Incarcerated Adolescents' Experiences as Perpetrators of Sexual Assault | Adolescent Medicine | JAMA Pediatrics | JAMA Network
[Skip to Navigation]
Sign In
August 2002

Incarcerated Adolescents' Experiences as Perpetrators of Sexual Assault

Author Affiliations

From the Department of Pediatrics, UCLA School of Medicine, Los Angeles, Calif (Drs Morris and Anderson); and the National Gang Crime Research Center, Chicago State University, Chicago, Ill (Dr Knox).

Arch Pediatr Adolesc Med. 2002;156(8):831-835. doi:10.1001/archpedi.156.8.831

Objectives  To identify self-reported characteristics of past sexual assaults perpetrated by incarcerated juveniles and to determine potential precursors of this behavior.

Participants and Methods  Participants completed an anonymous written questionnaire regarding their experiences of sexual abuse and/or assault as both a victim and a perpetrator. The questionnaire was incorporated into an existing annual survey of gang-related behaviors.

Results  A total of 805 adolescents (707 boys, 91 girls, and 7 gender-not-designated) participated. Of these teens, 79 males (11% of the males), 9 females (10% of the females), and 5 (71%) gender-not-designated said they had forced sex on someone (165 did not answer the question). Twenty-eight males (30% of the male perpetrators) and 4 females (44% of the female perpetrators) were both victims and perpetrators. Drugs and/or alcohol were frequently used by both the assailant (58% of the males and 55% of the females) and the victim (56% for the males and 66% for the females). The attributes of adolescent boys at highest risk of becoming perpetrators were all related to violence, including exposure to parents who were violent in the home (adjusted odds ratio, 2.68), being a victim of physical or sexual assault (adjusted odds ratio, 2.83), having parents who encouraged gang membership (adjusted odds ratio, 3.58), and knowing a perpetrator of sexual violence (adjusted odds ratio, 3.83).

Conclusions  Adolescent boys and girls were both victims and perpetrators of sexual assault. Programs that identify and target violence in young children's lives, particularly intrafamilial violence, may reduce the risk of teenagers becoming perpetrators of sexual violence. Anticipatory guidance during teen years, regarding risk and avoidance of sexual exploitation, may be helpful in reducing the chances of sexual assaults.

DETAINED YOUTH may have many negative life experiences linked to their delinquent and maladaptive behaviors.1-7 Physical and sexual abuse have been reported in juvenile murderers,1,2 chemically dependent youths,8 and youths involved in illegal behavior,3 including running away, prostitution, and drug abuse.4,8 It is clear from this that both physical and sexual maltreatment may contribute to delinquency behaviors including sexual and other forms of violence against others, thus turning victims into victimizers. Most crimes are not reported9-12; therefore, it is reasonable to assume that some incarcerated teens have committed additional crimes, such as sexual assaults, beyond those for which they were incarcerated.

We wished to explore self-reported perpetration of sexual assault by incarcerated male and female teens, focusing on potential precursors of this violent behavior. We chose the following areas to explore based on previous work relating violence and delinquency: family discord and/or stress,13-15 childhood maltreatment,16 sexual abuse,7,17 gang involvement,18 and drug and/or alcohol use19,20 based on significant numbers of positive drug screenings at the time of arrest.21

Participants and methods

The National Gang Survey is designed and implemented yearly by one of us (G.W.K.), 27 collaborators from various US institutions (eg, criminology, health care, special education, and criminal justice practitioners), and Chicago State University's National Gang Crime Research Center, Chicago, Ill. The National Gang Survey focuses on a different area of gang activity each year. The survey in 1996, titled GANGFACT, involved 10 000 juvenile and adult inmates in 75 institutions from 17 states. It posed 99 questions regarding bullying behavior, family dynamics, risks associated with joining gangs, gang society, and gang activity within correctional facilities.22

A representative of Chicago State University, who tells them they may ignore any or all questions, administered the anonymous paper-and-pencil survey simultaneously to detainees gathered in classrooms or dining halls after delivering a brief introductory explanation. Almost all residents of the facility took part, except those in court or sickbay. No attempt was made to determine how many youths were absent. The survey identifies those incarcerated youth with gang membership.22

Two of the 3 juvenile detention facilities in a large southwest city participated simultaneously in the survey in 1995 and 1996. The third hall did not participate because only 2 representatives of Chicago State University were available each year. The 3 facilities house primarily preadjudicated youth with a few who were sentenced and awaiting transfer. The average length of stay in the halls is 6 weeks. Youth are transferred daily back and forth among the 3 halls to balance the institutions' populations. Thus, the populations are almost identical and the tests must be administered simultaneously in both halls to avoid duplication of questionnaires.

A group of physicians and health educators from these 2 juvenile halls were invited by the gang surveyors to develop questions regarding juvenile detainees' experiences of sexual abuse and/or assault. These questions were added to the end of the survey. Data derived from these added questions were analyzed only for the 2 institutions, not nationally.

The questions were constructed to assess the "who," "what," and "where" of the youth's experiences of being a victim and of perpetrating sexual assault. Independent variables were constructed from the primary questionnaire that related to sexual assault victimization, family and social dynamics, gang membership, and the questions we added related to drug and/or alcohol use.21 Most of the questions in the primary questionnaire were dichotomous (ie, yes-no or true-false) with the exception of age and race questions, and questions such as "It is all right to demand my needs be met" where a Lickertlike scale was used (from always [5] to never [1]). In our analysis, youth choosing "always" were compared with all other answers less than always. We constructed dichotomous questions for "did an event happen," and multiple-choice questions for identities of victims and locations of assaults. In each question, sexual assault was defined as "sex someone did not want."

In 1995, after pilot testing by 20 incarcerated teens to determine readability and comprehension, the questions were administered to 281 juveniles from the same 2 juvenile detention centers as part of that year's National Gang Survey. Using the results of that survey, some questions were rewritten and additional questions were created regarding the identity of victims. This article presents only the 1996 data.

The chief juvenile court judge and the probation department administering the institutions approved the survey. The institutional review board of the National Gang Crime Research Center also approved the study.

Comparisons of 2 variables were done using the χ2 test and, where applicable, Fisher exact test. The attributes of adolescent boys who were perpetrators of sexual assault (79 of 707) were analyzed further using logistic regression. A pool of 14 potential predictor variables was selected based on the professional experience of the study designers and the previously quoted literature.1-7,13-18 The variables were dichotomous and included father only household; prior victim of sexual assault; started a fight in a facility; it is "always all right to demand that my needs be met"; knew another male perpetrator of sexual assault; "my parents" attitudes toward drugs, crime, or violence was favorable; parents had spent time in prison; "my parents were physically violent in my home"; juvenile gang membership; parents knew about gang membership; parents encouraged gang membership; parents would be embarrassed by gang membership; and 2 variables regarding ethnicity (African American vs other or Hispanic vs other). Logistic regression using both forward and backward selection (P = .05 for inclusion) was used to identify predictor variables for the final multivariate model. All predictors appearing in the final model of either of these methods was retained for further analysis. If 2 variables were so closely correlated that they each eliminated the other reciprocally during forward and backward elimination, they would have been combined into a composite variable. In this study, that effect was not seen. All values are reported as adjusted odds ratios (ORs).


A total of 805 adolescents (707 boys, 91 girls, and 7 gender-not-designated) participated in the survey study. Eighty-six percent were between 15 and 18 years old (overall mean, 16.2 years; 16.2 years for males and 16.1 years for females) (n = 692). The ethnicity of the youths reflected the general population of the halls, 33% black, 48% Hispanic, 8% white, 5% Asian, 1% Native American, and 5% mixed race. Sexual abuse and/or assault was reported by 147 teens (4 did not answer the question) (94 males [13% of the males], 48 females [52% of the females], and 5 gender-not-designated [71% of the GND]).

Adolescent girls as perpetrators

The adolescent girls made up 10% of the sample that is equal to their percentage in the halls. Nine (10% of the females) reported sexually assaulting at least 1 person, while 17 (19%) declined to answer the question. Four (44%) of 9 perpetrating females reported previous sexual victimization that was not significantly different than the rate of victimization among nonperpetrating females (44/82 or 54%).

Their victims were extrafamilial and known to them (Table 1). The assaults took place in homes (their own, relatives', or boyfriends'). Four (44%) used verbal coercion, 1 (11%) physical force, and 1 (11%) both verbal and physical force, to accomplish the assault.

Table 1. 
Relationship of Study Perpetrators to Their Victims of Forced Sexual Activity*
Relationship of Study Perpetrators to Their Victims of Forced Sexual Activity*

During the perpetration of sexual assault, 5 females (55%) always used drugs and/or alcohol, while 1 female, with more than 1 assault, sometimes used drugs and/or alcohol. Five (55%) reported their victims were using drugs and/or alcohol at the time of the assault and 1 (11%) said some of her victims were using drugs and/or alcohol.

Adolescent boys as perpetrators

Seventy-nine adolescent boys (11% of the males) sexually assaulted at least 1 person. One hundred forty-three declined to answer the question. Of the 94 male victims of sexual abuse and/or assault in this study, 28 (30%) became perpetrators compared with only 51 (8%) of 613 of the nonvictims becoming perpetrators (P<.001).

Adolescent boys almost always knew their victims, which included family and nonfamily persons (Table 1). Similar to adolescent girls, they assaulted in homes (their own, relatives', or girlfriends'). The remaining places were school (4), car (3), party (5), outside (5), with 9 in a variety of other locations. Thirty males (40%) used verbal coercion, 10 males (13%) used physical force, and 13 (16%) used both to accomplish the assault. When engaged in sexual assault, 28 males (35%) said they were always under the influence of drugs and/or alcohol and 18 (23%) were sometimes under the influence of drugs and/or alcohol. As for their victims, males reported 26 (33%) were using drugs and/or alcohol while 18 respondents (23%) said some of their victims were under the influence of drugs and/or alcohol.

Logistic regression using the 2 selection methods yielded the same model with 4 predictors (Table 2); living with violence at home (OR, 2.68), being a victim of sexual assault (OR, 2.83), parents encouraging violence (ie, gang membership) (OR, 3.58) as well as knowing a "role model" for sexual assault (OR, 3.83) independently increased the risk of engaging in sexual assaults. The Hosmer-Lemeshow goodness-of-fit statistic (P = .35) indicated that the model fit.23

Table 2. 
Logistic Regression Model of Variables Associated With Being a Male Perpetrator of Sexual Assault*
Logistic Regression Model of Variables Associated With Being a Male Perpetrator of Sexual Assault*


Adolescent girls as perpetrators

While females are often studied as victims of sexual assault, less is known about them as possible perpetrators. Monteleone et al24 reported that 85% of perpetrators are males but states that little is known about females and it is likely their offenses are underreported. The Minnesota Student Survey Report of 71 594 9th and 12th graders by Borowsky et al25 found 1.3% of females admitting to perpetrating sexual abuse and/or assault. Our delinquent population's rate of victimization (52%) and rate of perpetrating were higher (10%) than the general population, while the risk of becoming a perpetrator was not higher for our victims (8%) vs nonvictims (11%). However, we had very few female perpetrators and, thus, may have lacked the power to show an effect of sexual victimization. The review by Lodico et al7 of the Minnesota data had more victims becoming perpetrators (3.5%) compared with 1.3% of the nonvictims. This area deserves more study. Although we thought some female gang members might have been involved in sexual assaults as part of initiation rites, we found only 1 adolescent girl assaulted a gang member.

Adolescent boys as perpetrators

An equal number of males were victims (13%) and perpetrators (11%). We found that 30% of the male victims became perpetrators, a figure nearly identical to the rate of 27.6% reported by Lodico et al7 although their odds risk ratio was somewhat higher (4.4 vs 2.83 [ours]). This may be because our youth had many additional risk factors that affected their propensity to force sexual activity.

Unlike females, some males were not limited by the boundaries of the family and abused family members. In fact, all of the male victims in this study were family members. The reasons for this are unclear and deserve more study.

Girlfriends made up the largest group of victims targeted by our male perpetrators. Bennett and Fineran26 in 1998 also reported high school acquaintances or dating and/or ex-dating partners as most likely to engage in sexual or physical violence. Rickert and Wiemann's review of date rape found a prevalence of 20% to 68% for adolescents,27 which is much higher than our total figures for all sexual assaults. Risks for dating violence include parent-child violence, drug use, knowing a perpetrator of date violence, and especially for males believing that violence is justified to solve problems,28 which are similar to our findings. The Minnesota survey found similar results with the addition of frequent drug and/or alcohol use and suicidal behavior.25 Thus, violent experiences and psychoactive drug abuse remain the core elements leading to sexual assault. When parents held neutral attitudes regarding gang involvement, that is, they knew of the affiliation but did not actively support membership, the children had less risk of becoming an assailant. Further study of teens' beliefs regarding their parents' attitudes toward unlawful behavior would help complete our understanding of the parents' role in encouraging delinquency, especially violence.

We did not ask our respondents if they felt violence was acceptable in a relationship or whether male victims of forced sexual activity felt justified in forcing others. We did find that both males and females forced their partners to have sex. The area of date and/or acquaintance rape deserves more study, especially for delinquent youths. Adolescent care providers (ie, pediatricians, nurse practitioners, and health care workers in adolescent care sites) should provide anticipatory guidance to their patients about identifying risks of sexual assault and methods to avoid that risk, especially in romantic relationships.

Drug and/or alcohol use

We found high rates of drug and/or alcohol use in the episodes of sexual assault as have others,29 but not all assaults were accompanied by substance use. Rickert and Weimann27 cited alcohol or illicit drug or alcohol use as contributing to date rape and Johnson and Knight30 suggested that alcohol might play more of a role in coercive sexual acts than previously hypothesized. Since drug and/or alcohol use is so prevalent in delinquent youth,31 attention to this is imperative during rehabilitation for delinquency.

Limitations of the study

Although the questionnaire was anonymous, the emotional sensitivity of sexual subject matter may have influenced some answers. Placing these questions at the end of a long questionnaire was meant to reduce some of the emotional reaction to the sex abuse questions but the usefulness of this technique is unknown. The survey reports the results from a large number of youth that may help to reduce the effect of overreporting and underreporting of the activities. Nevertheless, self-reports are subject to bias and far fewer females participated. Generally youth self-report higher levels of criminal activity but not all investigators agree.32

Although face-to-face interviewing may result in more accurate responses, it is expensive and mandatory reporting laws make conducting nonanonymous sexual abuse surveys impossible. In addition, protecting the confidentiality of prisoners' answers remains an important concern. Anonymous written surveying remains one of the few ways of polling large numbers of detained youth. The results may not be generalizable to other cities owing to this study being performed only in one large urban city. We did not ask about circumstances leading to the assaults or the respondents' feeling regarding past and future possible assaults. We recommend that this be investigated further.

About 20% of youths (n = 160) did not answer the question regarding whether they were sexual assailants. A comparison of refusals to answer other questions revealed that demographic questions had low refusal rates, knowing other sexual perpetrators had a 20% refusal rate, and questions regarding gang membership and parental attitudes had higher refusal rates of 30% and 36%, respectively. After youths admitted to perpetrating sexual assaults, then only 12% to 20% of them refused to provide details except for their own age and that of victims, which approached a 35% refusal rate. It seems possible that some incarcerated youths were withholding information they felt could be used against themselves or their parents. It is reasonable to assume many or most of the 160 youths who refused to discuss the sexual assault questions were in fact assailants, but this remains conjecture.


There are no systematic studies of large programs designed to reduce the incidence of sexual abuse of children or their subsequent risk of becoming adolescent sexual assaulters. However, a few interventions have reduced child abuse generally. Between 1978 and 1980, Olds et al33-35 reported the implementation of home visits by nurses for pregnant high-risk poor white rural mothers in New York State continuing for 2 years post partum. After 15 years, the risk of child abuse, neglect, and subsequent delinquency was reduced by 75% compared with a control group.34,35 This economical home-visitation cost only $1582 per family.33 Kitzman et al36 provided similar services to African American women and found similar outcomes, especially fewer personal beliefs about childrearing that could lead to abuse. The program seemed to benefit those mothers who most needed help and might be at risk for raising delinquent children.

The National Longitudinal Study of Adolescent Health (ADD HEALTH), a longitudinal study of adolescents in grades 7 through 12, compared their living social contexts to health risk behaviors. Parent-family connectedness protected them from most health risk behaviors including violence. This leads to the conclusion that the family context has a strong effect on adolescents' behavior.37 Despite this, certain individual factors played the most significant role in a teen's involvement in violence such as a history of victimization and/or witnessing violence, weapon carrying, deviant behavior (ie, destruction of property, theft, skipping school in the past year, and ever being suspended or expelled from school), and drug selling. In actuality, except for the first, these are measures of violent and/or risk behavior and would be expected to be common among teens engaging in violence. This returns us to the family and its important role in early childhood violence. Pediatricians can provide a valuable service by screening for family violence and referring to appropriate services. Of course, family support services must be available.


These data highlight the fact that, in addition to physical violence, youthful offenders may also be engaging in sexual violence. Our data and that of others show an association between exposure to violence (ie, witnessed, condoned in the home, and/or being a victim) and an increased risk of becoming a perpetrator of sexual violence. Early intervention in troubled families appears to be a promising intervention strategy to prevent later teenage violence including sexual assaults. Anticipatory guidance for teens should include discussion of sexual assault. For youths already in the "system," rehabilitation of antisocial behavior should be broad and include sexual victimization.

Accepted for publication April 5, 2002.

What This Study Adds

Although it is commonly believed and self-reports of offenders support the premise that delinquents may commit a variety of crimes, there is little literature about the sexual offenses committed by juveniles incarcerated for other, unrelated crimes. We asked questions regarding sexual offenses in a longer questionnaire about violence and gang activity.

Data show that both adolescent boys and girls are involved in sexual exploitation. Exposure to and infliction of violence on youths along with parental approval of antisocial behavior are related to sexual victimization on the part of males. The study also reinforces the understanding that alcohol and other drugs play a role in these behaviors.

Reprints: Robert E. Morris, MD, 12-358 Marion Davies Children's Center, Department of Pediatrics, UCLA Center for Health Sciences, 10833 LeConte Ave, Los Angeles, CA 90095-1752 (e-mail:

Lewis  DOMoy  EJackson  LD  et al.  Biopsychosocial characteristics of children who later murder: a prospective study.  Am J Psychiatry. 1985;1421161- 1167Google Scholar
Lewis  DOPincus  JHBard  B  et al.  Neuropsychiatric, psychoeducational, and family characteristics of 14 juveniles condemned to death in the United States.  Am J Psychiatry. 1988;145584- 589Google Scholar
Burgess  AWHartman  CRMcCormack  A Abused to abuser: antecedents of socially deviant behaviors.  Am J Psychiatry. 0987;1441431- 1436Google Scholar
Famularo  RKinscherff  RFenton  TBolduc  SM Child maltreatment histories among runaway and delinquent children.  Clin Pediatr (Phila). 1990;29713- 718Google ScholarCrossref
Freeman-Longo  RE The impact of sexual victimization on males.  Child Abuse Negl. 1986;10411- 414Google ScholarCrossref
Widom  CS Victims of Childhood Sexual Abuse–Later Criminal Consequences.  Washington, DC US Dept of Justice, Office of Justice Programs, National Institute of Justice March1995;Research in Progress Seminar Series. Publication NJC151525
Lodico  MAGruber  EDiClemente  RJ Childhood sexual abuse and coercive sex among school-based adolescents in a midwestern state.  J Adolesc Health. 1996;18211- 217Google ScholarCrossref
Cavaiola  AASchiff  M Behavioral sequelae of physical and/or sexual abuse in adolescents.  Child Abuse Negl. 1988;12181- 188Google ScholarCrossref
Elliott  DSHuizinga  DAgeton  SS Explaining Delinquency and Drug Use.  Beverly Hills, Calif Sage Publications1985;
Empey  LT American Delinquency: Its Meaning and Construction.  Homewood, Ill Dorsey Press1982;
Williams  JRGold  M From delinquent behavior to official delinquency.  Soc Probl. 1972;29209- 229Google ScholarCrossref
Wolfgang  MThornberry  TFiglio  R From Boy to Man, From Delinquency to Crime.  Chicago, Ill University of Chicago Press1987;
Hanson  CLHenggeler  SWHaefele  WFRodick  JD Demographic, individual, and family relationship correlates of serious and repeated crime among adolescents and their siblings.  J Consult Clin Psychol. 1984;52528- 538Google ScholarCrossref
Loeber  RDishion  T Early predictors of male delinquency: a review.  Psychol Bull. 1983;9468- 99Google ScholarCrossref
Tolan  P Socioeconomic, family, and social stress correlates of adolescent antisocial and delinquent behavior.  J Abnorm Child Psychol. 1988;16317- 331Google ScholarCrossref
Paperny  DMDeisher  RW Maltreatment of adolescents: the relationship to a predisposition toward violent behavior and delinquency.  Adolescence. 1983;18499- 506Google Scholar
Truscott  D Adolescent offenders: comparison for sexual, violence and property offences.  Psychol Rep. 1993;73657- 658Google ScholarCrossref
Greydanus  DEFarrell  EGSladkin  KRypma  CV The gang phenomenon and the American teenager.  Adolesc Med. 1990;155- 70Google Scholar
Dembo  RDertke  MBorders  SWashburn  MSchmeidler  J The relationship between physical and sexual abuse and tobacco, alcohol, and illicit drug use among youths in a juvenile detention center.  Int J Addict. 1988;23351- 378Google Scholar
Simonds  JFKashani  J Specific drug use and violence in delinquent boys.  Am J Drug Alcohol Abuse. 1980;7305- 322Google ScholarCrossref
US Department of Justice, Table 4 .33 and Table 4 .35.  Bureau of Justice Statistics: Sourcebook of Criminal Justice Statistics–1992. Washington, DC US Dept of Justice, Office of Justice Programs1993;460463Google Scholar
Stone  Sed Contemporary Gang Issues: An Inside View.  Peotone, Ill New Chicago School Press Inc2000;
Hosmer  DWLemeshow  S Applied Logistic Regression.  New York, NY John Wiley & Sons1989;
Monteleone  JAGlaze  SBly  KMMonteleone  JAed Sexual abuse: an overview.  Child Maltreatment: A Clinical Guide and Reference. Vol 12nd ed St Louis, Mo GW Medical Publishing Inc1998;133Google Scholar
Borowsky  IWHogan  MIreland  M Adolescent sexual aggression: risk and protective factors.  Pediatrics [serial online]. 1997;100 ((6)) E7Available athttp://www.pediatrics.orgAccessed May 24, 2002Google Scholar
Bennett  LFineran  S Sexual and severe physical violence among high school students: power beliefs, gender, and relationship.  Am J Orthopsychiatry. 1998;68645- 652Google ScholarCrossref
Rickert  VIWiemann  CM Date rape among adolescents and young adults.  J Pediatr Adolesc Gynecol. 1998;11167- 175Google ScholarCrossref
Tontodonato  PCrew  BK Dating violence, social learning theory and gender: a multivariate analysis.  Violence Vict. 1992;73- 14Google Scholar
Seifert  SA Substance use and sexual assault.  Subst Use Misuse. 1999;34935- 945Google ScholarCrossref
Johnson  GMKnight  RA Developmental antecedents of sexual coercion in juvenile sexual offenders.  Sex Abuse. 2000;12165- 178Google Scholar
Morris  REHarrison  EAKnox  GWTromanhauser  EMarquis  DKWatts  LLMcCord  JedWidom  CSedCrowell  NAed Health risk behavioral survey from 39 juvenile correctional facilities in the United States.  J Adolesc Health. 1995;17334- 344Google ScholarCrossref
National Research Council and Institute of Medicine, Panel on Juvenile Crime: Prevention, Treatment and Control, Committee on Law and Justice and Board on Children, Youth, and Families, Not Available  Juvenile Crime, Juvenile Justice Washington, DC National Academy Press2001;31Google Scholar
Olds  DLHenderson Jr  CRPhelps  CKitzman  HHanks  C Effect of prenatal and infancy nurse home visitation on government spending.  Med Care. 1993;31155- 174Google ScholarCrossref
Olds  DLEckenrode  JHenderson Jr  CR  et al.  Long-term effects of home visitation on maternal life course and child abuse and neglect. fifteen-year follow-up of a randomized trial.  JAMA. 1997;278637- 643Google ScholarCrossref
Olds  DLEckenrode  JHenderson Jr  CR  et al.  Long-term effects of home visitation on children's criminal and antisocial behavior: 15-year follow-up of a randomized trial.  JAMA. 1998;2801238- 1273Google ScholarCrossref
Kitzman  HOlds  DLHenderson Jr  CR  et al.  Effect of prenatal and infancy home visitation by nurses on pregnancy, outcomes, childhood injuries, and repeated childbearing: a randomized controlled trial.  JAMA. 1997;278644- 652Google ScholarCrossref
Resnick  MDBearman  PSBlum  RW  et al.  Protecting adolescents from harm: findings from the National Longitudinal Study on Adolescent Health.  JAMA. 1997;278823- 832Google ScholarCrossref