In its report, Youth Risk Behavior Surveillance—United States 2005, the CDC asserts that “health-risk behaviors, which contribute to the leading causes of morbidity and mortality among youth and adults, often are established during childhood and adolescence, extend into adulthood, are interrelated, and are preventable [1]. Results from the CDC’s national Youth Risk Behavior Survey 2005 indicated that, during the prior 30 days, 23.0% had smoked cigarettes, 43.3% had consumed alcohol, 20.2% had used marijuana, and 2.1% reported past injection of an illegal drug [1]. Risky behavior among adolescents has decreased since 1991, but many still engage in behaviors which contravene achieving the goals of Healthy People 2010 [2]. Those who continue to engage despite policy interventions may be most at risk for negative outcomes.

The CDC calls for “more effective school health programs and other policy and programmatic interventions which are needed to reduce risk and improve health outcomes among youth” [1]. While school health programs may have some impact, it is critical for such programs to target those adolescents most at-risk. This research seeks to increase the effectiveness and target efficiency of health policies and programs whose aim is to reduce health-risk behaviors among adolescents, by exploring the role of resilience in adolescents’ risk for participation in these behaviors.

Efforts have been made to identify processes of both vulnerability and protection to reduce health compromising behaviors in adolescents [3]. Resilience refers to an individual’s capacity to engage in health-promoting behaviors in the face of adversity [4–9] and is an important element in human adaptation. It is a distinguishing quality necessary for adapting to changing demands of daily living. Work in psychology [5,7,8,10–12] introduces the notion of resilience to understand coping and adaptation, which can influence activity preferences, and supports the premise that resilient adolescents are more able to endure stress and avoid addiction [5,7,8]. Two individuals facing the same family and social pressures may respond differently based on unobserved attributes like resilience. Previous literature [5,7,13] has suggested that factors associated with resilience and vulnerability occur on three levels: personal, family, and community and that each contributes to an individual’s overall resilience. Social supports outside the family contribute to community-level resilience, family-level resilience is associated with parental function and relationships, and personal-resilience is associated with the individual’s psychological, cognitive, and emotional processes [5,7,8,13,14]. The indices we construct in this paper capture resilience in these areas. While an individual’s stock of resilience at any given point in time is also a function of these outside influences (community and family), it is the perception of one’s own situation that will ultimately drive behaviors such as smoking, drinking and drug use. There are three models of resilience in the literature. In this paper we build on the compensatory model of resilience that posits that the collective appraisals of risk and compensatory factors predict competence. For example, negative personal self image (risk factor) combined with a positive perception of peer support, respect and admiration (compensatory factor) may have a neutralizing effect on an individual’s propensity to engage in unhealthy activities [15–18]. Whether the factor is risky or compensatory is determined by the individual’s own self perception. We construct our index in line with this theoretical framework. The other two models provide alternative rationale for resilience in adolescents. The protective factor model identifies social and environmental assets and resources as factors that reduce the degree of risk [19]. The challenge model [18] focus on risk factors as facilitators of adaptation. Individuals build on their resilience capacities when confronted with stress that does not exceed their ability to cope.

Indications of not enjoying life, feeling sad, feeling hopelessness etc., imply lower personal competence and resilience. Alternatively, feeling one has good qualities and having a sense of accomplishment and pride are indicators of a healthy sense of personal competence and mastery and, consequently, of resilience [4,6,20–23]. In this research, we use indicators of these emotional attributes to construct measures of resilience in adolescents.

An adolescent functions in some dominant culture with customs, practices, ideologies, values, norms, and beliefs. The social environment, in addition to personal competence, is fundamental to fostering and exercising resilience, which, in adolescents, may be influenced by adverse experiences [13,24,25]. The social context might afford opportunities for feeling socially accepted, loved, and wanted (helping develop an adolescent’s resilience), or, conversely, might foster a sense of vulnerability, hindering development of resilience and resulting in initiation of activities that compromise health [6].

Adolescents engage in activities that satisfy their physiological and psychological needs and provide physical and mental comfort and security. The activities chosen (health-promoting or health-compromising) depend on the contexts in which the individual exists. Regardless of the activities’ benefits or detriments, engagement in these activities is contextually purposeful and meaningful to the individual [26]. Therefore, an adolescent’s overall resilience is not just a function of individual makeup, but also strongly impacted by social constructs. We also use indicators of an adolescent’s contextual/social attributes in our construction of resilience measures.

Factors such as peer and family influences, school, and policy environment have all proven to be predictors of early initiation of risky products like cigarettes [27–29], and policies have been enacted to reduce initiation, but with mixed results. At the margin, most of these policies have had some impact. The economics literature consistently finds that increased price reduces initiation [30]. However, it may be that “sin taxes” and advocacy programs have disproportionately provided incentives and information to deter such behavior among adolescents for whom addiction was an unlikely outcome [31,32]. To effectively reduce initiation and addiction, current policies may not be optimal since their focus is on adolescents in general, rather than those with the highest propensity for adverse outcomes. For example, the D.A.R.E. (Drug Abuse Resistance Education) program, once offered by 75% of American school districts to all adolescents and/or pre-adolescents in certain grades, may be eliminated by some districts because studies have not demonstrated the program’s effectiveness [33–35]. Some districts are considering replacing D.A.R.E. with alternative programs which assess “the levels of risk factors related to problem behaviors such as alcohol, tobacco and other drug use—and to identify the levels of protective factors that help guard against those behaviors” [36].

We suggest that one difficulty in achieving success with existing anti-addiction policy is the lack of research that identifies at-risk groups. Jessor (1991) believes the aim of research regarding risky adolescent behavior should be identifying its psychosocial antecedents and determinants, that is, “What are the risk factors for the risk factors?” [37], and that effective policy in reducing this behavior must address an adolescent’s capacity for coping with life as a factor influencing initiation of risky behaviors [4,9,20]. Our premise is that the protective factors which contribute to resilience help adolescents endure stress and avoid addiction-associated psychosocial impairments [5,7,8]. By developing resilience indices, we identify an important source of heterogeneity among adolescents in their propensities to cope, in order to identify those that may be more or less at risk for participating in health-compromising behavior.

Policy makers need to know who is at risk, the magnitude of that risk, and the factors that drive risky behavior, to develop target efficient policies aimed at improving outcomes among adolescents. Considering personal, psychosocial, and environmental factors when developing policy to reduce risky behavior should enable interventions to reduce participation and improve health more efficiently. Further, if resilience is a key factor in predicting outcomes, effective policies may attempt to improve resilience among adolescents, strengthening coping skills and the contextual factors that contribute to resilience [38]. The three models of resilience in the literature can be used to inform policy. Therefore, the ability to assess resilience in adolescents might help predict who is at risk for addiction, enabling addiction prevention policies to be targeted at those adolescents most likely to benefit. Using factor analysis on data from the nationally representative sample of middle/high school students in the National Longitudinal Study of Adolescent Health [39] (Add Health), we identified three, mutually-exclusive manifestations of resilience. Consistent with assumptions and prior literature, these are related to social, family, and individual resilience [6]. To identify the effect of adolescent resilience on the risk of engaging in health-compromising behaviors (smoking, drinking alcohol, and using illegal drugs) and subsequent addiction, we tested the following hypothesis: Adolescents who are less resilient will be more likely to smoke, drink alcohol, and use illegal drugs and will consume greater quantities of cigarettes, alcohol, and illegal drugs.

The results of our analyses of the influence of resilience on risky behavior are reported in Table 3. The effect of resilience persisted even after controlling for family, peer, and other environmental and personal factors. For each behavior (smoking, drinking alcohol, and using illegal drugs) we report the difference in the probability of participating (columns 1, 9 and 17) from the average probability (for a marginal increase in the resilience index or other variable) and then, for those who do participate, the incremental amount consumed (columns 5, 13 and 21). We modeled consumption thus to understand severity of the behavior among those who chose to participate. (Overall consumption results, which are similar to these, are also available, upon request.)

A one unit increase in the overall-resilience index reduces the probability of smoking by almost 5%, by 6% for drinking, and by 4% for illegal drug use. Those who were self/family-resilient but perhaps are weaker in social resilience were at a higher risk for participation. Individuals who score marginally higher in this index are 6–7% more likely to smoke or drink and 2% more likely to use illegal drugs than the average. Those who were self-resilient were marginally less likely to smoke or drink, but 1% more likely to use illegal drugs.

For consumption (conditional on participation), those who were overall-resilient consumed less alcohol and cigarettes with no difference in illegal drug consumption. Those who were self-resilient consumed more cigarettes, while those who were self/family-resilient smoked fewer cigarettes. Drinkers who were self-resilient consumed above average quantities of alcohol. Among drug users (smallest sample), there was no difference in consumption by resilience.

The confounding factors were also significant in predictable ways for both participation and consumption. The role of parents was especially important with parental education and participation in risky behaviors, as well as the relationship with the child, all being significant factors. After controlling for resilience, older adolescents, males, whites, and Hispanics were more likely to engage in risky behavior. Religious adolescents were less likely to do so.

The literature suggests three components of resilience (personal, family, and community) [8–11]. The data confirm three significant underlying components, with the bulk of what determines resilience captured by the first factor. However, two other factors, with sufficiently high eigenvalues and factor loadings, indicated residual commonalities worth including beyond the first resilience index. It was predicted that adolescents with high scores on these three factors had greater resilience and, in turn, would exhibit better outcomes than those with low scores, after controlling for other risk factors.

The results (see Table 3) indicate that self/family-resilience, in the absence of peer/community resilience, was inadequate to prevent participation in risky behaviors (Risk factors - smoking, drinking, and using drugs). Perhaps those who experience an absence of peer acceptance (compensatory factors) may be unable to resist risky behaviors due to peer pressure to “fit in.” However, strong self/family-resilience can lead to a reduction in the number of cigarettes smoked. These may be the experimenters or intermittent participants who smoke when out socially but with strong enough family support and self resilience to know when to stop. Alternatively, those who were overall-resilient may be better equipped to resist peer pressure to smoke, drink, or use drugs. Interestingly, those who were self-resilient, with little family or social support, were less likely to smoke or drink but those who did consumed larger quantities. In a model where we control for other risk factors like family characteristics, strong self resilience (coping skills) predictably reduces propensities to engage in such behaviors. There does appear to be a split among adolescents who fall into this resilience type since those who do participate are the target group at highest risk for addiction since among those who do participate, they are consuming closer to addictive levels. In addition, within this type are those who are more likely to turn to illegal drug use rather than face problems.

In sum, perhaps smokers who were overall-resilient were primarily experimental, smokers who were self/family-resilient were primarily experimental or intermittent (mainly to fit in socially), while smokers who were self-resilient were primarily regular smokers who may smoke as an escape, rather than for social acceptance. While those who were self-resilient consumed an average amount of drugs, they were more likely to use drugs, and any use of illegal drugs could be seeking an escape. These associations between resilience and consumption are particularly noteworthy because the consumption models contained only participants, who were more homogeneous in resilience and other risk factors, since these factors already significantly explained the propensity to participate.

Programs like the DARE program may have had little impact because they are not target efficient [27–29]. This work has implications for policies aimed at improving outcomes for adolescents at risk of engaging in the health compromising activities. School wide programs that are “one size fits all” may not yield desirable results without attempting to tailor the program to include incentives that will work for adolescents at risk. An alternative approach might be to identify at risk children in schools based on resilience scores and develop programs to increase self efficacy and compensate for lack of environmental assets and resources. Occupational therapists provide interventions in mental health and school-based settings that assist clients in developing compensatory strategies for building resilience and facilitating healthy adaptation. In school-based practice, occupational therapy can provide target specific interventions that meet the specific needs of at risk children and promote resilience [46].

One limitation of this study is that the Add Health data was not designed with the intent to construct indicators of resilience. So while the data are rich with indicators that overlap those used in the literature to construct resilience, the questions were not asked in a sequence consistent with how it might be done in a resilience study, and we had to pool together indicators throughout the survey piecemeal.

These results motivate further analysis of the relationships between adolescent resilience and risky behavior. We can further examine consumption findings, to test for level of addiction, such as whether those who were self-resilient were more likely to be regular smokers while those who were self/family-resilient were more likely to smoke intermittently. We also plan to further explore the psychometric properties of these resilience indices, comparing them to another measure of resilience and measures of related constructs (in another sample of adolescents) to assess concurrent and discriminant validity. Future research will also examine how resilience evolves over time and how this change affects addictive behavior. Data from future waves of Add Health (1996 and 2002) will be analyzed to further assess how adolescent resilience affects behavior throughout adult life. Also, interventions should be tested which are targeted toward vulnerable adolescents, engaging them in experiential learning and problem-solving to increase resilience and health-promoting behaviors when available personal, family, and peer/community resources are insufficient [47,48].

In conclusion, we hypothesized that adolescents vary in resilience and using a nationally representative sample of middle/high school students, we examined the role of individual resilience in predicting addictive behaviors: smoking, alcohol consumption, and illegal drug use. Examining addictive behavior in adolescents, controlling for confounding environmental and personal factors, the role of resilience was robust to various model specifications and significantly explains heterogeneous risks. We suggest that an adolescent’s propensity for detrimental risky behavior, and subsequent negative outcomes, relates to environmental and personal factors, including resilience. The value of knowing that resilience among adolescents tends to occur in one of three types: overall-resilient, self/family-resilient, or self-resilient, is that programs to reduce health-compromising behaviors can be more effective and efficient if they are targeted to adolescents with specific needs, which can be identified by their type of resilience.