04, [1.05, [1.06,
1.44] 1.63] 1.37] 1.36] 1.60]
0.02 1.68 3.28 1.76 1.51
(0.84) (0.82) (0.83) (0.96) (0.61)
⫺0.81–2.66 0.78 – 4.00 1.00 – 4.24 1.00 – 4.00 0.93– 4.00
1.74*** N/A 0.93 1.20* N/A
[1.44, 2.10]
⫺0.02
(0.78)
⫺1.18 –1.50
[0.79, 1.09] [1.05, 1.38]
2.37 1.39
(0.72) (0.68)
1.00 – 4.00 0.88 – 4.00
1.08*** 1.22*
[1.04, 1.12] [1.04, 1.43]
4.27 0.84
(3.09) (0.68)
0.00 –11.00 0.00 –3.00
1.60*** 1.13*
[1.42, 1.79] [1.03, 1.24]
2.33 2.14
(0.83) (0.98)
1.00 –5.00 1.00 –5.00
0.89 1.61*** N/A
[0.70, 1.14] [1.29, 2.00]
0.54 0.55
(0.50) (0.50)
0.00 –1.00 0.00 –1.00
* p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
a
smoking beliefs were risk factors at all three time periods, although results differed, depending on the type of belief. Contrary to expectations, smokers holding a stronger belief in the harmful effects of smoking were more likely to transition to regular smoking during both adolescence and young adulthood. However, these associations were owing to a suppression effect; believing that smoking is harmful was weakly related to smoking escalation when controlling only for demographic variables (all p ⬎ .20; results not shown). Smokers with a weaker belief in the addictiveness of smoking and those with a stronger belief in the positive nonhealth consequences of smoking were more likely to transition to regular smoking. However, controlling for other nondemographic and nonbelief variables in the trimmed models eliminated the association with addictiveness beliefs in young adulthood and the association with beliefs in the positive nonhealth consequences of smoking in adolescence.
Smoking environment. Friend smoking was a risk factor for transitioning to regular smoking during both adolescence and young adulthood when controlling for demographics and other aspects of the smoking environment. However, friend smoking remained a significant predictor only in middle adolescence when controlling for other variables in the trimmed models. Similarly, parental approval of smoking was a risk factor during both adolescence (significant) and young adulthood (marginally significant) when controlling for demographic and smoking environment variables. Controlling for other types of variables in the trimmed models eliminated the association at all three transition periods. Finally, the perceived prevalence of peer smoking, friend approval of smoking, and household smoking were not significant predictors of the transition to regular smoking (note that information on the perceived prevalence of peer smoking and household smoking were not available at grade 8).
TUCKER ET AL
320
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
Table 2. Predictors of the Transition to Regular Smoking From Grade 10 to Grade 12 (n ⫽ 2149) Grade 10 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
0.96 0.43** 0.80 0.79 1.06 0.94 0.96
[0.68, [0.24, [0.50, [0.42, [0.42, [0.67, [0.86,
1.35] 0.76] 1.28] 1.51] 2.68] 1.32] 1.06]
3.16
(1.49)
1.00 – 6.00
1.24* 1.61*** 1.19* 1.23* 1.32*
[1.03, [1.32, [1.02, [1.05, [1.03,
1.48] 1.96] 1.39] 1.45] 1.71]
⫺0.10 1.53 3.31 1.70 1.70
(0.66) (0.74) (0.80) (0.92) (0.57)
⫺0.65–2.49 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00 1.00 –3.57
1.37** 1.00 1.12 1.24* 1.29
[1.11, [1.00, [0.90, [1.03, [0.88,
1.69] 1.01] 1.41] 1.49] 1.90]
1.98 53.63 2.39 1.54 0.63
(0.81) (24.02) (0.67) (0.79) (0.48)
1.00 – 4.00 0.00 –100.00 1.00 – 4.00 1.00 – 4.00 0.00 –1.00
1.12*** 1.63**
[1.07, 1.18] [1.26, 2.10]
5.32 0.43
(3.11) (0.44)
0.00 –11.00 0.00 –3.00
1.17† 1.03
[0.99, 1.38] [0.90, 1.17]
2.37 2.57
(0.84) (1.35)
1.00 –5.00 1.00 – 6.00
0.66* 1.40*** 1.28***
[0.47, 0.92] [1.22, 1.59] [1.16, 1.42]
0.53 2.08 2.66
(0.50) (0.81) (0.97)
0.00 –1.00 1.00 – 4.00 1.00 – 6.00
p ⬍ .10; * p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
† a
Other problem behaviors. Greater alcohol consumption and deviant behavior were risk factors for the transition to regular smoking during middle and late adolescence but not during young adulthood. Controlling for other variables in the trimmed models, only alcohol consumption in late adolescence remained a significant risk factor for transitioning to regular smoking. Academic orientation. Poor grades and low academic intentions were significant risk factors for regular smoking during middle adolescence and young adulthood, and poor grades was a marginally significant risk factor in late adolescence. However, only in middle adolescence did these associations remain significant when controlling for other variables in the trimmed models. Parental bonding. Not having an intact nuclear family was a significant risk factor for transitioning to regular smoking during late adolescence only.
Poorer parental support was a risk factor during both adolescence and young adulthood, although only the associations in adolescence remained significant after controlling for other variables in the trimmed models. Mental health. Mental health information was unavailable at grade 8. Poorer mental health was a risk factor for regular smoking during late adolescence and young adulthood, although these associations did not remain significant in the final trimmed models. Controlling for Frequency and Length of Smoking Smoking frequency and the number of years smoking were added to the trimmed models. The significant associations shown in Table 4 remained at least marginally significant (p ⬍ .10) after controlling for these indicators of smoking dependence with the
April 2003
PREDICTORS OF TRANSITION TO REGULAR SMOKING
321
Table 3. Predictors of the Transition to Regular Smoking From Grade 12 to Age 23 (n ⫽ 1534) Grade 12 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
1.04 0.94 0.83 0.78 1.86 0.76† 1.02
[0.75, [0.63, [0.41, [0.58, [0.80, [0.57, [0.93,
1.44] 1.41] 1.67] 1.06] 4.33] 1.02] 1.11]
4.13
(2.24)
0.82–9.00
1.11 1.51*** 1.22* 1.15* 1.90***
[0.95, [1.27, [1.02, [1.02, [1.38,
1.30] 1.81] 1.46] 1.30] 2.60]
1.47 1.35 3.13 1.71 1.64
(1.00) (0.65) (0.83) (0.95) (0.50)
0.61–5.00 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00
1.46*** 1.00 0.92 1.19† 1.30
[1.20, [0.99, [0.75, [0.97, [0.85,
1.76] 1.00] 1.13] 1.44] 1.98]
2.08 51.88 2.41 1.72 0.58
(0.88) (24.24) (0.64) (0.89) (0.49)
1.00 – 4.00 0.00 –100.00 1.00 – 4.00 1.00 – 4.00 0.00 –1.00
1.05 1.30
[0.97, 1.14] [0.69, 2.44]
5.58 0.31
(3.16) (0.33)
0.00 –11.00 ⫺0.08 –2.15
1.14* 1.19*
[1.01, 1.28] [1.03, 1.36]
3.42 2.52
(1.54) (1.31)
1.00 – 8.00 1.00 – 6.00
1.16 1.21* 1.20*
[0.86, 1.57] [1.01, 1.45] [1.02, 1.40]
0.54 1.96 2.41
(0.50) (0.65) (0.94)
0.00 –1.00 1.00 – 4.00 1.00 – 6.00
p ⬍ .10; * p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
† a
following exceptions: intentions to smoke in both middle adolescence [odds ratio (OR) ⫽ 1.04, not significant] and late adolescence (OR ⫽ 1.02, not significant), as well as friend smoking in middle adolescence (OR ⫽ 1.17, not significant).
Discussion Whites were more likely than African-Americans, Hispanics, and Asians to transition from experimentation to regular smoking during middle adolescence. The difference between Whites and AfricanAmericans persisted into late adolescence, although there were no racial/ethnic group differences by early adulthood. This pattern of results is consistent with national surveys [2], as well as previous research focusing specifically on racial differences in the transition to regular smoking [21]. The lower rate of smoking escalation among African-Americans remained after controlling for demographic factors,
smoking attitudes and environment, engagement in other problem behaviors, and level of parental support. At least one other study has similarly found that such factors cannot explain this racial difference in smoking behavior [33]. Previous work has suggested that black–white differences in smoking may be owing to the underreporting of smoking by African-Americans [34,35]. In the present study, selfreported smoking in grade 8 was found to be highly accurate when validated against a biochemical marker of nicotine uptake, serum cotinine [36], and there is little reason to believe that reports of smoking became less valid over time. Thus, it is more likely that this difference is owing to a complex set of psychosocial and cultural influences that we could not fully assess in this study, rather than to the validity of self-reported smoking. It is interesting that the only demographic characteristic associated with the transition to regular smoking in young adulthood is age, with younger participants being
322
TUCKER ET AL
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
Table 4. Trimmed and Fully Adjusted Models: Predictors of the Transition to Regular Smoking Grade 8 –10
Grade 10 –12
Predictor Variables
OR
Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Friend smoking Parent approval of smoking Alcohol consumption Deviance Poor grades Low academic intentions Intact nuclear family Poor parental support Poor mental health
1.17 0.40** 0.60* 0.62** 1.25 1.02 1.06 1.20* 1.25* 1.22** 1.14* 1.18 1.31* 1.09 1.03 0.94 1.50*** 1.09*
[0.88, [0.21, [0.38, [0.44, [0.72, [0.83, [0.93, [1.03, [1.05, [1.06, [1.00, [0.96, [1.05, [0.94, [0.99, [0.80, [1.34, [1.00,
[95% CI]
1.32* N/A
[1.07, 1.63]
1.54] 0.77] 0.93] 0.87] 2.19] 1.25] 1.20] 1.39] 1.48] 1.41] 1.30] 1.44] 1.65] 1.25] 1.07] 1.10] 1.69] 1.19]
OR
Grade 12 to Age 23
[95% CI]
0.97 0.46* 0.75 0.92 1.04 0.98 0.97 1.31** 1.42** 1.16* 1.18* 1.11 1.14 1.10 1.08** 1.17 0.97
[0.69, [0.25, [0.46, [0.46, [0.38, [0.69, [0.87, [1.08, [1.14, [1.01, [1.02, [0.89, [0.96, [0.94, [1.02, [0.89, [0.83,
1.35] 0.86] 1.20] 1.85] 2.86] 1.39] 1.08] 1.59] 1.76] 1.34] 1.38] 1.39] 1.37] 1.30] 1.14] 1.55] 1.12]
0.65* 1.26** 1.07
[0.47, 0.91] [1.08, 1.47] [0.95, 1.21]
OR
[95% CI]
1.08 1.11 0.72 0.85 1.56 0.68* 1.06
[0.77, [0.71, [0.36, [0.54, [0.62, [0.49, [0.96,
1.51] 1.72] 1.42] 1.34] 3.89] 0.93] 1.18]
1.51*** 1.20* 1.11 1.72** 1.21 1.04
[1.25, [1.00, [0.97, [1.24, [0.99, [0.85,
1.82] 1.44] 1.28] 2.38] 1.48] 1.26]
1.10 1.15
[0.98, 1.24] [0.99, 1.34]
0.92 1.07
[0.75, 1.14] [0.90, 1.28]
Note. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy. * p ⬍ .05; ** p ⬍.01; *** p ⬍ .001. With the exception of demographic variables, only those variables associated with the transition to regular smoking at p ⬍ .10 in Tables 1–3 are included in these trimmed models.
more likely to escalate their smoking from grade 12 to age 23 years. It is unclear why experimenters who were young for their cohort escalated their smoking. One possibility is that they wanted to look more mature to their peers; however, if this were the case, one would expect to find this effect in adolescence as well. We expected that participants in our study had already formed strong smoking-related attitudes through their direct experience with the behavior and that these attitudes would be important antecedents of the transition to regular smoking. Indeed, prosmoking attitudes were the most consistent predictors of the transition to regular smoking over time in this study. In the analyses controlling for demographic variables and other smoking-related attitudes, experimenters who transitioned to regular smoking tended to have more difficulty resisting social pressures to smoke, stronger intentions to smoke in the future, weaker beliefs in the addictiveness of smoking, and stronger beliefs in the positive nonhealth consequences of smoking. The only exception to this pattern was found for RSE, which did not predict regular smoking in young adulthood. Although it may be argued that this lack of association is due to the longer (5-year) follow-up period that was used for the young adulthood analyses, the fact
that all other smoking-related attitudes remained significant risk factors in adulthood suggests an alternate explanation: Difficulty resisting social pressures to smoke plays a lesser role in the escalation of smoking over time because the peer pressures felt by adolescents to engage in substance use wane as they enter young adulthood. Attesting to the importance of smoking-related attitudes in the transition to regular smoking, these attitude variables remained significantly associated with smoking escalation in the trimmed models with only a few exceptions (positive nonhealth consequences of smoking during adolescence and the addictiveness of smoking in young adulthood). Supporting our hypothesis, exposure to friend smoking was associated with the transition to regular smoking over time, whereas no such associations were found for household smoking. However, perceived prevalence of peer smoking and approval of one’s smoking by friends were not associated with the transition to regular smoking. Given that both of these factors have been associated with smoking initiation [10], it may be the case that once an individual has begun experimenting with cigarettes, perceiving that others are smoking and that one’s friends approve of this behavior have little effect on whether the individual maintains or escalates his or
April 2003
her smoking over time. It is particularly noteworthy that friend smoking at grade 12 was associated with the transition to regular smoking by age 23 years in light of the significant changes in social networks that often occur during the transition out of high school. Perhaps experimenters who escalate their smoking during this period are choosing friends in part on the basis of their smoking behavior; previous research has provided some evidence for such a selection process [12,37]. In interpreting these findings on friend smoking, it is important to keep in mind that the associations in late adolescence and early adulthood became weaker and nonsignificant after controlling for other variables in the multivariate models. Although household smoking and parental approval of smoking were not significantly associated with the transition to regular smoking, it appears that family influences still play a role in smoking escalation. Parental smoking may be most relevant to the initiation process because it provides easy access to cigarettes and opportunities for experimentation. Likewise, parental disapproval of smoking may be a deterrent for initiating smoking, but, among those who have already started, other factors are likely more relevant to the escalation of use. However, experimenters who reported weaker family bonds were more likely to transition into regular smoking over the follow-up period. Thus, parents may play a role in smoking escalation to the extent that their children smoke in an attempt to mitigate the stress of a disrupted or discordant family environment [38]. Previous research has shown that adolescent depression or emotional distress is prospectively associated with adult smoking [23,39]. Results from this study further indicate that poorer mental health among young smokers is a risk factor for smoking escalation, although these associations did not remain significant in multivariate models. Alcohol consumption, engagement in deviant behavior, and weak academic orientation predicted the transition to regular smoking during adolescence but were not relevant risk factors during young adulthood. This finding is consistent with Problem Behavior Theory [16], which postulates that adolescent substance use often coexists with other problem behaviors because they have similar etiologies and serve common purposes. It also suggests that engagement in these other problem behaviors may play a role in the escalation of smoking during adolescence, suggesting that smoking prevention programs need to address engagement in other types of substance use and problem behaviors as important risk
PREDICTORS OF TRANSITION TO REGULAR SMOKING
323
factors. However, it is important to keep in mind that only academic orientation in middle adolescence and alcohol consumption in late adolescence remained significant predictors of the transition to regular smoking in multivariate models, indicating that the potential influence of engagement in problem behaviors on smoking escalation is overshadowed by other factors. Further, strong conclusions cannot be drawn about declines in the relevance of these factors over time because of the shorter follow-up period for the adolescent vs. young adult analyses.
Limitations Several study limitations should be noted. Results are based on a single sample from Oregon and California and may thus have limited generalizability. However, the schools from which participants were recruited were chosen to represent a broad spectrum of communities, socioeconomic status, and racial composition. Sample attrition is another limitation and necessitated the use of sample weights to effectively reduce attrition-related bias. Data are based on self-report and subject to the usual validity concerns; however, current tobacco use in this sample has been shown to be highly accurate when externally validated [36]. Finally, the longer length of follow-up for the young adult vs. adolescent analyses complicates the interpretation of our results in terms of possible age-related declines in the importance of certain risk factors. However, those variables that emerged as significant risk factors in the 5-year period from grade 12 to age 23 years are likely quite important to the smoking escalation process. Results from this study point to a number of smoking-related attitudes, social influences, and behaviors that prevention efforts may target to curb the escalation of smoking among experimenters. Altering prosmoking beliefs may be particularly important for both adolescent and young adult experimenters. This research was supported by University of California TobaccoRelated Disease Research Program Grant 10RT-0062.
References 1. Centers for Disease Control and Prevention. Youth risk behavior surveillance—United States, 1999. CDC Surveill Summ MMWR 2000;49:1–96. 2. Johnston LD, O’Malley PM, Bachman JG. Monitoring the Future national survey results on drug use, 1975–1999. Vol. I. Secondary school students. NIH Pub. No. 00-4802. Bethesda, MD: National Institute on Drug Abuse, 2000.
324
TUCKER ET AL
3. Fergusson DM, Horwood LJ. Transitions to cigarette smoking during adolescence. Addict Behav 1995;20:627–42. 4. Breslau N, Peterson EL. Smoking cessation in young adults: Age at initiation of cigarette smoking and other suspected influences. Am J Public Health 1996;86:214 –20. 5. Ellickson PL, Tucker JS, Klein DJ. High-risk behaviors associated with early smoking: Results from a 5-year follow-up. J Adolesc Health 2001;28:465–73. 6. Ary DV, Biglan A. Longitudinal changes in adolescent cigarette smoking behavior: Onset and cessation. J Behav Med 1988;11:361–81. 7. Botvin GJ, Baker E, Botvin EM, et al. Factors promoting cigarette smoking among black youth: A causal modeling approach. Addict Behav 1993;18:397–405. 8. Flay BR, Hu FB, Siddiqui O, et al. Differential influence of parental smoking and friends’ smoking on adolescent initiation and escalation of smoking. J Health Soc Behav 1994;35: 248 –65. 9. Robinson LA, Klesges RC, Zbikowski SM, et al. Predictors of risk for different stages of adolescent smoking in a biracial sample. J Consult Clin Psychol 1997;65:653–62. 10. Conrad KM, Flay BR, Hill D. Why children start smoking cigarettes: Predictors of onset. Br J Addict 1992;87:1711–24. 11. Chassin L, Presson CC, Sherman SJ, et al. Predicting the onset of cigarette smoking in adolescents: A longitudinal study. J Appl Soc Psychol 1984;14:224 –43. 12. West P, Sweeting H, Ecob R. Family and friends’ influences on the uptake of regular smoking from mid-adolescence to early adulthood. Addiction 1999;94:1397–1412. 13. Griffin KW, Botvin GJ, Doyle MM, et al. A six-year follow-up study of determinants of heavy cigarette smoking among high-school seniors. J Behav Med 1999;22:271–84. 14. Weinstein ND. Testing four competing theories of healthprotective behavior. Health Psychol 1993;12:324 –33. 15. Krause SJ. Attitudes and the prediction of behavior: A metaanalysis of the empirical literature. Pers Soc Psychol Bull 1995;21:58 –75. 16. Jessor R, Jessor SL. Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth. New York: Academic Press, 1977. 17. Larson RW, Richards MH, Moneta G, et al. Changes in adolescents’ daily interactions with their families from ages 10 to 18: Disengagement and transformation. Dev Psychol 1996; 32:744 –54. 18. Krosnick JA, Judd CM. Transitions in social influence at adolescence: Who induces cigarette smoking? Dev Psychol 1982;18:359 –68. 19. Chassin L, Presson CC, Sherman SJ, et al. Changes in peer and parent influence during adolescence. Longitudinal versus cross-sectional perspectives on smoking initiation. Dev Psychol 1986;22:327–34. 20. Chassin L, Presson CC, Rose JS, et al. From adolescence to adulthood: Age-related changes in beliefs about cigarette smoking in a Midwestern community sample. Health Psychol 2001;20:377–86.
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
21. Wang MQ, Fitzhugh EC, Green BL, et al. Prospective socialpsychological factors of adolescent smoking progression. J Adolesc Health 1999;24:2–9. 22. Escobedo LG, Reddy M, Giovino GA. The relationship between depressive symptoms and cigarette smoking in U.S. adolescents. Addiction 1998;93:433–40. 23. Orlando M, Ellickson PL, Jinnett K. The temporal relationship between emotional distress and cigarette smoking during adolescence and young adulthood. J Consult Clin Psychol 2001;69:959 –70. 24. Patton GC, Hibbert M, Rosier MJ, et al. Is smoking associated with depression and anxiety in teenagers? Am J Public Health 1996;86:225–30. 25. Ellickson PL, Bell RM. Drug prevention in junior high: A multi-site longitudinal test. Science 1990;247:1299 –305. 26. Collins RL, Ellickson PL, Bell RM. Simultaneous polydrug use among teens: Prevalence and predictors. J Subst Abuse 1999; 10:233–53. 27. McGuigan KA, Ellickson PL, Hays RD, et al. Adjusting for attrition in school-based samples: Bias, precision, and cost trade-offs of three methods. Eval Rev 1997;21:554 –67. 28. Ellickson PL, Saner H, McGuigan KA. Profiles of violent youth: Substance use and other concurrent problems. Am J Public Health 1997;87:985–91. 29. Stewart AL, Hays RD, Ware JE, Jr. The MOS Short-Form General Health Survey. Med Care 1988;26:724 –35. 30. Little RJA. Regression with missing X’s: A review. J Am Stat Assoc 1992;87:1227–37. 31. Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the fifth Berkeley Symposium on Mathematical Statistics Probability 1967;1: 221–33. 32. Stata Corp. Stata Statistical Software: Release 6.0. College Station, TX: Stata Corp., 1999. 33. Flint AJ, Yamada EG, Novotny TE. Black-white differences in cigarette smoking uptake: Progression from adolescent experimentation to regular use. Prev Med 1998;27:358 –64. 34. Bauman KE, Ennett SE. Tobacco use by black and white adolescents: The validity of self-reports. Am J Public Health 1994;84:394 –8. 35. Wagenknecht LE, Burke GL, Perkins LL, et al. Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. Am J Public Health 1992;82:33–6. 36. Reinisch E, Bell RM, Ellickson PL. How accurate are adolescent reports of drug use? Report No. N-3198-CHF. Santa Monica, CA: RAND, 1991. 37. Ennett ST, Baumann KE. The contribution of influence and selection to adolescent peer group homogeneity: The case of adolescent cigarette smoking. J Pers Soc Psychol 1994;6:653– 63. 38. Covey LS, Tam D. Depressive mood, the single-parent home, and adolescent cigarette smoking. Am J Public Health 1990; 80:1330 –3. 39. Kandel DB, Davies M. Adult sequelae of adolescent depressive symptoms. Arch Gen Psychiatry 1986;43:255–62.
ORIGINAL ARTICLE
Predictors of the Transition to Regular Smoking During Adolescence and Young Adulthood JOAN S. TUCKER, Ph.D., PHYLLIS L. ELLICKSON, Ph.D., AND DAVID J. KLEIN, M.S.
Purpose: To identify predictors of the transition from experimentation to regular smoking in middle adolescence, late adolescence, and young adulthood. Methods: California and Oregon students completed self-report surveys assessing the following potential predictors of the transition to regular smoking from grades 8 to 10 (n ⴝ 2496), grades 10 to 12 (n ⴝ 2149), and grade 12 to age 23 years (n ⴝ 1534): demographic characteristics; smoking-related attitudes, behaviors and environment; other problem behaviors; academic orientation; parental bonding; and mental health. Huberized regression techniques, which adjust for weighting and clustering of observations, were used to determine the independent associations of the predictor variables on subsequent smoking status. Results: Risk factors for the transition to regular smoking during middle adolescence included being white, prosmoking attitudes, friend smoking, weak academic orientation, and less parental support. During late adolescence, being African-American was protective, whereas risk factors included prosmoking attitudes, drinking, non-intact nuclear family, and less parental support. Risk factors in young adulthood included younger age and prosmoking attitudes. Conclusions: Results point to several smoking-related attitudes, social influences, and behaviors that prevention efforts may target to curb the escalation of smoking. © Society for Adolescent Medicine, 2003 KEY WORDS:
Adolescents Escalation Longitudinal Smoking Young adults
From RAND, Santa Monica, California. Address correspondence to: Joan S. Tucker, Ph.D., RAND, 1700 Main Street, P.O. Box 2138, Santa Monica, CA 90407-2138. Manuscript accepted November 1, 2002. 1054-139X/03/$–see front matter doi:10.1016/S1054-139X(02)00709-7
Experimentation with cigarette smoking is common during adolescence, with 78% of high school students having tried smoking by grade 12 [1]. Although most of these adolescents never become committed smokers, a sizable minority of those who experiment with cigarettes progress to regular smoking during the adolescent years. For example, two national surveys found that more than 20% of 12th graders reported smoking daily or almost daily in the past 30 days [1,2]. Regular smoking is much more prevalent among Whites than African-Americans or Hispanics and slightly more prevalent among male than female adolescents [1]. Although research examining the predictors of smoking has focused largely on smoking onset or initiation, it is important to understand the factors contributing to the escalation of smoking for several reasons. Smoking transitions tend toward heavier use; it is rare for adolescents who smoke regularly to revert to only occasional use [3]. Adolescents who are more committed smokers are also less likely to quit smoking than other adolescent smokers [4]. Furthermore, our own work has shown that adolescent smokers at grade 7 are at higher risk than experimenters both concurrently and 5 years later for academic difficulties, alcohol misuse and drug use, and delinquent behavior [5]. Thus, reducing the likelihood that an adolescent smoker will transition from experimentation to regular use is an important goal for prevention programs. Although the literature on smoking initiation and experimentation may provide some leads as to potentially important determinants of regular smoking, the risk factors for initial levels of smoking often differ from those for more advanced smoking [6 –9]. For example, family smoking has been identified as © Society for Adolescent Medicine, 2003 Published by Elsevier, 360 Park Avenue South, New York, NY 10010
April 2003
an important risk factor for smoking initiation in adolescence [10] and more strongly related to initiation than to the transition to regular smoking [11], perhaps because it allows the adolescent easy access to cigarettes and opportunities for experimentation. Once adolescents have experimented with cigarettes, peers may replace family members as the dominant social influence on smoking behavior. Consistent with this idea, several studies have found that “friend smoking” is a stronger predictor of the transition to regular use than “family smoking” [9,12], particularly when the majority of one’s friends smoke [13]. Cognitive models of health behavior such as the Theory of Reasoned Action give a central role to attitudes and beliefs in determining behaviors such as smoking [14]. Given that attitudes and beliefs tend to have a greater impact on behavior when they are based on direct experience [15], they may be particularly relevant to understanding the transition from experimentation to regular smoking. Chassin et al [11] found that the transition to regular smoking was best predicted by attitudes and beliefs about smoking, and behavioral intentions to smoke, among those who were already experimenting with smoking. There is also some evidence that parental and peer smoking influence smoking escalation indirectly through the adolescents’ beliefs about the negative health consequences of smoking and feelings of self-efficacy in refusing cigarette offers [8]. Other cross-sectional analyses have shown that 7thgrade regular smokers are more likely than experimenters to view cigarettes as affordable and smoking as having positive social consequences [9]. Some evidence suggests that deviancy and rebelliousness play a role in the transition to regular smoking. Chassin et al [11] found that a cluster of 17 variables chosen from Jessor and Jessor’s Problem Behavior Theory [16], such as having a devianceprone personality and low expectations for academic success, predicted the transition from experimental to regular smoking over a 1-year period. Other longitudinal research has indicated that having poor grades, experimenting with alcohol, and being inclined toward risk-taking in 7th grade are associated with heavy smoking 5 years later [13]. Rebelliousness has also been found to differentiate 7th-grade experimenters and regular smokers in cross-sectional analyses [9]. Important developmental changes occurring during adolescence and young adulthood may result in age-related differences in the predictors of smoking behavior [6]. Friends become increasingly important
PREDICTORS OF TRANSITION TO REGULAR SMOKING
315
during adolescence and the time spent with family decreases [17]. This suggests that peers may exert a stronger influence on smoking behavior as adolescents get older, although two studies of smoking initiation have yielded conflicting results. Whereas Krosnick and Judd [18] found that peer smoking was a stronger predictor of smoking initiation among adolescents than preadolescents, Chassin et al [19] did not find that peer influences were stronger risk factors with increasing age. Attitudes and beliefs also change during adolescence and young adulthood in ways that may be relevant to smoking transitions [20]. However, it is unclear whether such smoking attitudes become stronger or weaker predictors of the transition to regular smoking as adolescents mature. We know of only one study investigating the antecedents of the transition to regular smoking over multiple time periods in adolescence. Wang et al [21] found that the progression to regular smoking among both 12- to 15-year-olds and 16- to 18-yearolds was associated with being white, positive beliefs about smoking, depression, and not liking school. Additional predictors for those aged 12–15 years included poor academic performance; irregular school attendance; and having parents, friends, and teachers who smoked. Although that study provides a useful initial step in terms of understanding stability and change in the risk factors for regular smoking over time, its limitations include the confounding of smoking initiation with escalation and the possibility that age-related differences may have been owing to cohort effects, rather than developmental changes. The present study uses longitudinal data to identify predictors of the transition to regular smoking among initial experimenters during middle adolescence (grades 8 to 10), late adolescence (grades 10 to 12), and young adulthood (grade 12 to age 23 years). Based on prevalence rates of regular smoking by race/ethnicity and gender [1], Whites were expected to be more likely than other racial/ethnic groups to make the transition to regular smoking during each period, although significant gender differences were not necessarily expected. Prosmoking attitudes were also expected to be risk factors for the transition to regular smoking. Although Chassin et al [20] found age-related differences in smoking attitudes, we did not necessarily expect that their effect on the transition to regular smoking would differ between adolescence and young adulthood. In terms of exposure to a prosmoking environment, we anticipated that family influences (household smoking, parental approval of smoking) would be weaker predictors of
316
TUCKER ET AL
the transition to regular smoking than peer influences (perceived prevalence of peer smoking, friend smoking, and approval of smoking) at all three assessment periods. Based on Jessor and Jessor’s Problem Behavior Theory, we further hypothesized that adolescents who engaged in other problem behaviors (drinking, delinquency) had a weaker academic orientation, and poorer relations with parents would be more likely to transition to regular smoking, although we did not expect these associations to persist into young adulthood. Finally, a number of studies have linked emotional distress with smoking [22–24]. Based on this work, we expected that initial experimenters with poorer mental health would be more likely to transition to regular smoking during both adolescence and young adulthood.
Methods Participants Participants in the original full sample were 6527 males and females in the RAND Adolescent/Young Adult Panel Study, conducted to evaluate the effectiveness of the Project ALERT drug use prevention program for middle school children [25]. Students were recruited from 30 California and Oregon schools in grade 7 (1985) and assessed again in grades 8, 9, 10, and 12 and at age 23 years (1995). The schools were chosen to represent a wide range of community types (urban, suburban, and rural), socioeconomic status (18 schools drew from neighborhoods with household incomes below the median for their state), and racial and ethnic composition (nine schools had minority populations of 50% or more). This study focuses on three transition periods: grades 8 to 10 (1986 to 1988), grades 10 to 12 (1988 to 1990), and grade 12 to age 23 (1990 to 1995). Extensive tracking methods allowed us to retain 85% of the original baseline in 1988 (n ⫽ 5523), 67% of the sample in 1990 (n ⫽ 4390), and 53% of the sample in 1995 (n ⫽ 3466). We restricted our sample to adolescents who reported “ever smoking but less than weekly” at the beginning of each transition period. Additional individuals were dropped from analyses if they were missing smoking or ethnicity information. These exclusions resulted in final sample sizes for each transition period of n ⫽ 2496, n ⫽ 2149, and n ⫽ 1534, respectively. Of the final sample for the grades 10 to 12 transition period, which was weighted to reduce the bias associated with attrition, 37% were self-classified as a minority (13% African-
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
American; 12% Hispanic; 8% Asian; 4% Other), 47% were female, and 16% dropped out of school by the grade 12 assessment (the demographic profiles of the other two subsamples are nearly identical). Participants who could not be located or had withdrawn from the study differed from those who were retained on demographic characteristics, substance use, and grades [26]. Previous work has shown that the use of sample weights is the best method of adjusting for attrition in this school-based sample [27], and weights developed to deal with attrition by 1990 have been found to remove 90% or more of the bias exhibited in the unweighted sample [28]. For the present analyses, the sample was weighted to reduce bias associated with attrition occurring at each assessment. The weighting was conducted by using logistic regression to create predicted probabilities of responding to the 1988, 1990, and 1995 surveys that were derived from 7th grade information about each 7th grade respondent (including race, gender, family structure, delinquency, substance use, and grades). Measures A comprehensive review of previous empirical and theoretical work guided the development of the survey instruments for this project. Almost all of the items and measures have been successfully used in national surveys of adolescent drug use, although some were modified to match the reading level and experience of the sample. Smoking status. Participants were eligible for the analyses if they had “ever smoked but less than weekly” at the beginning of each transition period. Weekly smoking was defined as “having smoked at least 3 to 5 days in the past month.” Participants were considered to have made the transition to regular smoking at each follow-up if they reported weekly or more frequent smoking. Most of the experimenters were not current smokers at the beginning of each transition period, with the weighted percent who had not smoked in the past 30 days steadily increasing from grade 8 (83%) to grade 10 (87%) to grade 12 (92%). Demographics. Participants’ gender, race/ethnicity (white vs. African-American, Hispanic, Asian, and Other), age, and educational attainment of their father and/or mother at the beginning of each transition period were included as control variables in the main analyses.
April 2003
Smoking beliefs. Smoking resistance self-efficacy (RSE) at grade 8 was assessed by asking participants what they would do if they were offered a cigarette on a date and did not want it (“1” ⫽ “I would tell my date no” to “3” ⫽ “I would smoke the cigarette”) and how strongly they agreed with the statement “If all my friends were smoking cigarettes at a party, I’d feel pretty out of it unless I smoked, too” (“1” ⫽ “strongly disagree” to “4” ⫽ “strongly agree”). Two additional items at grade 10 assessed whether they could resist smoking if offered a cigarette when their best friend was smoking and they were bored. Four items at grade 12 indicated whether they could resist smoking when their best friend was smoking, they were bored, they were at a party at which all their friends were smoking, and their date was smoking (“1” ⫽ “very sure I could” to 5 ⫽ “not at all sure I could”) (␣ ⫽ .73 to .96). Participants indicated whether they intended to smoke in the next 6 months (“1” ⫽ “definitely no” to “4” ⫽ “definitely yes”). Single items were used to assess the harmfulness of smoking cigarettes occasionally (“1” ⫽ “a lot” to “4” ⫽ “not at all”) and whether they might get addicted if they smoked cigarettes every weekend (“1” ⫽ “strongly agree” to “4” ⫽ “strongly disagree”). Beliefs about the “nonhealth consequences of smoking” were assessed by four items at grades 8 and 10 and seven items at grade 12 (e.g., it relaxes you; makes you feel more at ease with others). Higher scores indicated more positive beliefs (grade 8: ␣ ⫽ .66; grade 10: ␣ ⫽ .65; grade 12: ␣ ⫽ .77). Smoking environment. We assessed five aspects of the participants’ smoking-related social environment: friend smoking; perceived prevalence of peer smoking; friend and parent approval of smoking; and any household smoking. Friend smoking was assessed using two items: how often they are with people their age who are smoking (“1” ⫽ “never” to “4” ⫽ “often”) and whether (grade 8) or how often (grades 10 and 12; “1” ⫽ “never” to “4” ⫽ “often”) their best friend smokes (␣ ⫽ .69 at grade 8; ␣ ⬎ .70 at grades 10 and 12). Perceived prevalence of peer smoking at grades 10 and 12 was assessed by asking participants to estimate the percentage of their classmates who had smoked in the last month. Parental and friend approval of smoking were rated in terms of how each would feel if they found out that the participant smoked (parent rating: “1” ⫽ “very upset” to “4” ⫽ “not at all upset or would not care”; friend rating: “1” ⫽ “they would disapprove and stop being my friends” to “4” ⫽ “they would approve or would not care”). Any household smoking
PREDICTORS OF TRANSITION TO REGULAR SMOKING
317
at grades 10 and 12 was assessed in terms of whether any of these individuals smoked (“0” ⫽ “no”, “1” ⫽ “yes”): an older sibling, the adult male and female who are most important to them (presumably the parents). Alcohol consumption and deviance. Participants provided information on their history and frequency of alcohol use; this information was recoded on a scale ranging from “0” (“never used”) to “11” (“used 20 or more days in past month”). Engagement in delinquent behavior was assessed by four items at grade 8 and eight items at grade 10 (e.g., school problems, stealing, broken into or damaged property: “0” ⫽ “not at all” to “3” ⫽ “more than three times in past year”; ␣ ⫽ .63 at grade 8 and .78 at grade 10, respectively). At grade 12 participants indicated how often they had engaged in each of 26 problem behaviors (e.g., school problems, stealing, violent behavior, damaging property, selling drugs: “0” ⫽ “not at all” to “5” ⫽ “20 or more times in past year”; ␣ ⫽ .86). Academic orientation. Participants provided information on their grades at grade 10 (“1” ⫽ “mostly A’s” to “5” ⫽ “mostly F’s”) and grade 12 (“1” ⫽ “mostly A’s” to “8” ⫽ “mostly below D”). They also indicated their academic intentions at grades 10 and 12 in terms of how much schooling they anticipated completing (“1” ⫽ “attend graduate or professional school” to “6” ⫽ “some high school.” Because this information was not available at the grade 8 assessment, we used grade 7 information as a reasonable proxy for grade 8 academic orientation (grades: “1” ⫽ “mostly A’s” to “5” ⫽ “mostly F’s”; academic intentions: “1” ⫽ “attend graduate or professional school” to “5” ⫽ “may not finish high school”). Parental bonding. Participants indicated whether they were living with both biological parents (“0” ⫽ “no,” “1” ⫽ “yes”); not having an intact nuclear family would suggest that the child’s bond with one or both biological parents was weakened by diminished contact. At grade 8 parental support was assessed in terms of whether participants indicated that they would be most likely to talk to their parents about a personal problem (“0” ⫽ “yes”; “1” ⫽ “no”). The measure of parental support contained three items at grade 10 (parents listen to my opinions, respect my feelings, don’t trust my judgment) with three items added at grade 12 (parents keep track of how I am doing in school, sense when I am upset, know where I am and what I am doing: “1” ⫽
318
TUCKER ET AL
“strongly agree” to “4” ⫽ “strongly disagree”). Higher scores indicate weaker parental support (␣ ⬎ .70 at grades 10 and 12). Mental health. Perceived mental health status was assessed at grades 10 and 12 by the mental health subscale (MHI-5) of the Medical Outcomes Study Short-Form Health Survey (SF-20) [29]. Items ask how much of the time, during the past month, participants have: “been a very nervous person”; “been a happy person”; “felt calm and peaceful”; “felt downhearted and blue”; “felt so down in the dumps that nothing could cheer them up” (“1” ⫽ “all of the time” to “6” ⫽ “none of the time”). Higher scores on the scale indicated poorer perceived mental health (␣ ⫽ .50 at grade 10; ␣ ⫽ .83 at grade 12).
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
as covariates in each of these predictor models. Trimmed models that contained only variables that were at least marginally significant (p ⬍ .10) predictors of the transition to regular use were then examined. These trimmed models were also run controlling for smoking frequency and the number of years the participant had smoked to determine whether the psychosocial and behavioral variables remained significant predictors of the transition to regular smoking after accounting for these indicators of smoking dependence. We checked for any effect of participation in the drug prevention program in grades 7 and 8 by adding treatment to the relevant models; because it was not significant, we have not included it in the tables of results.
Statistical Methods
Results
Participants at each wave of data collection generally completed the entire survey. However, there was a small amount of missing data for most variables (1% to 6%), which can result in a sizable sample reduction for multivariate analyses that require complete information for all included variables. To circumvent this problem, we used least squares regression imputation on continuous predictors and logistic regression imputation on dichotomous predictors, with the exception of race/ethnicity and smoking status [30]. These imputation procedures generated a predicted value (the criterion) from other variables that were strongly correlated with the criterion value (e.g., demographic characteristics, substance use, academic performance, deviant behaviors) and then substituted the predicted value for the missing value. The clustering of observations within schools and use of sample weights were accounted for by computing Huber standard errors [31,32]. Huberized regression techniques were used to determine the independent associations of the predictor variables on subsequent smoking status. Logistic regression analyses were initially conducted to identify predictors of the transition to regular smoking. Given the large number of predictor variables in the analyses, seven separate models were tested. The initial model included demographic variables only (gender, ethnicity, parental education, and age). The remaining six models each included one of the conceptually related groups of psychosocial and behavioral predictors (smoking beliefs, smoking environment, alcohol consumption and deviance, academic orientation, parental bonding, and mental health). Demographic variables were entered
Prevalence of Transitioning to Regular Smoking The number (and weighted percent) of initial experimenters who became regular smokers by the end of each transition period was 473 (19%) individuals from grades 8 to 10, 329 (16%) individuals from grades 10 to 12, and 259 (17%) individuals from grade 12 to age 23 years. Predictors of the transition to regular smoking are shown in Table 1 for the period from grades 8 to 10 (middle adolescence), Table 2 for grades 10 to 12 (late adolescence), and Table 3 for grade 12 to age 23 years (young adulthood). Trimmed final models are shown in Table 4. Demographics. Ethnicity was a significant predictor of the transition to regular smoking during adolescence but not young adulthood. Results from the trimmed models indicated that African-Americans, Hispanics, and Asians were all at significantly lower risk, compared with Whites, of transitioning to regular smoking during middle adolescence. The lower risk associated with being Hispanic and Asian dissipated during late adolescence, and the lower risk associated with being African-American dissipated during young adulthood. Younger smokers were at higher risk of transitioning to regular smoking but only during young adulthood. Gender and parental education were not significant predictors of the transition to regular smoking. Smoking attitudes. At all ages, a stronger intention to smoke in the future was a risk factor for transitioning to regular smoking. Those with lower RSE were also more likely to escalate their smoking in adolescence, although not in young adulthood. Pro-
April 2003
PREDICTORS OF TRANSITION TO REGULAR SMOKING
319
Table 1. Predictors of the Transition to Regular Smoking From Grade 8 to Grade 10 (n ⫽ 2496) Grade 8 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
1.16 0.36** 0.72 0.55** 1.17 1.14 0.98
[0.90, [0.19, [0.48, [0.40, [0.65, [0.93, [0.86,
1.49] 0.68] 1.07] 0.76] 2.09] 1.41] 1.12]
1.86
(1.05)
0.00 –3.00
1.22* 1.42*** 1.19* 1.19* 1.30*
[1.04, [1.23, [1.04, [1.05, [1.06,
1.44] 1.63] 1.37] 1.36] 1.60]
0.02 1.68 3.28 1.76 1.51
(0.84) (0.82) (0.83) (0.96) (0.61)
⫺0.81–2.66 0.78 – 4.00 1.00 – 4.24 1.00 – 4.00 0.93– 4.00
1.74*** N/A 0.93 1.20* N/A
[1.44, 2.10]
⫺0.02
(0.78)
⫺1.18 –1.50
[0.79, 1.09] [1.05, 1.38]
2.37 1.39
(0.72) (0.68)
1.00 – 4.00 0.88 – 4.00
1.08*** 1.22*
[1.04, 1.12] [1.04, 1.43]
4.27 0.84
(3.09) (0.68)
0.00 –11.00 0.00 –3.00
1.60*** 1.13*
[1.42, 1.79] [1.03, 1.24]
2.33 2.14
(0.83) (0.98)
1.00 –5.00 1.00 –5.00
0.89 1.61*** N/A
[0.70, 1.14] [1.29, 2.00]
0.54 0.55
(0.50) (0.50)
0.00 –1.00 0.00 –1.00
* p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
a
smoking beliefs were risk factors at all three time periods, although results differed, depending on the type of belief. Contrary to expectations, smokers holding a stronger belief in the harmful effects of smoking were more likely to transition to regular smoking during both adolescence and young adulthood. However, these associations were owing to a suppression effect; believing that smoking is harmful was weakly related to smoking escalation when controlling only for demographic variables (all p ⬎ .20; results not shown). Smokers with a weaker belief in the addictiveness of smoking and those with a stronger belief in the positive nonhealth consequences of smoking were more likely to transition to regular smoking. However, controlling for other nondemographic and nonbelief variables in the trimmed models eliminated the association with addictiveness beliefs in young adulthood and the association with beliefs in the positive nonhealth consequences of smoking in adolescence.
Smoking environment. Friend smoking was a risk factor for transitioning to regular smoking during both adolescence and young adulthood when controlling for demographics and other aspects of the smoking environment. However, friend smoking remained a significant predictor only in middle adolescence when controlling for other variables in the trimmed models. Similarly, parental approval of smoking was a risk factor during both adolescence (significant) and young adulthood (marginally significant) when controlling for demographic and smoking environment variables. Controlling for other types of variables in the trimmed models eliminated the association at all three transition periods. Finally, the perceived prevalence of peer smoking, friend approval of smoking, and household smoking were not significant predictors of the transition to regular smoking (note that information on the perceived prevalence of peer smoking and household smoking were not available at grade 8).
TUCKER ET AL
320
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
Table 2. Predictors of the Transition to Regular Smoking From Grade 10 to Grade 12 (n ⫽ 2149) Grade 10 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
0.96 0.43** 0.80 0.79 1.06 0.94 0.96
[0.68, [0.24, [0.50, [0.42, [0.42, [0.67, [0.86,
1.35] 0.76] 1.28] 1.51] 2.68] 1.32] 1.06]
3.16
(1.49)
1.00 – 6.00
1.24* 1.61*** 1.19* 1.23* 1.32*
[1.03, [1.32, [1.02, [1.05, [1.03,
1.48] 1.96] 1.39] 1.45] 1.71]
⫺0.10 1.53 3.31 1.70 1.70
(0.66) (0.74) (0.80) (0.92) (0.57)
⫺0.65–2.49 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00 1.00 –3.57
1.37** 1.00 1.12 1.24* 1.29
[1.11, [1.00, [0.90, [1.03, [0.88,
1.69] 1.01] 1.41] 1.49] 1.90]
1.98 53.63 2.39 1.54 0.63
(0.81) (24.02) (0.67) (0.79) (0.48)
1.00 – 4.00 0.00 –100.00 1.00 – 4.00 1.00 – 4.00 0.00 –1.00
1.12*** 1.63**
[1.07, 1.18] [1.26, 2.10]
5.32 0.43
(3.11) (0.44)
0.00 –11.00 0.00 –3.00
1.17† 1.03
[0.99, 1.38] [0.90, 1.17]
2.37 2.57
(0.84) (1.35)
1.00 –5.00 1.00 – 6.00
0.66* 1.40*** 1.28***
[0.47, 0.92] [1.22, 1.59] [1.16, 1.42]
0.53 2.08 2.66
(0.50) (0.81) (0.97)
0.00 –1.00 1.00 – 4.00 1.00 – 6.00
p ⬍ .10; * p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
† a
Other problem behaviors. Greater alcohol consumption and deviant behavior were risk factors for the transition to regular smoking during middle and late adolescence but not during young adulthood. Controlling for other variables in the trimmed models, only alcohol consumption in late adolescence remained a significant risk factor for transitioning to regular smoking. Academic orientation. Poor grades and low academic intentions were significant risk factors for regular smoking during middle adolescence and young adulthood, and poor grades was a marginally significant risk factor in late adolescence. However, only in middle adolescence did these associations remain significant when controlling for other variables in the trimmed models. Parental bonding. Not having an intact nuclear family was a significant risk factor for transitioning to regular smoking during late adolescence only.
Poorer parental support was a risk factor during both adolescence and young adulthood, although only the associations in adolescence remained significant after controlling for other variables in the trimmed models. Mental health. Mental health information was unavailable at grade 8. Poorer mental health was a risk factor for regular smoking during late adolescence and young adulthood, although these associations did not remain significant in the final trimmed models. Controlling for Frequency and Length of Smoking Smoking frequency and the number of years smoking were added to the trimmed models. The significant associations shown in Table 4 remained at least marginally significant (p ⬍ .10) after controlling for these indicators of smoking dependence with the
April 2003
PREDICTORS OF TRANSITION TO REGULAR SMOKING
321
Table 3. Predictors of the Transition to Regular Smoking From Grade 12 to Age 23 (n ⫽ 1534) Grade 12 Variables Model 1: Demographics only Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Model 2: Smoking attitudes (⫹ demographics) Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Model 3: Smoking environment (⫹ demographics) Friend smoking Perceived prevalence of peer smoking Friend approval of smoking Parent approval of smoking Any household smoking Model 4: Other problem behaviors (⫹ demographics) Alcohol consumption Deviance Model 5: Academic orientation (⫹ demographics) Poor grades Low academic intentions Model 6: Parental bonding (⫹ demographics) Intact nuclear family Poor parental support Model 7: Poor mental health (⫹ demographics)
OR
[95% CI]
Mean
(SD)
Rangea
1.04 0.94 0.83 0.78 1.86 0.76† 1.02
[0.75, [0.63, [0.41, [0.58, [0.80, [0.57, [0.93,
1.44] 1.41] 1.67] 1.06] 4.33] 1.02] 1.11]
4.13
(2.24)
0.82–9.00
1.11 1.51*** 1.22* 1.15* 1.90***
[0.95, [1.27, [1.02, [1.02, [1.38,
1.30] 1.81] 1.46] 1.30] 2.60]
1.47 1.35 3.13 1.71 1.64
(1.00) (0.65) (0.83) (0.95) (0.50)
0.61–5.00 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00 1.00 – 4.00
1.46*** 1.00 0.92 1.19† 1.30
[1.20, [0.99, [0.75, [0.97, [0.85,
1.76] 1.00] 1.13] 1.44] 1.98]
2.08 51.88 2.41 1.72 0.58
(0.88) (24.24) (0.64) (0.89) (0.49)
1.00 – 4.00 0.00 –100.00 1.00 – 4.00 1.00 – 4.00 0.00 –1.00
1.05 1.30
[0.97, 1.14] [0.69, 2.44]
5.58 0.31
(3.16) (0.33)
0.00 –11.00 ⫺0.08 –2.15
1.14* 1.19*
[1.01, 1.28] [1.03, 1.36]
3.42 2.52
(1.54) (1.31)
1.00 – 8.00 1.00 – 6.00
1.16 1.21* 1.20*
[0.86, 1.57] [1.01, 1.45] [1.02, 1.40]
0.54 1.96 2.41
(0.50) (0.65) (0.94)
0.00 –1.00 1.00 – 4.00 1.00 – 6.00
p ⬍ .10; * p ⬍ .05; ** p ⬍ .01; *** p ⬍ .001. Range reflects imputed values. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy.
† a
following exceptions: intentions to smoke in both middle adolescence [odds ratio (OR) ⫽ 1.04, not significant] and late adolescence (OR ⫽ 1.02, not significant), as well as friend smoking in middle adolescence (OR ⫽ 1.17, not significant).
Discussion Whites were more likely than African-Americans, Hispanics, and Asians to transition from experimentation to regular smoking during middle adolescence. The difference between Whites and AfricanAmericans persisted into late adolescence, although there were no racial/ethnic group differences by early adulthood. This pattern of results is consistent with national surveys [2], as well as previous research focusing specifically on racial differences in the transition to regular smoking [21]. The lower rate of smoking escalation among African-Americans remained after controlling for demographic factors,
smoking attitudes and environment, engagement in other problem behaviors, and level of parental support. At least one other study has similarly found that such factors cannot explain this racial difference in smoking behavior [33]. Previous work has suggested that black–white differences in smoking may be owing to the underreporting of smoking by African-Americans [34,35]. In the present study, selfreported smoking in grade 8 was found to be highly accurate when validated against a biochemical marker of nicotine uptake, serum cotinine [36], and there is little reason to believe that reports of smoking became less valid over time. Thus, it is more likely that this difference is owing to a complex set of psychosocial and cultural influences that we could not fully assess in this study, rather than to the validity of self-reported smoking. It is interesting that the only demographic characteristic associated with the transition to regular smoking in young adulthood is age, with younger participants being
322
TUCKER ET AL
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
Table 4. Trimmed and Fully Adjusted Models: Predictors of the Transition to Regular Smoking Grade 8 –10
Grade 10 –12
Predictor Variables
OR
Gender African-American Hispanic Asian Other Older age relative to cohort Parental education Low smoking RSE Intentions to smoke Smoking is harmful Smoking is not addictive Smoking has positive effects Friend smoking Parent approval of smoking Alcohol consumption Deviance Poor grades Low academic intentions Intact nuclear family Poor parental support Poor mental health
1.17 0.40** 0.60* 0.62** 1.25 1.02 1.06 1.20* 1.25* 1.22** 1.14* 1.18 1.31* 1.09 1.03 0.94 1.50*** 1.09*
[0.88, [0.21, [0.38, [0.44, [0.72, [0.83, [0.93, [1.03, [1.05, [1.06, [1.00, [0.96, [1.05, [0.94, [0.99, [0.80, [1.34, [1.00,
[95% CI]
1.32* N/A
[1.07, 1.63]
1.54] 0.77] 0.93] 0.87] 2.19] 1.25] 1.20] 1.39] 1.48] 1.41] 1.30] 1.44] 1.65] 1.25] 1.07] 1.10] 1.69] 1.19]
OR
Grade 12 to Age 23
[95% CI]
0.97 0.46* 0.75 0.92 1.04 0.98 0.97 1.31** 1.42** 1.16* 1.18* 1.11 1.14 1.10 1.08** 1.17 0.97
[0.69, [0.25, [0.46, [0.46, [0.38, [0.69, [0.87, [1.08, [1.14, [1.01, [1.02, [0.89, [0.96, [0.94, [1.02, [0.89, [0.83,
1.35] 0.86] 1.20] 1.85] 2.86] 1.39] 1.08] 1.59] 1.76] 1.34] 1.38] 1.39] 1.37] 1.30] 1.14] 1.55] 1.12]
0.65* 1.26** 1.07
[0.47, 0.91] [1.08, 1.47] [0.95, 1.21]
OR
[95% CI]
1.08 1.11 0.72 0.85 1.56 0.68* 1.06
[0.77, [0.71, [0.36, [0.54, [0.62, [0.49, [0.96,
1.51] 1.72] 1.42] 1.34] 3.89] 0.93] 1.18]
1.51*** 1.20* 1.11 1.72** 1.21 1.04
[1.25, [1.00, [0.97, [1.24, [0.99, [0.85,
1.82] 1.44] 1.28] 2.38] 1.48] 1.26]
1.10 1.15
[0.98, 1.24] [0.99, 1.34]
0.92 1.07
[0.75, 1.14] [0.90, 1.28]
Note. OR ⫽ odds ratio; CI ⫽ confidence interval; RSE ⫽ resistance self-efficacy. * p ⬍ .05; ** p ⬍.01; *** p ⬍ .001. With the exception of demographic variables, only those variables associated with the transition to regular smoking at p ⬍ .10 in Tables 1–3 are included in these trimmed models.
more likely to escalate their smoking from grade 12 to age 23 years. It is unclear why experimenters who were young for their cohort escalated their smoking. One possibility is that they wanted to look more mature to their peers; however, if this were the case, one would expect to find this effect in adolescence as well. We expected that participants in our study had already formed strong smoking-related attitudes through their direct experience with the behavior and that these attitudes would be important antecedents of the transition to regular smoking. Indeed, prosmoking attitudes were the most consistent predictors of the transition to regular smoking over time in this study. In the analyses controlling for demographic variables and other smoking-related attitudes, experimenters who transitioned to regular smoking tended to have more difficulty resisting social pressures to smoke, stronger intentions to smoke in the future, weaker beliefs in the addictiveness of smoking, and stronger beliefs in the positive nonhealth consequences of smoking. The only exception to this pattern was found for RSE, which did not predict regular smoking in young adulthood. Although it may be argued that this lack of association is due to the longer (5-year) follow-up period that was used for the young adulthood analyses, the fact
that all other smoking-related attitudes remained significant risk factors in adulthood suggests an alternate explanation: Difficulty resisting social pressures to smoke plays a lesser role in the escalation of smoking over time because the peer pressures felt by adolescents to engage in substance use wane as they enter young adulthood. Attesting to the importance of smoking-related attitudes in the transition to regular smoking, these attitude variables remained significantly associated with smoking escalation in the trimmed models with only a few exceptions (positive nonhealth consequences of smoking during adolescence and the addictiveness of smoking in young adulthood). Supporting our hypothesis, exposure to friend smoking was associated with the transition to regular smoking over time, whereas no such associations were found for household smoking. However, perceived prevalence of peer smoking and approval of one’s smoking by friends were not associated with the transition to regular smoking. Given that both of these factors have been associated with smoking initiation [10], it may be the case that once an individual has begun experimenting with cigarettes, perceiving that others are smoking and that one’s friends approve of this behavior have little effect on whether the individual maintains or escalates his or
April 2003
her smoking over time. It is particularly noteworthy that friend smoking at grade 12 was associated with the transition to regular smoking by age 23 years in light of the significant changes in social networks that often occur during the transition out of high school. Perhaps experimenters who escalate their smoking during this period are choosing friends in part on the basis of their smoking behavior; previous research has provided some evidence for such a selection process [12,37]. In interpreting these findings on friend smoking, it is important to keep in mind that the associations in late adolescence and early adulthood became weaker and nonsignificant after controlling for other variables in the multivariate models. Although household smoking and parental approval of smoking were not significantly associated with the transition to regular smoking, it appears that family influences still play a role in smoking escalation. Parental smoking may be most relevant to the initiation process because it provides easy access to cigarettes and opportunities for experimentation. Likewise, parental disapproval of smoking may be a deterrent for initiating smoking, but, among those who have already started, other factors are likely more relevant to the escalation of use. However, experimenters who reported weaker family bonds were more likely to transition into regular smoking over the follow-up period. Thus, parents may play a role in smoking escalation to the extent that their children smoke in an attempt to mitigate the stress of a disrupted or discordant family environment [38]. Previous research has shown that adolescent depression or emotional distress is prospectively associated with adult smoking [23,39]. Results from this study further indicate that poorer mental health among young smokers is a risk factor for smoking escalation, although these associations did not remain significant in multivariate models. Alcohol consumption, engagement in deviant behavior, and weak academic orientation predicted the transition to regular smoking during adolescence but were not relevant risk factors during young adulthood. This finding is consistent with Problem Behavior Theory [16], which postulates that adolescent substance use often coexists with other problem behaviors because they have similar etiologies and serve common purposes. It also suggests that engagement in these other problem behaviors may play a role in the escalation of smoking during adolescence, suggesting that smoking prevention programs need to address engagement in other types of substance use and problem behaviors as important risk
PREDICTORS OF TRANSITION TO REGULAR SMOKING
323
factors. However, it is important to keep in mind that only academic orientation in middle adolescence and alcohol consumption in late adolescence remained significant predictors of the transition to regular smoking in multivariate models, indicating that the potential influence of engagement in problem behaviors on smoking escalation is overshadowed by other factors. Further, strong conclusions cannot be drawn about declines in the relevance of these factors over time because of the shorter follow-up period for the adolescent vs. young adult analyses.
Limitations Several study limitations should be noted. Results are based on a single sample from Oregon and California and may thus have limited generalizability. However, the schools from which participants were recruited were chosen to represent a broad spectrum of communities, socioeconomic status, and racial composition. Sample attrition is another limitation and necessitated the use of sample weights to effectively reduce attrition-related bias. Data are based on self-report and subject to the usual validity concerns; however, current tobacco use in this sample has been shown to be highly accurate when externally validated [36]. Finally, the longer length of follow-up for the young adult vs. adolescent analyses complicates the interpretation of our results in terms of possible age-related declines in the importance of certain risk factors. However, those variables that emerged as significant risk factors in the 5-year period from grade 12 to age 23 years are likely quite important to the smoking escalation process. Results from this study point to a number of smoking-related attitudes, social influences, and behaviors that prevention efforts may target to curb the escalation of smoking among experimenters. Altering prosmoking beliefs may be particularly important for both adolescent and young adult experimenters. This research was supported by University of California TobaccoRelated Disease Research Program Grant 10RT-0062.
References 1. Centers for Disease Control and Prevention. Youth risk behavior surveillance—United States, 1999. CDC Surveill Summ MMWR 2000;49:1–96. 2. Johnston LD, O’Malley PM, Bachman JG. Monitoring the Future national survey results on drug use, 1975–1999. Vol. I. Secondary school students. NIH Pub. No. 00-4802. Bethesda, MD: National Institute on Drug Abuse, 2000.
324
TUCKER ET AL
3. Fergusson DM, Horwood LJ. Transitions to cigarette smoking during adolescence. Addict Behav 1995;20:627–42. 4. Breslau N, Peterson EL. Smoking cessation in young adults: Age at initiation of cigarette smoking and other suspected influences. Am J Public Health 1996;86:214 –20. 5. Ellickson PL, Tucker JS, Klein DJ. High-risk behaviors associated with early smoking: Results from a 5-year follow-up. J Adolesc Health 2001;28:465–73. 6. Ary DV, Biglan A. Longitudinal changes in adolescent cigarette smoking behavior: Onset and cessation. J Behav Med 1988;11:361–81. 7. Botvin GJ, Baker E, Botvin EM, et al. Factors promoting cigarette smoking among black youth: A causal modeling approach. Addict Behav 1993;18:397–405. 8. Flay BR, Hu FB, Siddiqui O, et al. Differential influence of parental smoking and friends’ smoking on adolescent initiation and escalation of smoking. J Health Soc Behav 1994;35: 248 –65. 9. Robinson LA, Klesges RC, Zbikowski SM, et al. Predictors of risk for different stages of adolescent smoking in a biracial sample. J Consult Clin Psychol 1997;65:653–62. 10. Conrad KM, Flay BR, Hill D. Why children start smoking cigarettes: Predictors of onset. Br J Addict 1992;87:1711–24. 11. Chassin L, Presson CC, Sherman SJ, et al. Predicting the onset of cigarette smoking in adolescents: A longitudinal study. J Appl Soc Psychol 1984;14:224 –43. 12. West P, Sweeting H, Ecob R. Family and friends’ influences on the uptake of regular smoking from mid-adolescence to early adulthood. Addiction 1999;94:1397–1412. 13. Griffin KW, Botvin GJ, Doyle MM, et al. A six-year follow-up study of determinants of heavy cigarette smoking among high-school seniors. J Behav Med 1999;22:271–84. 14. Weinstein ND. Testing four competing theories of healthprotective behavior. Health Psychol 1993;12:324 –33. 15. Krause SJ. Attitudes and the prediction of behavior: A metaanalysis of the empirical literature. Pers Soc Psychol Bull 1995;21:58 –75. 16. Jessor R, Jessor SL. Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth. New York: Academic Press, 1977. 17. Larson RW, Richards MH, Moneta G, et al. Changes in adolescents’ daily interactions with their families from ages 10 to 18: Disengagement and transformation. Dev Psychol 1996; 32:744 –54. 18. Krosnick JA, Judd CM. Transitions in social influence at adolescence: Who induces cigarette smoking? Dev Psychol 1982;18:359 –68. 19. Chassin L, Presson CC, Sherman SJ, et al. Changes in peer and parent influence during adolescence. Longitudinal versus cross-sectional perspectives on smoking initiation. Dev Psychol 1986;22:327–34. 20. Chassin L, Presson CC, Rose JS, et al. From adolescence to adulthood: Age-related changes in beliefs about cigarette smoking in a Midwestern community sample. Health Psychol 2001;20:377–86.
JOURNAL OF ADOLESCENT HEALTH Vol. 32, No. 4
21. Wang MQ, Fitzhugh EC, Green BL, et al. Prospective socialpsychological factors of adolescent smoking progression. J Adolesc Health 1999;24:2–9. 22. Escobedo LG, Reddy M, Giovino GA. The relationship between depressive symptoms and cigarette smoking in U.S. adolescents. Addiction 1998;93:433–40. 23. Orlando M, Ellickson PL, Jinnett K. The temporal relationship between emotional distress and cigarette smoking during adolescence and young adulthood. J Consult Clin Psychol 2001;69:959 –70. 24. Patton GC, Hibbert M, Rosier MJ, et al. Is smoking associated with depression and anxiety in teenagers? Am J Public Health 1996;86:225–30. 25. Ellickson PL, Bell RM. Drug prevention in junior high: A multi-site longitudinal test. Science 1990;247:1299 –305. 26. Collins RL, Ellickson PL, Bell RM. Simultaneous polydrug use among teens: Prevalence and predictors. J Subst Abuse 1999; 10:233–53. 27. McGuigan KA, Ellickson PL, Hays RD, et al. Adjusting for attrition in school-based samples: Bias, precision, and cost trade-offs of three methods. Eval Rev 1997;21:554 –67. 28. Ellickson PL, Saner H, McGuigan KA. Profiles of violent youth: Substance use and other concurrent problems. Am J Public Health 1997;87:985–91. 29. Stewart AL, Hays RD, Ware JE, Jr. The MOS Short-Form General Health Survey. Med Care 1988;26:724 –35. 30. Little RJA. Regression with missing X’s: A review. J Am Stat Assoc 1992;87:1227–37. 31. Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the fifth Berkeley Symposium on Mathematical Statistics Probability 1967;1: 221–33. 32. Stata Corp. Stata Statistical Software: Release 6.0. College Station, TX: Stata Corp., 1999. 33. Flint AJ, Yamada EG, Novotny TE. Black-white differences in cigarette smoking uptake: Progression from adolescent experimentation to regular use. Prev Med 1998;27:358 –64. 34. Bauman KE, Ennett SE. Tobacco use by black and white adolescents: The validity of self-reports. Am J Public Health 1994;84:394 –8. 35. Wagenknecht LE, Burke GL, Perkins LL, et al. Misclassification of smoking status in the CARDIA study: A comparison of self-report with serum cotinine levels. Am J Public Health 1992;82:33–6. 36. Reinisch E, Bell RM, Ellickson PL. How accurate are adolescent reports of drug use? Report No. N-3198-CHF. Santa Monica, CA: RAND, 1991. 37. Ennett ST, Baumann KE. The contribution of influence and selection to adolescent peer group homogeneity: The case of adolescent cigarette smoking. J Pers Soc Psychol 1994;6:653– 63. 38. Covey LS, Tam D. Depressive mood, the single-parent home, and adolescent cigarette smoking. Am J Public Health 1990; 80:1330 –3. 39. Kandel DB, Davies M. Adult sequelae of adolescent depressive symptoms. Arch Gen Psychiatry 1986;43:255–62.