Profiles of Protection from Substance Useamong Adolescents

Amy K. Syvertsen & Michael J. Cleveland &

Jochebed G. Gayles & Melissa K. Tibbits &

Monique T. Faulk

# Society for Prevention Research 2009

Abstract The purpose of this study was to explore whetheradolescents (N=10,287) could be classified into homoge-neous subgroups based on their protective factors and, if so,whether these constellations of protection differentiallyrelate to adolescents’ lifetime and 30-day alcohol andtobacco use. Latent class analysis with eight protectivefactors—four internal and four external—were used toidentify the underlying latent class structure. Five profilesof protection emerged: Adequate Protection (54%), Ade-quate External Protection (9%), Adequate Protection withLow Adult Communication (16%), Adequate Protectionwith Risky Friends (9%), and Inadequate Protection (12%).Lifetime alcohol use was associated with only a modestincrease in odds of belonging to the Adequate External orLow Adult Communication latent classes, but an enormousincrease in odds of having Inadequate Protection or RiskyFriends. Similar effects were found for past month alcoholuse. Unlike alcohol use, which was related most stronglywith membership in the Risky Friends latent class (relativeto Adequate Protection), cigarette use was most stronglyrelated to membership in the Inadequate Protection latentclass. Findings can be used to inform prevention programsas they illustrate the relationships that exist betweenadolescents’ profiles of protection and substance use.

Keywords Adolescents . Protective factors . Substance use

Profiles of Protection from Substance UseAmong Adolescents

The study of adolescents in prevention science has—overthe past quarter century—undergone a metamorphosis.Whereas earlier research was steeped in the deficitorientation characteristic of the traditional medical model,prevention research today takes a decidedly more balancedperspective: acknowledging not only adolescents’ deficitsbut also their assets (see Catalano et al. 2002; Schwartz etal. 2007). Stemming, in part, from increased attention toresiliency (e.g., Masten 2001; Werner 1989) and positiveyouth development (e.g., Damon 2004; Silbereisen andLerner 2007), prevention science has come to value theimportant role adolescents’ positive dispositional attributesand attachments play in preventing risk behaviors. Shiftingfrom a deficit- to a strength-based paradigm has opened aline of inquiry in prevention science focused on the personalcharacteristics and developmental contexts that insulateagainst risk and promote well-being. To date, however,much of this empirical research has been limited, oftenfailing to demonstrate how an individual’s constellation ofprotective factors cultivates protection.

The present study builds on and extends the literature onassets that protect adolescents from substance use byexploring multiple dimensions of protection and theirrelationship with adolescents’ alcohol and tobacco use.Toward this end, we summarize empirical evidencedemonstrating the relationship between a select subset ofadolescent protective factors and substance use behaviors.Then, using a person-centered approach, we test whetheradolescents can be profiled based on the presence, orabsence, of various protective factors. Finally, we examinethe relationship between identified profiles of protectionand adolescent substance use.

A. K. Syvertsen (*) :M. J. Cleveland : J. G. Gayles :M. K. TibbitsThe Pennsylvania State University,University Park, PA, USAe-mail: syvertsen@psu.edu

M. T. FaulkUniversity of Southern California,Los Angeles, CA, USA

Prev SciDOI 10.1007/s11121-009-0154-9

Defining Protection in Adolescence

The variables that safeguard adolescents from risk behav-iors and promote well-being (e.g., school bonding, familyboundaries) have variously been called protective factors(Rutter 1979), developmental assets (Scales and Leffert2004), and supports and opportunities (Roth et al. 1998).While the differences in these terms are almost entirelysemantic (see Schwartz et al. 2007), they originate fromunique—albeit overlapping—areas of adolescent researchon prevention science and positive youth development. Inthe positive youth development paradigm, developmentalassets are routinely characterized as outcomes; whereas,prevention science typically introduces protective factors asmediators or moderators of risk behavior. In the currentstudy, we identify factors adolescents have that aid in theirwell-being as well as help to buffer or moderate underageuse of licit substances. We focus specifically on conglom-erate profiles of protection and their links to substance use.

We use the term protective factors to refer to variablesthat either decrease or delay adolescents’ alcohol, tobacco, orother drug (ATOD) use. The prevention science (Weissbergand Greenberg 1998) and positive youth development(Benson et al. 2004) literatures use a common organizationalschema for discussing protective factors: those that residewithin the individual (internal) versus those that reside intheir environment (external). Internal protective factorsinclude an individual’s health-promoting dispositional attrib-utes, commitments, values, and competencies. While, incontrast, external protective factors consist of the health-promoting aspects of adolescents’ ecology such as family,peers, school, neighborhood, and community. Internal andexternal protective factors serve complementary roles andco-occur along with various risk factors (e.g., Pollard et al.1999).

Internal Protection Against Substance Use

A variety of internal protective factors have been identifiedwithin the study of adolescent substance use (Hawkins et al.1992). Internal protective factors can be classified asaffective, cognitive, behavioral, or social (Jenson 2004).As a full examination of all internal protective factors isbeyond the scope of this study, we concentrate on oneinternal protective factor from each of these classes whichhas been linked to a lower likelihood of adolescentsubstance use behaviors; specifically, life satisfaction(affective), planful competence (cognitive), physical activity(behavioral), and positive school orientation (social).

Life Satisfaction Correlational studies, such as the researchconducted by Topolski et al. (2001), reveal a positiverelationship between adolescents’ self-reported quality of

life and abstaining from health-risk behaviors (e.g., ATODuse, risky sexual behavior). In the same vein, other researchshows that adolescents with low levels of life satisfactionare more likely to report earlier lifetime use of alcohol,cigarettes, marijuana, and cocaine compared to those withhigher life satisfaction (e.g., Zullig et al. 2001). Althoughthe specific causal pathways linking life satisfaction andsubstance use await empirical validation, extant theory (e.g.,Newcomb et al. 1986; Raphael 1996) suggestions that higherlife satisfaction may decrease (or delay) substance useindirectly by empowering adolescents with a sense of worthwhich, in turn, promotes healthy decision-making.

Planful Competence A similar indirect mechanism ishypothesized to relate planful competence (i.e., goal-setting) with substance use avoidance. Researchers positthat adolescents who set goals and have personal expect-ations regarding their future are more likely to abstain fromor defer substance use behaviors (Carroll et al. 1997;Griffin et al. 2004). In particular, adolescents who haveformulated plans (e.g., health, educational, moral) for theirfuture as well as a tentative course of action for followingthrough with these plans are more likely to make choicesand exhibit behaviors that help actualize their goals(Clausen 1991; Crockett and Silbereisen 2000). Thus, wehypothesize that adolescents who exhibit planful compe-tence will be more likely to abstain from behaviors, likesubstance use, that may compromise their health and abilityto achieve future goals.

Physical Activity Whereas life satisfaction and planfulcompetence represent affective and cognitive facets ofinternal protection, physical activity taps the more behav-ioral dimension—i.e., how one’s actions buffer against arange of physiological and psychological risks. Physicalactivities provide adolescents with alternative free timeactivities and reinforce messages about the importance ofmaintaining a healthy lifestyle (Kulig et al. 2003; Mooreand Werch 2005). A number of studies corroborate thebenefits of physical activity by showing markedly lowerrates of substance use among participants compared to non-participants (e.g., Elder et al. 2000). At odds with thisnotion, however, is evidence linking sports participation tosubstance use (Moore and Werch 2005). While an impor-tant qualification when discussing physical activity as aprotective factor, the association between participation inathletics and substance use varies widely by sport, school-sponsorship, level of involvement as well as adolescentgender and race (Aaron et al. 1995; Pate et al. 2000).Despite this caveat, for most adolescents, engaging inphysical activity makes them more aware of their bodies aswell as how the decisions they make impact themphysically.

Prev Sci

Positive School Orientation Children and adolescents’orientation towards, or attachment to, school has long beenrecognized as a key internal source of protection. Establish-ing a positive orientation towards school—an institutionwith clear standards about appropriate behavior—protectsagainst substance use by encouraging adolescents to act inways that are consistent with the expectations communi-cated by teachers, administrators, and through the formaland informal academic curriculum (Hawkins et al. 1999).Positive school orientation has been linked not only todecreased substance use, but also school dropout, delin-quency, teen pregnancy, and violence (e.g., Hawkins et al.2001; Simons-Morton et al. 1999). Collectively, studiessuggest that a positive school orientation is an importantmediating mechanism with an enduring effect on youngpeople’s developmental trajectories.

External Protection Against Substance Use

External factors that protect against substance use cutacross the ecology of adolescents’ daily lives. Aside fromsocietal factors such as laws and community anti-substanceuse norms (see Hawkins et al. 2004), most externalprotection is relational. That is, they concern the influencesof parents, friends, and other adults on adolescents’ ATODuse.

Parents Parents are often thought of as the most proximalbuffers against adolescent substance use. This importanceexplains why parents are heavily targeted by substance useprevention programs (e.g., The Adolescent TransitionsProgram; Dishion and Kavanagh 2003; StrengtheningFamilies Program 10–14; Molgaard et al. 2001). Parentsdeter adolescent substance use by modeling healthybehavior, setting limits, and supervising activities (Milleret al. 2000). By providing structure and establishingboundaries, parents relay expectations to their adolescentsregarding acceptable behavior. In contrast, low parentalmonitoring has been positively associated with adolescentATOD use and delinquency (e.g., Biglan et al. 1995;Fletcher et al. 1995). In addition to providing structure andsupervision, parents can also discourage substance use bymaking themselves physically and emotionally available.As Kafka and London (1991) demonstrated, adolescentswho perceive at least one parent to be accessible andapproachable are less inclined to turn to ATOD as a meansof dealing with problems.

Friends Although often overlooked, friendships can alsofulfill an important protective function. In contrast to themore popular image of adolescent friends as sources ofdeviant peer influence (Gilliam and Balles 2001), it is

plausible that friends who avoid risk behaviors encourageone another to make healthy choices. According to sociallearning theory (Bandura 1977), non-ATOD using friendsmay prevent one another’s use by reinforcing prosocialbehaviors (Prinstein et al. 2001) and negatively definingand sanctioning ATOD use. Controlling for selectioneffects, Maxwell (2002) illustrates both the negative andpositive sides of peer influence revealing not only that risk-engaging friends (i.e., using ATOD, having sex) predictadolescents’ own risk behavior, but also that non-riskengaging friends may encourage one another to stop oravoid some behaviors (e.g., drinking alcohol, chewingtobacco).

Non-parental Adults External protective factors extendbeyond the reach of intimate family and peer relationshipsas well. Indeed, non-parental adults (e.g., youth groupleaders, relatives, teachers) are an important source ofsocial capital for adolescents (e.g., Wooley and Bowen2007). Studies in the mentoring literature consistentlydemonstrate direct and indirect associations between ado-lescents’ involvement in non-parental adult relationshipslasting at least 12 months and positive outcomes includinglower rates of alcohol use and risk-taking (e.g., DuBois andSilverthorn 2005; Rhodes et al. 2005). Unlike parent–adolescent relationships, adolescents’ relationships withnon-parental adults are usually free of the normativeconflicts witnessed in families (e.g., disagreements aboutchores, curfews, homework). Accordingly, research con-ducted by Beam et al. (2002) showed that adolescents maybe more inclined to disclose some information to the non-parental adults in their lives than their parents. Manyadolescents in this study cited their relationship with non-parental adults as one where they could get advice from atrusted adult about sensitive topics such as drugs and sexwithout the repercussions (e.g., parent disapproval offriendship, punishment) they feared might result if theyhad the same conversation with a parent.

Aims of Present Study

The present study explored whether adolescents could beclassified into homogeneous profiles based on theirprotective factors and, if so, whether these constellationsof protection differentially relate to adolescents’ alcoholand tobacco use. The eight protective factors—four internaland four external—described above were used in thepresent research.

Past studies of protection have relied primarily onvariable-centered approaches that examine these influenceseither correlationally, in multiple regression—in which caseit is looking at its unique effects—or as part of cumulative

Prev Sci

indices that presume all factors provide equal protection.Extending recent research suggesting that the link betweenrisk factors and health outcomes is more complex than acumulative risk factor model allows (e.g., Stevens et al. 2006),the current study uses a person-centered analytic approach,which does not assume the sample is homogeneous but,instead, allows for the identification of qualitatively differentgroups of individuals. In other words, person-centeredapproaches allow us to describe the multidimensionalpatterning of protective factors as they exist in individuals.It may be that a single protective factor operates differentlyalone than in combination with other protective factors(Kraemer et al. 1997). By identifying subtypes of individualswho exhibit similar patterns of protective factors, we canlink group membership with alcohol and tobacco use.

Method

The data for this study were drawn from the eighth andtenth grade survey of the 2006 Monitoring the Futureproject (Johnston et al. 2007). Monitoring the Future is anannual survey conducted with a nationally representativesample of students enrolled in approximately 420 publicand private middle and high schools across the contermi-nous USA. Participants are recruited using a multi-stagerandom sampling procedure which selects based ongeographic area, school, and then classroom. A total of11,150 eighth and tenth grade students participated in the2006 wave of the Monitoring the Future survey. Fifty-onepercent of participants were female. The ethnic backgroundof participants was 60% White, 13% Hispanic, and 12%Black. An additional 15% of participants self-identified aseither some other ethnicity or did not respond to thequestion. Using participants’ self-reports, the approximatemedian level of parent education involved some collegeeducation.

Measures

The measures used in the present study are based onadolescents’ self-assessments. All of the internal and externalprotective factor measures were scored such that higherscores indicate greater protection. With the exception ofsinge-item indicators, all measures were standardized to easeinterpretation.

Internal Protective Factors Four measures of internalprotection were assessed: life satisfaction, planful compe-tence, physical activity, and positive school orientation. LifeSatisfaction was measured using an 11-item cognitive-judgmental assessment of one’s global quality of life (e.g.,“On the whole, I’m satisfied with myself;” α=0.90).

Planful Competence, the ability to think about futureaspirations and set attainable goals, was assessed usingtwo items: “How often do you think about your futurebeyond high school?” and “Which best describes yourplans after high school?”(r=0.39). A single item was usedto gauge Physical Activity: “How often do you activelyparticipate in sports, athletics, or exercising?” PositiveSchool Orientation was measured with seven items assess-ing adolescents’ interest in school, compliance withteachers’ behavioral expectations, and willingness to putforth effort (α=0.73).

External Protective Factors To assess the range of protectivefactors experienced across the daily contexts of adolescents’lives, four external aspects of protection were measured:parental monitoring and limit-setting, parent–child communi-cation, low friends’ATOD use, and communication with non-parental adults. Four items were combined in the ParentalMonitoring and Limit-Setting measure to tap adolescents’perceptions that their parents supervised their behavior andestablished rules (e.g., “How often do your parents check onwhether you have done your homework?” α=0.62). A singleitem gauged Parent–Child Communication: “If you werehaving problems in your life, do you think you would talkthem over with your parents?” Low Friends’ ATOD Use wasmeasured with four items that asked adolescents to reporthow many of their friends smoke cigarettes, drink alcohol,get drunk once a week, and smoke marijuana (α=0.88).These items were reverse-coded to assess the presence oflow ATOD using friends. Other Adult Communication wasassessed with one item: “Other than your parents, is thereone other adult you could talk to if you were havingproblems?”

Substance Use Behaviors Substance use behaviors wereassessed by measuring adolescents’ alcohol and tobacco use.Participants were asked separate questions about theirlifetime (experimentation) and 30-day (regular use) con-sumption (i.e., “more than just a few sips”) of alcoholicbeverages. Responses for both lifetime and 30-day use werescored on the same seven-point scale, ranging from 0occasions (1) to 40 or more occasions (7). Lifetime cigaretteuse was assessed with a single item, “Have you ever smokedcigarettes?” Participants were provided with five responseoptions: never, once or twice, occasionally but not regularly,regularly in the past, and regularly now. To measure recentcigarette use, participants were asked: “How frequently haveyou smoked cigarettes during the last 30 days?” Responseswere coded on a seven-point scale (1=Not at All to 7=2+Packs/Day). Because the distributions of the substance usebehaviors were skewed toward non-use, dichotomouslifetime and 30-day use variables were created to indicateuse vs. non-use of each substance.

Prev Sci

Missing Data

To be included in the present analyses, adolescents needed toprovide data on: (a) at least one of the internal or externalprotection variables (i.e., indicator variables), and (b) theirlifetime and 30-day substance use behaviors (i.e., covariates).Parameters are estimated by maximum likelihood using theEM algorithm (Lanza et al. 2008). Use of this procedureallows us to utilize data from participants who may not haveresponded to all of the internal and external protectionvariables. As missing data on the covariates cannot beaccounted for in the latent class model, our sample had to berestricted to participants with complete data on the substanceuse measures (N=10,287). Table 1 provides additional detailregarding patterns of missingness. A series of t-tests wereconducted to compare the means of the internal and externalprotective factors between the full sample and the restrictedsample used in the analyses. The results indicated that, withone exception (Other Adult Communication), adolescentsexcluded from the analyses due to incomplete data on thesubstance use outcomes reported lower levels of protectionthan those with complete data (all t-values>2.00, ps<0.05).Thus, the current sample may underrepresent adolescentswho are at high risk.

Analysis Plan

Latent class analysis (LCA) is a statistical model designedto answer the question: Are there underlying types orgroups of individuals who share certain characteristics?That is, LCA models estimate a categorical latent variablethat divides a population into mutually exclusive andexhaustive latent classes (Goodman 1974). In the currentstudy, internal and external protective factors were used ascategorical indicators of latent classes of adolescentprotection from substance use. Dichotomizing these indi-cators allows us to identify the underlying subgroups ofindividuals by simplifying what would otherwise (ifdimensional latent indicators were used) be a large complexarray of data. Protective factors can, arguably, be oper-ationalized according to their continuous (degree on aresponse scale) or discrete (adequate vs. inadequate) levelsof presence in individuals (Collins and Lanza 2009). Usingthe latter approach as the lens through which we examineour data allows us to distinguish the various profiles ofprotection based on whether adolescents have achievedadequate levels of protection for each of the internal andexternal factors. This approach to dichotomization allowsus to concentrate our analyses on the presence of protectionand facilitates the identification of those adolescents whoare not at high risk for substance use due to the synergisticaction of multiple protective factors (Farrington and Loeber2000; Kraemer et al. 1997).

The LCA model can be expressed as a function of twosets of parameters. First, the latent class membershipprobabilities represent the proportion of the population ineach latent class. Second, the conditional item-responseprobabilities represent the distribution of responses to eachmeasured item, within each latent class. More details aboutthe LCA statistical model and an empirical example appearin Lanza et al. (2007).

An important extension of LCA allows exogenouspredictors (i.e., covariates) to be added to the basic modelvia multinomial logistic regression (Lanza et al. 2007). Thecurrent study explored the relation between the latent classesof protection and four measures of substance use—lifetimeand 30-day use of alcohol and cigarettes. These modelsrepresent the increase in odds of belonging to a certain latentclass, relative to a reference class, given that an adolescentreported the behavior (e.g., used cigarettes in the last30 days). All analyses were conducted using PROC LCA(Lanza et al. 2008).

Results

Model Specification and Selection

In order to specify the latent class models, the distributionsof the four internal and four external indicators wereexamined, then dichotomized (see Table 1). The distribu-tions of all the indicators were skewed toward higher levels ofprotection; thus, we selected a cut point that best allowed us todistinguish between adolescents who were at the extreme lowend and those at the more moderate to high ends of thedistribution. For the five scale indices (life satisfaction, planfulcompetence, positive school orientation, parents’ monitoringand limit-setting, and low friends’ ATOD use), a score ofinadequate (1) was assigned to those in the bottom 25thpercentile and a score of adequate (2) was given to thoseabove the 25th percentile. Similarly, single–item indices(physical activity, parent–child communication, other adultcommunication) were assigned a value of adequate (2) if aparticipant’s response indicated—based on the extant litera-ture discussed above—a satisfactory level of protection.

A series of latent class models were compared todetermine the optimal model in terms of balancing modelfit and parsimony. Using several indicators including thelikelihood-ratio G2 statistic, Akaike’s Information Criterion(AIC; Akaike 1974), the Bayesian Information Criterion(BIC; Schwartz 1978), and interpretability of results, it wasdetermined that the five-class latent model provided a moreoptimal solution than the one-, two-, three-, or four-classmodels (G2[df=211]=327.92, AIC=415.92, BIC=734.42).Specifically, compared to models with one to four classes,the five-class model had the lowest AIC value and only a

Prev Sci

slightly higher BIC value than the four-class model (BIC=733.42). We also considered the interpretability, degree oflatent class separation (i.e., degree to which the latentclasses can be clearly distinguished from one another), andsize of the latent classes in selecting the optimal model. Thelatent classes in the five-class solution were easilyinterpretable, showed high latent class separation, and nosingle class comprised less than 5% of the sample. Themodal G2 of the five-class solution was also the lowest in72 out of 100 random starting values indicating the solutionwas adequately identified. Additional models with six- andseven-classes were also explored; however, these more

complex models exhibited identification and convergenceproblems with no single best solution emerging.

Latent Classes of Protection

Table 2 presents the prevalence estimates for the five latentclasses and the probability that members of each classreported an adequate level of protection for each of theeight protective factors (i.e., received a score of “2” for theindicator). According to the model, Adequate Protectionwas the most common latent class, comprising more thanhalf the sample (54%). This class was characterized by very

Code Label Frequency (valid %)

Indicator variables

Internal protective factors

Life satisfaction 1 Inadequatea 2,237 (21.8%)

2 Adequate 6,986 (67.9%)

Missing 1,064 (10.3%)

Planful competence 1 Inadequatea 2,231 (21.7%)

2 Adequate 7,833 (76.1%)

Missing 223 (2.2%)

Physical activity 1 Inadequateb 2,610 (25.4%)

2 Adequate 7,633 (74.2%)

Missing 44 (0.4%)

Positive school orientation 1 Inadequatea 2,483 (14.1%)

2 Adequate 7,772 (75.6%)

Missing 32 (0.3%)

External protective factors

Parents’ monitoring/limit-setting 1 Inadequatea 2,004 (19.5%)

2 Adequate 6,016 (58.5%)

Missing 2,267 (22.0%)

Parent–child communication 1 Inadequatec 2,105 (20.5%)

2 Adequate 5,871 (57.1%)

Missing 2,311 (22.5%)

Low friends’ ATOD use 1 Inadequatea 2,393 (23.3%)

2 Adequate 7,321 (71.2%)

Missing 573 (5.6%)

Other adult communication 1 Inadequatec 1,871 (18.2%)

2 Adequate 6,100 (59.3%)

Missing 2,316 (22.5%)

Substance use behaviors covariates

Lifetime alcohol use 1 Use 5,025 (48.9%)

2 Non-use 5,262 (51.2%)

30-day alcohol use 1 Use 7,702 (74.9%)

2 Non-use 2,585 (25.1%)

Lifetime cigarette use 1 Use 7,164 (69.6%)

2 Non-use 3,123 (30.4%)

30-day cigarette use 1 Use 9,144 (88.9%)

2 Non-use 1,143 (11.1%)

Table 1 Creation ofdichotomous latent classindicators and covariates

a The standardized indicator wasdichotomized using quartiles:bottom 25% (inadequate) andtop 75% (adequate)b The original (single-item)five-point response scale wasdichotomized such that physicalactivity once or twice a month, afew times a year, or never wascoded as inadequate and activityat least once a week or almostevery day was coded asadequatec The original (single-item)three-point response scale wascoded: no (inadequate) andsome, most or all (adequate).

Prev Sci

high probabilities of reporting adequate protection acrossall eight protective indictors. In contrast, a small—butnoteworthy—proportion of adolescents (11%) belonged tothe Inadequate Protection latent class, set apart by low ormoderate probabilities of endorsing each of the eightprotective factors. The remaining three latent classes weredifferentiated by a mix of adequate and inadequateprotection for each of the internal and external protectivefactors. Approximately 10% of participants were in theAdequate External Protection class, which was characterizedby a high probability of endorsing the external factors butlow or moderate probabilities of each of the internal factors,particularly positive school orientation. The AdequateProtection with Low Adult Communication (Low AdultCommunication) latent class (17%) was comprised ofadolescents with low probabilities of talking to their parents(0.31) or other adults (0.44) about their problems, butmoderate to high probabilities of endorsing each of theremaining six protective factors. The final latent class,Adequate Protection with Risky Friends (Risky Friends),revealed a similar pattern. This class, represented by 9% ofthe sample, was distinguished by adequate protection oneach of the internal factors and three of the four externalfactors, but a low probability (0.25) of having non-ATODusing friends (i.e., high levels of ATOD using friends).

Relating Latent Class Membership and Substance Use

In the next set of analyses, the LCA model was extended toinclude the four substance use behaviors. Binary indicators

of lifetime and 30-day use of alcohol and cigarettes wereentered into these models as covariates to predict latentclass membership. Prior research suggests that associationsbetween risk and protective factors may differ across typesof substances or stages of use (Cleveland et al. 2008; Flay etal. 1998; Jackson 1997). Thus, each substance use behaviorwas added in a separate multinomial logistic regressionmodel with the Adequate Protection latent class serving asthe reference class.

Figure 1 shows the log-odds of belonging to each latentclass, relative to the Adequate Protection class, given thatthe adolescent had used alcohol in their lifetime (top panel)or in the past 30 days (bottom panel). As seen in the figure,lifetime alcohol use was associated with only a modestincrease in odds of belonging to the Adequate External orLow Adult Communication latent classes, but an enormousincrease in odds of having Inadequate Protection (OR=35.46) or Risky Friends (OR=81.88). Similar effects werefound for past month alcohol use. Unlike alcohol use,which was related most strongly with membership in theRisky Friends latent class (relative to Adequate Protection),cigarette use was most strongly related to membership inthe Inadequate Protection latent class relative to the otherlatent classes (see Fig. 2).

Discussion

Given that these five protective profiles showed differentialassociations with regard to alcohol and cigarette use it can

Table 2 Latent class membership probabilities and conditional item-response probabilities of latent class membership

Conditional item-response probabilities of adequate protectiona

InadequateProtection

AdequateExternal

Adequate Protection w/Low Adult Communication

Adequate Protection w/Risky Friends

AdequateProtection

Latent class membership probabilities 11% 10% 17% 9% 54%

Internal protective factors

Life satisfaction 0.29b 0.49c 0.62d 0.81d 0.93d

Planful competence 0.53c 0.55c 0.77d 0.88d 0.86d

Physical activity 0.56c 0.53c 0.71d 0.82d 0.82d

Positive school orientation 0.19b 0.39b 0.83d 0.65d 0.94d

External protective factors

Parents’ monitoring/limit-setting 0.43c 0.62d 0.63d 0.67d 0.89d

Parent–child communication 0.16b 1.00d 0.31b 0.71d 0.95d

Low friends’ ATOD use 0.42c 0.61d 0.86d 0.25b 0.90d

Other adult communication 0.44c 0.86d 0.44c 0.94d 0.89d

a Item-response probabilities indicate a response of adequate protection (i.e., received a score of “2” for the indicator)b Low (<0.40) strength of the probabilitycModerate (0.40–0.60) strength of the probabilityd High (>0.60) strength of the probability

Prev Sci

be concluded that the relationships among protective factorsand substance use behaviors are not uniform for alladolescents. Modeling both internal and external factorsconcurrently allowed us to capture the interactive relation-ships that exist between these factors within individuals(Magnusson 1988, 1995). Rather than examining the linear

associations that exist between variables like most previousresearch, the present study examined how protective factorswere organized within adolescents and then grouped thosewho exhibited similar patterns. Aligned with our interest incapturing the co-occurrence of various protective factors,this person-centered approach provided more information

35.5

1.5 1.6

81.9

1.0

0

10

20

30

40

50

60

70

80

90

100

Inadequate

Protection

Adequate External Adequate Protection

w/ Low Adult

Communication

Adequate Protection

w/ Risky Friends

Adequate Protection

21.5

2.5 1.5

29.1

1.0

0

10

20

30

40

50

60

70

80

90

100

Inadequate

Protection

Adequate External Adequate Protection

w/ Low Adult

Communication

Adequate Protection

w/ Risky Friends

Adequate Protection

Fig. 1 The log-odds of belong-ing to each latent class, relativeto the Adequate Protection class,given that the adolescent hadused alcohol in their lifetime(top panel) or in the previous30 days (bottom panel). Note.Adequate Protection is thereference class

44.3

3.1 2.2

25.4

1.0

0

10

20

30

40

50

60

70

80

90

100

Inadequate

Protection

Adequate External Adequate Protection

w/ Low Adult

Communication

Adequate Protection

w/ Risky Friends

Adequate Protection

97.2

5.2 2.6

49.8

1.0

0

10

20

30

40

50

60

70

80

90

100

Inadequate

Protection

Adequate External Adequate Protection

w/ Low Adult

Communication

Adequate Protection

w/ Risky Friends

Adequate Protection

Fig. 2 The log-odds of belong-ing to each latent class, relativeto the Adequate Protection class,given that the adolescent hadsmoked cigarettes in theirlifetime (top panel) or in theprevious 30 days (bottompanel). Note. AdequateProtection is the reference class

Prev Sci

about the multiplicative nature of protection in adolescents(Bergman et al. 2003).

Specifically, according to these data, there exist sub-populations of adolescents for whom alcohol and cigaretteuse differ depending on their constellation of protectivefactors. Not surprisingly, individuals in the AdequateProtection group were at lowest risk for substance use asassessed by both lifetime and recent alcohol and cigaretteuse. This finding corresponds with studies demonstrating adirect relationship between protective factors and abstainingfrom ATOD (e.g., Jelicic et al. 2007; Theokas and Lerner2006).

Given the important role that parents and adults play asprotective factors (see Hawkins et al. 1992; Wooley andBowen 2007), it was revealing to find that the use ofalcohol or cigarettes was associated with a small increase inthe odds of membership in the Adequate Protection withLow Adult Communication group relative to the AdequateProtection group. This finding suggests that in the contextof multiple other protective factors (i.e., internal protection,non-ATOD using peers, parental monitoring/limit-setting)which may serve a compensatory role, the absence ofcommunication with parents and adults may be offset. Thesame “compensatory” reasoning does not, however, holdfor the Adequate Protection with Risky Friends group.Despite reporting levels of internal and external protectioncomparable to their peers in the Adequate Protection classfor seven of the eight protective factors considered, thesetwo groups showed substantially different associations withalcohol and cigarette use. This corroborates previousresearch showing that adolescents who use ATOD likelyhave friends who engage in similar behaviors (e.g., Hamm2000). Our analysis further bolsters the argument thatfriends’ influence is pervasive in adolescence by showingthat the relationship between friends’ ATOD use andadolescents’ ATOD use holds despite the adequate presenceof other internal and external protective factors. Looking atthe low friends’ ATOD use indicator across the profilesreveals another possible interpretation for the AdequateProtection with Low Adult Communication group’s rela-tively low association with alcohol or cigarette use.Adolescents fitting this profile reported—much like theAdequate Protection group—a very high probability ofhaving friends who do not use ATOD. It is possible that nothaving parents and other adults to share problems with is ofless consequence for these youth in light of their positiveorientation towards school and having friends who avoidATOD.

The Risky Friends and Inadequate Protection groups hadthe highest levels of association with alcohol and cigaretteuse. While alcohol use (lifetime and 30-day) was associatedwith substantially greater odds of belonging to the RiskyFriends group, adolescents who reported lifetime and 30-day

cigarette use were most likely to belong to the InadequateProtection group. These results suggest that the link betweenprotection and ATOD use may differ based on the type ofsubstance (Cleveland et al. 2008; Flay et al. 1998; Jackson1997). Finding that alcohol use, more than cigarette use, wasstrongly associated with having few non-ATOD usingfriends (or, stated in reverse, having ATOD using friends)may be indicative of the social and addictive nature of thesesubstances: While, for most adolescents, alcohol consump-tion is primarily a social activity done with a group offriends, the addictive power of nicotine compels adolescentsto smoke when the craving hits regardless of whether theyhave a friend to join in (Gardner and Steinberg 2005).

A comparison across the five profiles of protectionreinforces the idea that protective factors are not equallyweighted. While it is beyond the scope of the modelspresented here to assign a protective value to each factor, thedifferential associations between the profiles—each with aunique constellation of adequate protective factors—andadolescent substance use behaviors suggests inadequatelevels of some protective factors (e.g., low ATOD-usingfriends) are more consequential than others.

Limitations and Future Research

In interpreting these findings a few limitations should benoted. First, the protective factors and substance usebehaviors examined in this study were not exhaustive.Thus, we do not ascribe these findings to the wholeconglomerate of protection in adolescence. Religiosity andcommunity involvement, for example, are not representedin our model but both have been empirically shown toprotect against alcohol and drug use (Oman et al. 2004).Additionally, because we only looked at profiles ofprotection, our study tells us little about the presence ofrisk factors or how the constellation of risk and protectivefactors together are linked with substance use. Thesignificant interaction of risk and protection in theprediction of problem behaviors, as demonstrated by Jessoret al. (1995), punctuates this point. Similarly, Pollard et al.(1999) provide a persuasive cross-sectional demonstrationsuggesting that simply promoting protective factors is anincomplete solution for reducing the prevalence of sub-stance use and delinquency in adolescents. This was alsotrue for risk: Exclusively reducing risk factors proved justas ineffective. In short, risk and protective factors need tobe considered separately as well as in concert to ascertainboth their unique and multiplicative contributions toadolescent outcomes.

Second, the data used for these analyses, althoughnationally representative, were cross-sectional (eighth andtenth grade combined). Specifically, it is unknown whetherthese latent classes are causally related to substance use.

Prev Sci

Similarly, it is unknown how the presence or absence ofprotective factors changes over time. Thus, no conclusionsare drawn with respect to causality or developmentalprocesses. It is only suggested that during adolescencethere are interesting and differential associations among theidentified profiles of protection and substance use.

Building on the present study, future research shouldexamine demographic differences (e.g., sex, race, SES) inthe profiles of protective factors, including comparisons ofeighth and tenth graders. Such research would allow us toelucidate the culturally, socially, and historically boundnature of adolescent protection (Masten and Curtis 2000).Similarly, future research may consider relating profiles ofadolescent protection to prosocial behaviors in addition toindicators of psychological well-being. The field has cometo recognize that psychological well-being, and better optimaldevelopment, is more than the absence of problems. To bewell, one must have the skills, competencies, and copingresources necessary to thrive and overcome difficulties(Cowen 1994). Protective factors are integral to improvingpsychological well-being (Durlak 1995). Knowing howadolescents’ diverse profiles of protection relate to substanceuse does not allow us to make substantial claims regardingtheir well-being.

Implications for Prevention Science

Findings from this study have important implications forour understanding of co-occurring protective factors as wellas the design of prevention programs targeting adolescentalcohol and cigarette use. As a science, we are movingtoward creating a more holistic understanding of individu-als (Bergman and Magnusson 1997; von Eye and Bergman2003). This involves seeing individuals as an organizedwhole and focusing on the patterns of risk and protectionoperating at the individual level. Understanding how riskand protective factors cluster in adolescents providesimportant insight on how to negate or promote healthyfunctioning (Farrell 2008). By identifying unique profilesof protection for adolescents, these results illustrate thepositive synergistic effect of co-occurring protective factorsas well as the negative impact of missing key sources ofprotection. As illustrated with the Adequate Protection withRisky Friends class, the absence of a single protective factorcan be linked to significantly higher odds of using alcoholand cigarettes; thus, underscoring the need for promotiveprevention efforts that target not only individual adoles-cents but also the full ecology of their daily lives (Dishionet al. 2002). While implementing such programs may bechallenging, collaborative family, school, community, andresearch partnerships may remedy some of these barriers.For optimal development, adolescents need protectionacross individual, family, peer, school, and community

levels. This, like many others, we argue should be theprimary endeavor of prevention science.

When it comes to prevention, adolescents have differentneeds, different risks, and different sources of protection.Although most ATOD prevention programming is deliveredin a group or classroom setting, prevention research canbenefit from important information obtained when consid-ering individual-level differences in the protective factorsthat buffer against substance use behaviors. For example,adaptive intervention strategies that tailor the particulardosage or type of treatment across individuals in order tomatch the intervention strategy with the level of theindividual’s risk or the absence of a particular protectivefactor have emerged as new approaches (Collins et al.2004). Person-centered research on risk and protection shedlight on these differences and create a unique opportunity todevelop empirically-based interventions that focus on theindividual as a whole and as they exist in the real world.

Acknowledgements An earlier version of this paper was presentedat the Society for Prevention Research conference (2008, May) in SanFrancisco, CA as part of the Sloboda and Bukoski SPR Cup Competition(Eddy & Martinez 2008). The authors’ time was supported by NIDAtraining grants T32 DA017629, F31 DA024535, and F31 DA024916-01 and NIDA Center Grant P50 DA100075 to The Methodology Centerat The Pennsylvania State University. The contents of the article aresolely the responsibility of the authors and do not necessarily representthe official views of the National Institutes of Health.

The authors would like to thank Stephanie Lanza, PhD and MarkGreenberg, PhD for their valuable and constructive comments onearlier drafts of the manuscript.

References

Aaron, D.J., Dearwater, S.R., Anderson, R., Olsen, T., Kriska, A.M.,& Laporte, R.E. (1995). Physical activity and the initiation ofhigh-risk health behaviors in adolescents. Medicine & Science inSports & Exercise, 27, 1639–1645.

Akaike, H. (1974). A new look at the statistical model identification.IEEE Transactions on Automatic Control, 19, 716–723.

Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ:Prentice Hall.

Beam, M.R., Chen, C., & Greenberger, E. (2002). The nature ofadolescents’ relationships with their “very important” nonpar-ental adults. American Journal of Community Psychology, 30,305–325.

Benson, P.L., Mannes, M., Pittman, K., & Ferber, T. (2004). Youthdevelopment, developmental assets, and public policy. In R. M.Lerner & L. Steinberg (Eds.), Handbook of adolescent psychology(pp. 781–814, 2nd ed.). New York: Wiley.

Bergman, L.R., & Magnusson, D. (1997). A person-oriented approachin research on developmental psychopathology. Development andPsychopathology, 9, 291–319.

Bergman, L.R., Magnusson, D., & El-Khouri, B.M. (2003). Studyingindividual development in an interindividual context: A person-oriented approach. Mahwah, NJ: Lawrence Erlbaum.

Biglan, A., Duncan, T.E., Ary, D.V., & Smolkowski, K. (1995). Peerand parental influences on adolescent tobacco use. Journal ofBehavioral Medicine, 18, 315–330.

Prev Sci

Carroll, A., Durkin, K., Hattie, J., & Houghton, S. (1997). Goal settingamong adolescents: A comparison of delinquent, at-risk, and not-at-risk youth. Journal of Educational Psychology, 89, 441–450.

Catalano, R.F., Hawkins, J.D., Berglund, M.L., Pollard, J.A., &Arthur, M.W. (2002). Prevention science and positive youthdevelopment: Competitive or cooperative frameworks? Journalof Adolescent Health, 31, 230–239.

Clausen, J.S. (1991). Adolescent competence and the shaping of thelife course. American Journal of Sociology, 96, 805–842.

Cleveland, M.J., Feinberg, M.E., Bontempo, D.E., & Greenberg, M.T.(2008). The role of risk and protective factors in substance useacross adolescence. Journal of Adolescent Health, 43, 157–164.

Collins, L.M., & Lanza, S.T. (2009). Latent class and latent transitionanalysis for applications in the social, behavioral, and healthsciences. New York: Wiley (in press).

Collins, L.M., Murphy, S.A., & Bierman, K. (2004). A conceptualframework for adaptive preventive interventions. PreventionScience, 3, 185–196.

Cowen, E. L. (1994). The enhancement of psychological wellness:Challenges and opportunities. American Journal of CommunityPsychology, 22, 149–179.

Crockett, L.J., & Silbereisen, R.K. (2000). Social changes andadolescent development: issues and challenges. In L. J. Crockett& R. K. Silbereisen (Eds.), Negotiating adolescence in times ofsocial change (pp. 1–13). New York: Cambridge UniversityPress.

Damon, W. (2004). What is positive youth development? Annals ofthe American Academy of Political and Social Science, 591, 13–24.

Dishion, T.J., & Kavanagh, K. (2003). Intervening in adolescentproblem behavior: A family-centered approach. New York:Guilford.

Dishion, T.J., Kavanagh, K., Schneiger, A., Nelson, S., & Kaufman,N.K. (2002). Preventing early adolescent substance use: Afamily-centered strategy for the public middle school. PreventionScience, 3, 191–201.

DuBois, D.L., & Silverthorn, N. (2005). Natural mentoring relation-ships and adolescent health: Evidence from a national study.American Journal of Public Health, 95, 518–524.

Durlak, J.A. (1995). School-based prevention programs for childrenand adolescents. Thousand Oaks, CA: Sage.

Eddy, J.M., & Martinez, C.R. (2008). An introduction to the Slobodaand Bukoski Society for Prevention Research (SPR) Cup.Prevention Science, 9, 1–3.

Elder, C., Leaver-Dunn, D., Wang, M.Q., Nagy, S., & Green, L.(2000). Organized group activity as a protective factor againstadolescent substance use. American Journal of Health Behavior,24, 108–113.

Farrell, A.D. (2008). The Multisite Violence Prevention Project:Impact of a universal school-based violence prevention programon social-cognitive outcomes. Prevention Science, 9, 231–244.

Farrington, D.P., & Loeber, R. (2000). Some benefits of dichotomi-zation in psychiatric and criminological research. CriminalBehavior and Mental Health, 10, 100–122.

Flay, B.R., Hu, F.B., & Richardson, J. (1998). Psychosocial predictorsof different stages of cigarette smoking among high schoolstudents. Preventive Medicine, 27, A9–A18.

Fletcher, A.C., Darling, N., & Steinberg, L. (1995). Parentalmonitoring and peer influences on adolescent substance use. InJ. McCord (Ed.), Coercion and punishment in long-termperspective (pp. 259–271). New York: Cambridge UniversityPress.

Gardner, M., & Steinberg, L. (2005). Peer influence on risk taking,risk preference, and risky decision making in adolescence andadulthood: An experimental study. Developmental Psychology,41, 625–635.

Gilliam, F.D., & Balles, S.N. (2001). Strategic frame analysis:Reframing America’s youth. Social Policy Report, XV, 65.

Griffin, K.W., Botvin, G.J., Nichols, T.R., & Scheier, L.M. (2004).Low perceived chances for success in life and binge drinkingamong inner-city minority youth. Journal of Adolescent Health,34, 501–507.

Goodman, L.A. (1974). Exploratory latent structure analysis usingboth identifiable and unidentifiable models. Biometrika, 61, 215–231.

Hamm, J.V. (2000). Do birds of a feather flock together? The variablebases for African American, Asian American, and EuropeanAmerican adolescents’ selection of similar friends. Developmen-tal Psychology, 36, 209–219.

Hawkins, J.D., Catalano, R.F., & Miller, J.Y. (1992). Risk andprotective factors for alcohol and other drug problems inadolescence and early adulthood: Implications for substanceabuse prevention. Psychological Bulletin, 112, 64–105.

Hawkins, J.D., Catalano, R.F., Kosterman, R., Abbott, R., & Hill, K.G. (1999). Prevention of adolescent health-risk behaviors bystrengthening protection during childhood. Archives of Pediatrics& Adolescent Medicine, 153, 226–234.

Hawkins, J.D., Guo, J., Hill, K.G., Battin-Pearson, S., & Abbott, R.D.(2001). Long-term effects of the Seattle Social DevelopmentIntervention on school bonding trajectories. Applied DevelopmentalScience, 5, 225–236.

Hawkins, J.D., Van Horn, M.L., & Arthur, M.W. (2004). Communityvariation in risk and protective factors and substance useoutcomes. Prevention Science, 5, 213–220.

Jackson, C. (1997). Initial and experimental stages of tobacco andalcohol use during late childhood: Relation to peer, parent, andpersonal risk factors. Addictive Behavior, 22, 685–698.

Jelicic, H., Bobek, D.L., Phelps, E., Lerner, R.M., & Lerner, J.V.(2007). Using positive youth development to predict contributionand risk behaviors in early adolescence: Findings from the firsttwo waves of the 4-H Study of Positive Youth Development.International Journal of Behavioral Development, 31, 263–273.

Jenson, J.M. (2004). Risk and protective factors for alcohol and otherdrug use in childhood and adolescence. In M. W. Fraser (Ed.),Risk and resilience in childhood: An ecological perspective (pp.183–208, 2nd ed.). Washington, DC: NASW.

Jessor, R., Van Den Bos, J., Vanderryn, J., Costa, F.M., & Turbin, M.S. (1995). Protective factors in adolescent problem behavior:Moderator effects and developmental change. DevelopmentalPsychology, 31, 923–933.

Johnston, L.D., Bachman, J.G., & O’Malley, P.M. (2007). Monitoringthe Future: A continuing study of the lifestyles and values ofyouth (12th Grade Survey), 1977–2006. [Data file]. Conductedby University of Michigan, Survey Research Center. Ann Arbor:Inter-university Consortium for Political and Social Research,Institute for Social Research.

Kafka, R.R., & London, P. (1991). Communication in relationshipsand adolescent substance use: The influence of parents andfriends. Adolescence, 26, 587–598.

Kraemer, H.C., Kazdin, A.E., Offord, D.R., Kessler, R.C., Jensen, P.S., & Kupfer, D.J. (1997). Coming to terms with the terms ofrisk. Archives of General Psychiatry, 54, 337–343.

Kulig, K., Brener, N.D., & McManus, T. (2003). Sexual activity andsubstance use among adolescents by category of physical activityplus team sports participation. Archives of Pediatrics andAdolescent Medicine, 157, 905–912.

Lanza, S.T., Collins, L.M., Lemmon, D.R., & Schafer, J.L. (2007).PROC LCA: A SAS procedure for latent class analysis.Structural Equation Modeling, 14, 671–694.

Lanza, S.T., Lemmon, D.R., Schafer, J.L., & Collins, L.M. (2008).PROC LCA user’s guide version 1.1.5. University Park:Pennsylvania State University, The Methodology Center.

Prev Sci

Magnusson, D. (1988). Individual development from an interactionalperspective: A longitudinal study. Hillsdale, NJ: LawrenceErlbaum.

Magnusson, D. (1995). Individual development: A holistic, integratedmodel. In P. Moen, G. H. Jr Elder, & K. Lüscher (Eds.),Examining lives in context: Perspectives on the ecology of humandevelopment (pp. 19–60). Washington, DC: American Psycho-logical Association.

Masten, A.S. (2001). Ordinary magic: Resilience processes indevelopment. American Psychologist, 56, 227–238.

Masten, A.S., & Curtis, W.J. (2000). Integrating competence andpsychopathology: Pathways toward a comprehensive science ofadaptation in development. Development and Psychopathology,12, 529–550.

Maxwell, K.A. (2002). Friends: The role of peer influence acrossadolescent risk behaviors. Journal of Youth and Adolescence, 31,267–277.

Miller, M.A., Alberts, J.K., Hecht, M.L., Trost, M., & Krizek, R.L.(2000). Adolescent relationships and drug use. Hillsdale, NJ:Lawrence Erlbaum.

Molgaard, V.K., Kumpfer, K., & Fleming, E. (2001). StrengtheningFamilies Program: For parents and youth 10-13. A video basedcurriculum. (Leader Guide). Ames: Iowa State UniversityExtension.

Moore, M.J., & Werch, C.E. (2005). Sport and physical activityparticipation and substance use among adolescents. Journal ofAdolescent Health, 36, 486–493.

Newcomb, M.D., Bentler, P.M., & Collins, C. (1986). Alcohol use anddissatisfaction with self and life: A longitudinal analysis of youngadults. Journal of Drug Issues, 63, 479–494.

Oman, R.F., Vesely, S., Aspy, C., McLeroy, K.R., Rodine, S., &Marshall, L. (2004). The potential protective effect of youthassets on adolescent alcohol and drug use. American Journal ofPublic Health, 94, 1425–1430.

Pate, R.R., Trost, S.G., Levin, S., & Dowda, M. (2000). Sportsparticipation and health-related behaviors among US youth.Archives of Pediatrics & Adolescent Medicine, 154, 904–911.

Pollard, J.A., Hawkins, J.D., & Arthur, M.W. (1999). Risk andprotection: Are both necessary to understand diverse behavioraloutcomes in adolescence? Social Work Research, 23, 145–157.

Prinstein, M.J., Boergers, J., & Spirito, A. (2001). Adolescents’and their friends’ health-risk behaviors: Factors that alter oradd to peer influence. Journal of Pediatric Psychology, 26,287–298.

Raphael, D. (1996). The determinants of adolescent health: Evolvingdefinitions, recent findings, and proposed research agenda.Journal of Adolescent Health, 19, 6–16.

Rhodes, J.E., Reddy, R., & Grossman, J.B. (2005). The protectiveinfluence of mentoring on adolescents’ substance use: Directand indirect pathways. Applied Developmental Science, 9, 31–47.

Roth, J., Brooks-Gunn, J., Murray, L., & Foster, W. (1998). Promotinghealthy adolescents: Synthesis of youth development programevaluations. Journal of Research on Adolescence, 8, 423–459.

Rutter, M. (1979). Protective factors in children’s response to stressand disadvantage. In W. M. Kent & J. E. Rolf (Eds.), Primaryprevention of psychopathology, vol. 3 (pp. 49–74). Hanover, NH:University Press of New England.

Scales, P.C., & Leffert, N. (1999).Developmental assets: A synthesis ofthe scientific research on adolescent development. Minneapolis,MN, US: Search Institute.

Schwartz, G. (1978). Estimating the dimension of a model. Annals ofStatistics, 6, 461–464.

Schwartz, S.J., Pantin, H., Coatsworth, J.D., & Szapocznik, J. (2007).Addressing the challenges and opportunities for today’s youth:Toward an integrative model and its implications for research andintervention. Journal of Primary Prevention, 28, 117–144.

Silbereisen, R.K., & Lerner, R.M. (Eds.) (2007). Approaches topositive youth development. London: Sage.

Simons-Morton, B.G., Crump, A.D., Haynie, D.L., & Saylor, K.E.(1999). Student–school bonding and adolescent problem behavior.Health Educational Research, 14, 99–107.

Stevens, G.D., Seid, M., Mistry, R., & Halfon, N. (2006). Disparitiesin primary care for vulnerable children: The influence of multiplerisk factors. Health Services Research, 41, 507–531.

Theokas, C., & Lerner, R.M. (2006). Observed ecological assets infamilies, schools, and neighborhoods: Conceptualization, mea-surement and relations with positive and negative developmentaloutcomes. Applied Developmental Science, 10, 61–74.

Topolski, T.D., Patrick, D.L., Edwards, T.C., Huebner, C.E., Connell,F.A., & Mount, K.K. (2001). Quality of life and health-riskbehaviors among adolescents. Journal of Adolescent Health, 29,426–435.

von Eye, A., & Bergman, L.R. (2003). Research strategies indevelopmental psychopathology: Dimensional identity and theperson-oriented approach [Special Issue: Conceptual, methodo-logical, and statistical issues in developmental psychopathology:A special issue in honor of Paul E. Meehl]. Development andPsychopathology, 15, 553–580.

Weissberg, R.P., & Greenberg, M.T. (1998). School and communitycompetence-enhancement and prevention programs. In W. Damon(Series Ed.), I. E. Sigel, & K. A., Renniger (Vol. Eds.), Handbookof child psychology: Vol. 4. Child psychology in practice (5th ed.,pp. 877–954). New York: Wiley.

Werner, E.E. (1989). Children of the garden island. ScientificAmerican, 260, 107–111.

Wooley, M.E., & Bowen, G.L. (2007). In the context of risk:Supportive adults and the school engagement of middle schoolstudents. Family Relations, 56, 92–105.

Zullig, K.J., Valois, R.T., Huebner, E.S., Oeltmann, J.E., & Drane, W.(2001). Relationship between perceived life satisfaction andadolescents’ substance abuse. Journal of Adolescent Health, 29,279–288.

Prev Sci