To better understand the motivation for adolescent smoking and drinking and to identify the underlying risk and protective factors associated with these behaviors among adolescents.
Cross-sectional, school-based survey of students in grades 5 through 12.
A nationally representative sample of 2574 boys and 2939 girls in grades 7 through 12 from 297 public, private, and parochial schools across the United States who participated in The Commonwealth Fund Survey of the Health of Adolescent Girls and Boys in 1997.
Main Outcome Measures
Sex-specific adjusted relative risks (RRs) and 95% confidence intervals (CIs) comparing self-reported regular smokers and regular drinkers by risk and protective factors with adolescents reporting none of these behaviors.
Adolescent boys and girls were equally likely to be regular smokers (11.2%). The prevalence rate of regular drinking was only slightly higher for boys (22.4%) than it was for girls (19.3%). The rates of both health-risk behaviors were significantly higher for those reporting risk factors, and the strengths of associations varied by sex. Sex differences also emerged in motivation for engaging in these behaviors. When we adjusted for demographic characteristics, exposure to childhood abuse (RR, 4.1; 95% CI, 2.4-7.0) and stressful life events (RR, 2.4; 95% CI, 1.1-5.4) were strongly associated with increased risk for boys' regular smoking. Similar associations were found for regular drinking. For girls, a history of abuse (RR, 1.8; 95% CI, 1.1-2.8), violence within the family (RR, 2.2; 95% CI, 1.6-3.2), depressive symptoms (RR, 1.6; 95% CI, 1.0-2.4), and stressful life events (RR, 3.1; 95% CI, 1.8-5.6) were significantly associated with increased risk for regular smoking. Similar associations were again found for regular drinking. Parental support was protective against both health-risk behaviors for both sexes. Participation in extracurricular activities was associated with lower risk for regular smoking for boys (RR, 0.4; 95% CI, 0.2-0.7) and for girls (RR, 0.3; 95% CI, 0.2-0.5); however, there was no significant association between drinking behavior and participation in activities.
The increased risk for regular smoking and regular drinking among adolescents with a history of abuse, family violence, depressive symptoms, and stressful life events suggests that routine screening for abuse, violence, and other family experiences should be an essential component of adolescent health care visits. Effective prevention programs to reduce smoking and drinking among adolescents should recognize that health-risk behaviors may be associated with other negative life experiences and that the strength of these associations differs by sex.
HEALTH-RISK BEHAVIORS, especially smoking and drinking, remain a major problem among US adolescents. These behaviors are associated with the leading causes of mortality and morbidity, posing immediate risks to health during adolescence and increasing the likelihood of excess preventable morbidity and death in adulthood.1- 3
Alcohol continues to be the most common substance of abuse among US adolescents.4 Nearly 80% of high school students report having used alcohol at some time in their lives.5 Recent statistics show that 62.3% of 12th graders, 48.9% of 10th graders, and 24.8% of 8th graders reported that they had been drunk at least once in 1999.6 Binge drinking (consuming ≥5 drinks in succession on one occasion) is alarmingly common, with 31% of seniors, 26% of sophomores, and 15% of 8th graders having done so within the previous 2 weeks.6 Initial use of alcohol often occurs in early adolescence. Studies that focus on youth reveal that 39% of 6th to 8th graders consumed alcohol at least once in the past year.7,8 Use of alcohol is associated with the major causes of death in adolescents and young adults, including unintentional injury, suicide, and homicide.9 Furthermore, adolescents who drink become addicted to alcohol more rapidly than do adults who drink, especially when drinking begins before age 15.10
Despite the well-known adverse effects of smoking on health,11- 13 rates of current smoking among adolescents remain at unacceptably high levels, with initiation of smoking occurring at progressively younger ages.7,8 More than one third (34.6%) of seniors, 25.6% of sophomores, and 17.5% of 8th graders report smoking 1 or more cigarettes in the last 30 days.14 The onset of tobacco addiction occurs primarily among children, at an average age of 12.11 Most adolescent smokers are addicted to nicotine and report that they want to quit but are unable to do so.15
Significantly, the rates of smoking and drinking have remained at high levels even though major intervention programs to prevent youth from initiating these behaviors have been developed. Recently, a large body of research has focused on identifying factors that heighten or decrease the risk for engaging in health-compromising behaviors.16- 19 Accumulated findings have distinguished lower socioeconomic status,20- 22 psychological distress,23- 25 exposure to violence and abuse,26- 30 and family stresses31,32 as putting adolescents at greater risk of regular smoking or regular alcohol use. More recently, research efforts have also begun to explore the impact of protective influences on the probability of adolescents' engaging in risky behaviors.33- 35
In the present study, we examine the relationships between rates of smoking and drinking and a wide range of life experiences, abuse, violence, negative life events, and depressive symptoms, using a large nationally representative sample of adolescents and spanning an important age range. Our analysis examines the effects of exposure to factors thought to increase the risk of smoking and drinking. We also examine factors believed to attenuate or protect against health-risk behaviors. In addition, our analysis explores whether the associations of smoking and drinking with risk and protective factors differ by sex.
Our analyses are based on the Commonwealth Fund Survey of the Health of Adolescent Girls and Boys,36,37 a nationally representative stratified sample of students in 297 public, private, and parochial schools in the United States, which was conducted by Louis Harris and Associates. The schools were drawn from a database of approximately 80,000 schools maintained as a national survey resource by the National Center for Educational Statistics. Two hundred sixty-five schools were selected randomly in proportion to their students' representation in the population, and an additional 32 urban schools were selected to oversample minority youths. Any school declining to participate was replaced by randomly choosing 1 of 5 schools whose demographics and geography most closely matched the declining school. The school selection procedure and survey method are discussed in more detail in previously published studies.38,39
Within each participating school, students in a randomly selected English classroom from a randomly selected grade were surveyed anonymously during class. Louis Harris and Associates obtained approval for the study from appropriate officials in accordance with the policies governing each school and collected data from students in compliance with participating schools' internal review processes regarding informed consent.40 Surveys required about 45 minutes to complete and were administered by teachers in the classroom. Students were given envelopes in which to seal their completed surveys before returning them to the teacher. The survey instrument was anonymous; at no point was the student asked to provide his or her name. In addition, participants were instructed that they could skip any questions they did not feel comfortable in answering. Some items of a sensitive nature were omitted from the questionnaire used with students in the fifth through eighth grades, and, similarly, versions of the questionnaires differed slightly across sex. Because of the anonymous nature of the survey, the teachers in each sampled classroom were instructed to collect the survey forms in sealed envelopes. A total of 6748 adolescents completed the survey between December 1996 and June 1997. Twenty questionnaires were excluded because of inconsistent answers. While a number of steps were taken to maximize response rates,38 due to the methods of sampling, survey distribution, and collection by the teachers, response rates for eligible sampled youth could not be determined. The final sample was weighted to adjust for the oversample of urban schools and to allow the results to reflect the in-school distribution of US adolescents. Of the 6728 adolescents who completed the survey, 54% were white (non-Hispanic); 14%, black (non-Hispanic); 9%, Hispanic; 4%, Asian; 3%, other; and 16%, unknown ethnicity. A comparison of the resulting final sample demographics and prevalence rates of key behavioral variables revealed that the sample distribution matched expected rates based on national, federally sponsored surveys of adolescent health (L. Greenberg, PhD, unpublished data, 1997).
Given the low incidence of smoking (3%) and drinking (2%) among youth in the 5th and 6th grades, the analysis in this study was limited to the sample of 5513 adolescents in grades 7 through 12.
The questionnaire asked adolescents to describe their use of cigarettes and alcohol. Six response categories existed for smoking: never smoked, tried 1 or 2 cigarettes, smoked sometimes, smoked several cigarettes in the last week, smoked a pack or more in the last week, or used to smoke but quit. For the analysis, we created 3 smoker categories: nonsmoker (never smoked or tried it once), occasional smoker (smoke sometimes), and regular smoker (several cigarettes or more per week). Adolescents who reported that they used to smoke but quit (6% of the sample) were excluded from the multinomial regression analyses. While these adolescents were nonsmokers at the time of the survey, they did not fit the criteria for inclusion in the nonsmoker category. To be sure that their exclusion did not bias our results, we examined a model with a 4-category smoker variable, which included former smokers. The final results were the same as those obtained using the 3-category smoker variable. Therefore, for simplicity of exposition, we decided to report results from multinomial regressions with a 3-category smoker variable.
In the section on drinking, the survey asked adolescents to select among 5 possible descriptions of how often they drink: never, once or twice in a lifetime, drink once in a while, drink at least once a month, or drink at least once a week. To gauge the intensity of drinking, those who said they ever drink were also asked how many drinks they typically consume during an evening when they drink, with possible responses ranging from 1 to 6 or more drinks. A follow-up question asked how often the youth drinks "enough to feel buzzed, tipsy, or drunk." In the analysis, we group adolescents into 3 main categories: nondrinker (never drink or tried it once or twice), occasional drinker (drink once in a while or at most once a month and consume <3 drinks when drinking, or no frequency specified), or regular drinker (drink at least once a month or more frequently and consume at least 3 to 4 drinks when drinking).
In addition to information about demographic factors, data on family situations and experiences (including violence and abuse), mental health, communication with parents, and participation in extracurricular activities were gathered. A history of abuse was assessed with 2 questions: "Have you ever been sexually abused?" and "Have you ever been physically abused?" Adolescents who responded yes to either of those questions were considered as having been abused. The survey also asked, "Has violence in your home, or the threat of violence, ever made you want to leave your home, even just for a short while?" In the analysis, this variable was used as a general measure of concern with violence at home.
Adolescents were also asked whether disruptive or stressful life events had occurred in the last year, including the death of a close friend, parents' divorce or separation, mother or father losing a job, or parent getting in trouble with the law. Based on their answers, we constructed a scale of the number of stressful life events in the last year that ranged from 0 to 4.
A depressive symptom measure was assessed using the modified version of the Children's Depression Inventory.41 The modified version consisted of 14 of the 27-item scale. Each item included 3 choices, scored as 0 (none/mild), 1 (moderate), or 2 (severe). The youth used the options to rate the degree to which each statement described him or her for the last 2 weeks. In the analysis, we use levels of 9 or higher (the sample mean plus 1 SD, 4.9 ± 4.5) to define moderate to high levels of depressive symptoms. To score 9 or higher, a student had to have indicated some level of depressive symptoms on at least 9 of 14 items—a score of 1 on each—or severe levels on at least 4 of 14 measures, each of which was an extreme statement of mental distress.
Responses to several questions were combined to assess protective factors. Adolescents were considered to have good parent support if they identified their parents as one of several possible choices in response to the question, "Who do you usually talk to when feeling stressed, overwhelmed, or depressed?" To assess teen and parent communication, an index was created from responses naming the mother or father as a main source of information about health care issues, elicited by the following questions: "If you wanted to know about health care issues, who would you ask first?" "Who else would you ask about health issues?" "What sources do you rely on for information about health care?" "Have you learned about birth control, contraception, or preventing pregnancy from your parents?" "If you wanted to get condoms, spermicides, or birth-control pills, where would you go?" Other questions disclosed whether parents had initiated discussions on sensitive topics: "Have your parents discussed the following topics with you?" A list followed that included acquired immunodeficiency syndrome, alcohol abuse, drug abuse, having sex, how women become pregnant, how to prevent pregnancy, sexual abuse of children by adults, STDs, use of condoms, violence, guns, and gangs. Adolescents whose index score was 3 or higher and who also indicated that they turned to their parents for support when stressed, overwhelmed, or depressed were hypothesized as having parental-support protective factor. They were compared with adolescents who had infrequent communication (index score, 0 – 1) and who did not name their parents as someone they turn to for support. (Before construction of the index, a factor analysis using the principal factor method was undertaken on the intercorrelation of the support variable and communication variables to see how they clustered. Each of the items loaded significantly on one main principal factor, which had an eigenvalue of 2.27 and accounted for 37.8% variance of the data, suggesting that the decision to combine these items into a single scale has construct validity.42)
For extracurricular activities, the survey asked about participation in school sports teams and group or individual exercise activities as well as after-school clubs. We constructed a variable indicating participation in any of these activities, coded as 1 for participation and 0 for no activity.
All analyses were conducted using Software for Statistical Analysis of Survey Data (Stata).43 We used the multinomial logistic regression procedure in Stata (svymlog), which took into account the 2-stage probability sample and cluster design of the survey. The primary sampling unit was the classroom. Observations were weighted to reflect grade enrollment, region, ethnicity, and sex. Initially, χ2 and t tests were used to examine the relationship of the potential explanatory variables and adolescents' rates of smoking and drinking. Then, sex-specific multinomial logistic regression models were created to identify the relative significance of each factor for each health behavior. The 2 outcome measures were cigarette use and alcohol use, each of which had 3 levels of severity (none, occasional, and regular). In each case, nonusers were the referent group. The multinomial model allowed us to predict the relative risk (RR) of being an occasional or a regular user compared with a nonuser. From these logistic regressions, we calculated RRs and 95% confidence intervals (CIs) associated with each risk and protective factor, adjusting for grade level, ethnicity, parental education, family structure, and other demographic variables. In this study, we were specifically interested in knowing which factors increase the risk for frequent or regular use of cigarettes and alcohol; thus, we report the relative risks for regular smoking and regular drinking compared with the groups reporting none of this type of behavior.
The sample was approximately evenly divided between boys and girls, and nearly 60% of respondents were white (Table 1). The sample was also similarly distributed across grade levels. More than one third (39.9%) of adolescents had at least 1 parent with a college or higher education. About two thirds (65.4%) of adolescents lived in households with 2 parents.
Table 1 also presents weighted percentages of adolescents reporting smoking and drinking for the entire sample and by sex. Overall, 11.2% of adolescents reported regular smoking, 20.9% reported regular drinking, and 19.2% reported using drugs in the preceding month. The prevalence of these behaviors was similar for both sexes. Notable sex differences emerged for risk and protective factors. Overall, 17.7% of adolescents reported being physically or sexually abused. Girls were significantly more likely than were boys to report a history of abuse. Girls were also more likely to say that violence made them want to leave home and were more likely to report moderate to high depressive symptoms. Girls were more likely than were boys to have had parental support, as measured by their discussing health issues with them and considering parents as support persons when they were in need, but both sexes were equally likely to participate in extracurricular activities.
The prevalence of regular smoking and regular drinking differed significantly by grade level (Table 2). Adolescents in grades 11 and 12 were 3 times as likely as those in grades 7 and 8 to smoke regularly and nearly 4 times as likely to drink regularly. Adolescents in grades 11 and 12 were also twice as likely as the younger group (25.3% vs 12.7%; P<.01) to report the use of drugs. The rates of both behaviors were highest among white adolescents, adolescents whose parents had less than a high school education, and those who lived with 1 parent or in a nontraditional living arrangement. Notably, rates of regular smoking and drinking were significantly lower for blacks than they were for white or Hispanic adolescents.
The reported rates of regular smoking and regular drinking were significantly higher for adolescents who reported a history of abuse, family violence, negative life events, or high depressive symptoms (Table 3). Adolescents with a history of physical or sexual abuse were nearly 3 times as likely to report regular smoking and nearly twice as likely to report regular drinking as their nonabused peers. These differences in rates of regular smoking and regular drinking were marked across all risk factors, and the strengths of associations varied by sex. More extreme use of alcohol by boys who reported physical or sexual abuse was also evident (data not shown). Boys with a reported history of abuse were more than twice as likely as girls who reported abuse to consume 6 or more drinks in succession (34.2% vs 14.7%; P<.01) and nearly twice as likely to get drunk every time they drink (37.1% vs 19.9%; P<.01).
Adolescents who had parental support or participated in extracurricular activities had significantly lower prevalences of regular smoking and regular drinking than did adolescents without these protective factors.
Although a common reason for smoking given by boys (50.1%) and girls (56.4%) was that they smoke because they are surrounded by others who smoke, there were significant sex differences on reported rates of other reasons why they smoke (Table 4). Among regular smokers, adolescent girls were much more likely than were adolescent boys to say they smoke to relieve stress and to help them stay slim. Boys were more likely than were girls to say that smoking makes them feel "cool."
Nevertheless, the 3 primary reasons for regular drinking were similar for boys and girls. Most said they drink because drinking is fun, because it helps them relieve stress, and because it helps them forget their problems. As with smoking, some sex differences in reported rates were evident. Girls were more likely than were boys to say they drink to relieve stress and to forget their problems. In contrast, boys were more likely than were girls to say that they drink because all the popular kids drink.
Relationships between risk and protective factors and regular smoking and drinking were examined further, controlling for grade level, ethnicity, parents' education, and family structure (Table 5). For boys, a history of abuse was associated with a 4-fold higher risk for regular smoking (RR, 4.1) and a 2-fold higher risk for drinking (RR, 2.2). Experiencing negative life events was most strongly associated with drinking behavior (RR, 4.8) and with a more than 2-fold higher risk for smoking (RR, 2.4). The associations of other risk factors with either behavior were not significant.
For girls, significant associations were found between each of the risk factors and health-risk behaviors, with adjusted RRs ranging from 1.5 to 3.1 for regular smoking and from 1.4 to 2.0 for regular drinking.
Of the protective factors, parental support remained significant after controlling for demographic characteristics and diminished by one half the risk for regular smoking and drinking among boys and girls. Participation in extracurricular activities also had a protective effect on smoking, with RRs ranging from 0.3 for girls to 0.4 for boys. However, the association of extracurricular activities with drinking behavior was not significant.
Adolescents who report physical or sexual abuse, violence within the family, stressful life events, or moderate to high depressive symptoms are more likely to report regular smoking and regular drinking. The strength of these associations varies across behaviors and by sex. Consistent with previous findings, depression was a significant risk factor for girls' smoking and drinking, but depressive symptoms were not significant for boys when other characteristics were controlled for in the model.44
For girls, experiencing even 1 or 2 stressful events was significantly associated with increased risk of regular smoking and drinking. Experiencing 3 or more events was especially strongly related to smoking. That girls report a major motivation for drinking and smoking was to escape or alleviate stress arising from depressive symptoms has also been previously reported.45,46 For boys, experiencing 3 or more negative events was a significant risk factor for both behaviors, and this was most strongly associated with drinking. Consistent with previous literature,47 these findings suggest that boys may be more resilient and less susceptible to stressors. Previous studies have shown that girls often rate negative events more stressful than do boys. These differences may be due to sex differences in coping styles. As a result of the different socialization of boys and girls, girls may perceive themselves as less resourceful and may develop more passive styles of responding to threats or distressing events as opposed to a more active style of coping. In contrast, boys may react in anger or respond more physically.48,49 This sex difference in coping may also contribute to girls' higher rates of depressive symptoms relative to boys.50,51
Our study supports previous findings that physically or sexually abused adolescents are much more likely to engage in health-risk behaviors than are other adolescents.52,53 Although girls were more likely than were boys to have experienced abuse—and there was a strong association between reports of ever being abused and current smoking and drinking for both sexes—abuse was an especially strong risk factor for boys' smoking and drinking. Consistent with findings from earlier studies,54,55 more extreme use of alcohol by boys who experienced abuse was also evident in this study.
In this analysis, protective factors were conceptualized as distinct from risk factors. Parental support as a protective factor encompassed several dimensions, including emotional support, closeness, and communication. Strong parental support was significantly associated with reduced risks of smoking and drinking for both sexes, and these protective factors had an independent contribution when controlling for all other characteristics. Studies that have measured other dimensions of family environment have found similar results.34 Adolescents who reported a "connectedness" to their parents were the least likely to engage in risky behaviors. These adolescents felt close to their parents, believed their parents and family members cared for them, and were satisfied with their family relationships.35 In addition, other investigators have found that perceived high expectations by parents regarding school achievement and a sense of parental connection to the school also protected young people from a variety of risk behaviors.33
The second protective factor, participation in extracurricular activities such as exercise or after-school sports clubs, was also associated with a decreased risk of smoking for boys and girls, but had no impact on the risk of drinking. Lower smoking rates among those involved in extracurricular activities could indicate that participation provides a protective factor, or it might be that adolescents who joined exercise or after-school sports clubs may have chosen a healthier way of life. The cross-sectional nature of the study does not allow us to infer causality. Notably, the study did not find a similar extracurricular-activity protective impact on alcohol use.
This study had a number of strengths that increase the applicability of the data to the larger population of youth. First, the survey used in the present study included a wide range of risk indexes. Second, the study directly examined the independent and collective effect of risk factors as well as influences of protective factors that mitigate the exposure to risk. In addition, by using the multinomial model, we were able to focus on adolescents who were regular smokers and regular drinkers, the group at highest risk for becoming addicted to nicotine and alcohol as adults.56 Finally, a major strength of the study was the use of a large nationally representative sample of adolescents, thus allowing for the generalization of findings to other youths in the United States. Our findings provide a greater understanding of the contributing factors and adolescent motivation for smoking and drinking and are especially pertinent to clinicians and other professionals providing health care to youth.
Nevertheless, in drawing conclusions from this study, one must consider its limitations. First, the data are cross-sectional; thus, our observations are associations and do not provide information about causal direction. Cross-sectional studies can rule out possible causes when relationships between variables are not found. They do not, however, determine with certainty which came first, smoking or the variable correlated with it. Second, the data apply only to adolescents who attend school. Absentees and dropouts, presumably at higher risk, were not included. Moreover, while a number of steps were taken to maximize participation, the actual response rates for eligible sampled youth could not be determined. The specific grade was selected randomly in each school, without detailed information about the size of any individual class. Although all students were instructed to return the completed questionnaires in sealed envelopes, we have no information on how many students were absent on the day of the survey. Based on the comparisons of survey responses with those of other national studies of adolescent health, we believe, however, that the final sample is nationally representative of the target adolescent, in-school population.
As with all self-reported data, the question of validity is of concern; this is particularly the case when sensitive questions about abuse and substance use are asked. Although the accuracy of self-reported information on physical and sexual abuse is not clearly documented,57 efforts were taken to ensure confidentiality and anonymity of responses. Furthermore, other studies indicate that self-report data of health behaviors are generally reliable.58
Our findings have important implications for professionals providing health care. The strong associations between negative life experiences and depressive symptoms with regular smoking and regular drinking indicate the need for increased awareness among providers of the possible links between health and underlying concerns and behaviors. In addition to standard questions about drinking and smoking, professionals providing health care need to elicit information from adolescents about stressful life events and find productive ways such as outreach and counseling to help adolescents to cope. One way is to increase social support and provide community resources where they can turn for help.
The significant rates of drinking and smoking among adolescents who have reported physical or sexual abuse confirm the importance of screening for abuse during adolescent health care visits, as has been recommended in the American Medical Association's Guidelines for Adolescent Preventive Services.59 A reported history of physical or sexual abuse should alert professionals providing health care to the possible occurrence of other risky behaviors, and, conversely, certain adolescent problem behaviors should alert the clinician to question adolescents more closely about physical or sexual abuse.
Effective prevention efforts should also recognize that the motivation for engaging in risky behavior differs between boys and girls. Adolescent girls may believe that smoking keeps their weight down, as has been shown prospectively in earlier studies.60- 62 These views should prompt clinicians to suggest healthier and more effective methods of weight control for teenage girls as a part of routine smoking intervention programs.
The sex differences found in associations between risk factors and rates of regular smoking and drinking also have implications for clinicians working with adolescents. Programs to target depression that include alternative coping mechanisms and social support may be effective in smoking and alcohol prevention for girls. In contrast, intensive programs targeted at adolescents exposed to multiple life events may be more effective in combating problem drinking for adolescent males.
Parental support and school activities were associated with lower rates of risk behaviors. For adolescents without strong connections to supportive parents, the findings suggest the need for providing them with other access to caring adults for social and emotional support. The importance of mentoring relationships for at-risk youth has been lately rediscovered.63 Adolescents should also be encouraged to participate in sports and after-school activities, and schools need to create opportunities for students to learn healthy behaviors.
Considering these associations, a review of family structure, school, abuse history, and other life experiences remains an important aspect of routine health maintenance for adolescents. Further study is needed to identify causal relationships between programs and activities and long-term health or school outcomes.
Accepted for publication April 20, 2000.
Reprints: Elisabeth Simantov, PhD, Department of Research and Evaluation, The Commonwealth Fund, One E 75th St, New York, NY 10021 (e-mail: firstname.lastname@example.org).
Simantov E, Schoen C, Klein JD. Health-Compromising Behaviors: Why Do Adolescents Smoke or Drink?Identifying Underlying Risk and Protective Factors. Arch Pediatr Adolesc Med. 2000;154(10):1025-1033. doi:10.1001/archpedi.154.10.1025