Background. Violence is a large contributor to morbidity and mortality among adolescents. Most studies assessing markers for violent injury are cross-sectional. To guide intervention, we conducted a case-control study to explore factors associated with assault injury and locations to reach at-risk adolescents.
Objective. 1) To assess risk and protective factors for adolescent assault injury compared with 2 control groups of youth with unintentional injuries and noninjury complaints presenting to the emergency department and 2) to assess locations of contact with assault-injured youth for prevention programs.
Methods. Face-to-face and phone interviews were conducted with systematic samples of youth aged 12 to 19 years presenting to the emergency department with assault injury, unintentional injury, and noninjury complaints. Youth with intentional injuries were matched to youth in the 2 control groups on age ±1 year, gender, race, and residency.
Results. One hundred forty-seven 147 assault-injured youth completed interviews. One hundred thirty-three assault-injured youth were matched to 133 unintentionally injured and 133 noninjured youth presenting to the emergency department. Compared with the 2 control groups, assault-injured youth were more likely to have had more fights in the past year (odds ratio [OR]: 3.91; 95% confidence interval [CI]: 2.02, 7.58; OR: 4.00; 95% CI: 2.23, 7.18) and fights requiring medical treatment (OR: 35.49; 95% CI: 8.71, 144.68; OR: 80.00; 95% CI: 11.13, 574.80). Eighty percent of assault-injured youth had been in 1 or more fights in the last 12 months compared with 55% and 46% in unintentional and noninjured controls, respectively. Assault-injured youth were more likely to have had previous weapon injuries (OR: 9.50; 95% CI: 3.39, 26.6; OR: 8.50; 95% CI: 3.02, 23.95) and have seen someone shot (OR: 2.00; 95% CI 1.12, 3.58; OR: 2.00; 95% CI: 1.12, 3.58). Eighty-six percent of assault-injured youth had a regular health care provider with 82% reporting a visit within the last year. There were no differences between cases and controls with regard to physician contact, extracurricular activity involvement, school or church attendance, police contact, weapon access or weapon-carrying, or witnessing nonweapon-related violence.
Conclusions. Fighting was common among all groups. Assault-injured youth were more likely to have had previous weapon injuries and were high-risk for future injury. Past fights, past fight injuries, and seeing someone else shot were markers associated with assault injury. Health providers do have access to at-risk teens for clinical risk assessment and intervention.
Improved public health and medical care have markedly reduced the impact of common childhood infections and other medical conditions. However, injuries and violence have come to occupy an increasingly prominent role in the mortality and morbidity of children and adolescents. This is especially true in the United States where the leading causes of death in youth aged 10 to 19 include unintentional injury, suicide, and homicide.1,2 Although the number of youth dying from violent injuries is alarming, it represents only a fraction of violent injuries that occur in this age group. For every death attributable to violent injury in youth in the District of Columbia, there were 8 hospitalizations and 108 emergency department (ED) visits.3 To reduce morbidity and mortality attributable to violence, we must first study risk and protective factors for adolescent assault injury and determine potential sites for intervention.
Much is known about general risk factors for adolescent injury and delinquency, but less is known about injury among youth presenting to care for intentional injuries. High rates of injury are associated with behavior patterns and risk factors common to adolescent development. These include male gender, previous injuries, alcohol/drug use, conflict with parents, pattern of parental supervision, weapon-carrying, delinquency, and pubertal development.4–16 Co-varying risk factors for violent behavior are similar: male gender, poor mental health, drug use, lack of parental affection and support, weapon-carrying, school drop-out, exposure to violence, victimization, and delinquency.17–21 Careful attention must be given to the context in which injuries occur, with focus on interventions targeted at factors that may be modifiable.
Known risk/protective factors for involvement in violence and injury may be viewed as associations that identify at-risk individuals or as contributing mediators or co-factors to the outcome.22 Regardless of the degree of causality, ascertainment of epidemiologically-associated factors can identify a subgroup at risk for violence and its adverse consequences. Many advocate a proactive, preventive approach that identifies risk and escalating problem behavior early for interruption and intervention. The literature and the 2001 Surgeon General’s Report on Youth Violence strongly suggest that a multitude of factors contribute to an individual’s propensity to behave aggressively.23,24 An acknowledgment of this multifactorial reality forms the conceptual basis for our assessment of multiple risk and protective factors. Unfortunately, many studies assessing markers for violent injury are cross-sectional without comparison groups. Identifying these predictive factors is an important step in approaching prevention.
The American Academy of Pediatrics25,26and other organizations have advocated that health professionals be involved in the identification of youth at risk for perpetration or victimization of violence. It has been proposed that primary care providers address violence prevention in child health supervision. It is not known, however, whether assault-injured youth access primary care and what locations of contact might be appropriate for prevention programs. To guide strategies, we conducted a case-control study with 2 objectives: 1) to explore risk and protective factors associated with assault injury, and 2) to identify locations or populations of at-risk adolescents for intervention.
Interviews were conducted with samples of adolescents aged 12 to 19 presenting to 2 urban hospitals. Cases included adolescents with interpersonal assault injuries excluding sexual assault, child abuse, or legal intervention (International Classification of Diseases, Ninth Revision, Clinical Modification supplementary classification of external causes of injury and poison codes [E-codes] E960, 961–966, 968 and 969). Two control groups included matched adolescents presenting with 1) unintentional injury, and 2) noninjury complaints (medical illness). These 2 groups were felt to represent the population at risk for assault injury and those who sought care at the same sites. We recruited cases and control presenting to the ED or hospitalized during a period from January 1997–January 2000 at a large urban, academic children’s hospital and from December 1997–December 1998 at a large urban private hospital in Washington, DC. The study protocol was approved by the institutional review boards at the 2 institutions. Eligibility criteria are listed in Table 1.
We identified all eligible youth who presented to the ED or were hospitalized for assault injury. Cases were identified from ED logs and computer printouts of hospitalized patients. Research assistants performed interviews in the ED, hospital ward or by phone. Systematic assignment of ED shifts covered all 7 days of the week during the hours of 8 am to 12 midnight. Surveillance of adolescent injuries found that >90% of adolescents with injuries presented during these hours. Because we wished to oversample more serious injuries, we attempted to interview all assault-injured adolescents hospitalized for injuries. Patients were contacted by a study staff member before discharge whenever possible. When in-person contact was not made, families were contacted by phone.
Control recruitment was designed to enroll adolescents who would have come to the same facility if they had an assault injury. Cases were matched to youth in the 2 control groups on age ±1 year, gender, race, and residency (District of Columbia or Maryland). The unintentional injury control group included patients with discharge diagnoses of unintentional injuries. The uninjured control group included adolescents with nontraumatic conditions (eg, fever, sore throat, appendicitis, first asthma episode). We excluded patients with chronic medical conditions or sexual complaints because we wished to assess primary care use and sexual risk behavior. For ED patients, cases and controls were enrolled on an ongoing basis and were matched with controls at the time of analysis. All hospitalized patients were recruited. After enrollment of a hospitalized case, the next appropriately matched control was recruited. We attempted to match all hospitalized cases 1-to-1 with hospitalized controls.
Adolescents were approached by a research assistant who screened their eligibility for the study. Written informed consent was obtained. If the adolescent was a minor (<18 years), parental consent (in person or by phone) and adolescent assent were obtained. If the adolescent was of legal age, their written consent was obtained.
The interview consisted of 2 components: a verbal response component and an audiotape (Walkman, Sony Corporation, Japan) component. For the audiotape component, subjects listened to questions asked on the Walkman and wrote numeric answers on an answer sheet that did not have the printed questions. For phone interviews, we sought to complete the interview as quickly as possible after the injury event to minimize the likelihood of recall bias. For the audiotape component, questions were asked over the phone and participants answered by pressing the appropriate button on the touch tone keypad that was interpreted by the interviewer using Digit Grabber dialed digit meter (model TPM-32; Metro Tel Corporation; Jericho, NY). Adolescents were given an audiotape player (Walkman) in appreciation of their participation. Quality control was maintained by extensive training and close supervision of research assistants by senior investigators.
Development of the interview included 16 initial focus groups with adolescents on injury and violence,27 review of the literature on adolescent injury, and adaptation of models on adolescent risk-taking behavior.8,28,29 Risk and protective factors assessed included: 1) social and community factors (including individual and family demographics, school and church attendance, primary health care contact, and extracurricular activity involvement); 2) behavioral factors (including substance use, sexual activity, behavioral and emotional problems, and suicidal ideation); and 3) violence factors (including past violent injury, fighting frequency, weapon-carrying, weapon access, police contact, and exposure to violence).
National Health Interview Survey questions on demographics, primary care source, past injuries, and use of medical and mental health services were included. The questions on extracurricular activity involvement and police contact were adapted from the Denver Youth Survey and Boston City Hospital Interview.30,31 We used questions on fighting, past injury, weapon-carrying, and behavioral factors from the middle school version of the Youth Risk Behavior Survey, which has been found to have good test-retest reliability.32 A measure of exposure to violence developed by Richters and Martinez was modified for this study.33,34 The interview instrument underwent 2 phases of pretesting and 2 months of pilot testing.
Data were entered into Microsoft Access (Microsoft Corporation, Bellevue, WA) and analyzed using SAS (SAS Institute, Cary, NC).
Assessment of Sampling Bias
Cases were compared with unmatched cases and refusers on age, sex and race/ethnicity. Comparisons were assessed using analysis of variance for continuous variables and χ2 for categorical variables. The Kruskal-Wallis rank order test was used for variables that were not normally distributed.
In an effort to increase the study efficiency, individual cases were 1-to-1 matched with 1) a youth in the unintentional injury control group and 2) a youth in the noninjured control group. Matching variables included those demographic factors that may be associated with the outcome variable and thus confound the analysis of risk factors. These included gender, race, age, and state of residency. Risk and protective factors included the social and community, behavioral, and violence factors described above. The association between each risk or protective factor and intentional injury outcome was measured using odds ratios (ORs) with 95% confidence intervals (CIs). Where necessary, interview item results were dichotomized to allow calculation of the OR. Two ORs were calculated for each variable using the cases as a referent group compared with each of the 2 control groups, respectively. These paired comparisons take advantage of the 1-to-1 matching between groups and compared each patient to his or her matched control subject. The database was maintained using Microsoft Access, and the SAS logistic regression program was used to obtain ORs and CIs. Pearson χ2 analysis was used to evaluate with whom the youth predominately lived.
Figure 1 shows the results of case recruitment. Of all eligible assault-injured patients, 80% completed interviews and 73% were successfully matched to a control in each of the 2 groups. Of those completing interviews, 64% were unarmed assaults, 13% were gun assaults, 10% were knife assaults, and 13% were assaults with other weapons. On signed consent, the adolescent was defined as enrolled. Losses of eligible patients from enrollment were attributable to inability to contact the patient, inability to obtain consent, or refusal to participate. Refusals were attributable to lack of interest, lack of time, or feeling too ill to participate. Table 2 compares enrolled and matched cases with those not enrolled, those who refused, or those who were missed. No significant differences were found across the enrollment categories on age, gender, or race.
In the unintentional and uninjured control groups, 85% (345/404) and 75% (265/353), respectively, of eligible patients completed interviews. For ED patients, control groups were oversampled, and in analysis, the first matched case was used. For those hospitalized, we attempted to match 1-to-1 with controls during recruitment. Of all cases, 26% (34/133) were hospitalized patients. All were matched to hospitalized controls except for 5 cases failing to match to a noninjured hospitalized control and 3 cases failing to match to an unintentionally injured control. These cases were matched to ED controls. Of all case and control interviews completed, 74% were face-to-face interviews completed in the hospital with the remainder completed by phone.
Age, race, gender, and residency were matching criteria. Case and control participants had a mean age of 15 years. Ninety-five percent were black, 69% were male, and 77% lived in the District of Columbia. Of cases, 37% (49/133) were injured by a weapon (E-codes 965, 966, and 968.2). Cases were more likely than controls in the 2 groups to live in a group home and were much more likely to live without parental presence (P = .0046) as seen in Table 3.
Table 4 presents frequencies and ORs of risk and protective factors for assault injury by study group. We queried church attendance and involvement in 9 types of prosocial extracurricular activity. There were no differences between cases and the 2 control groups on church attendance or the number or type of extracurricular activity involvement. More than 86% of cases and controls reported having a regular health provider and >81% reported having had a checkup in the last year. There were no differences between cases and the 2 control groups on having a regular health care provider or having had a check-up.
Table 4 compares the responses of case participants with the 2 control groups for selected behavioral and violence risk factors. Behavioral risk factors included smoking cigarettes or marijuana, alcohol use, drug use including cocaine and injected drugs, sexual activity, suicide, history of behavioral and emotional problems, seat belt use, and bicycle helmet use. Cases were significantly more likely than controls in 1 of the control groups to have ever tried cigarettes or marijuana, drank alcohol in the last month, ever had sex, or ever been or gotten someone pregnant. Cases were more likely than controls in both groups to have >1 sexual partner.
For violence risk factors, assault injury was associated with more lifetime fights, fights in the past year, fight injuries, been shot or injured with a weapon compared with both control groups. For fights in the past year >1, sample size of 133 per group had 80% power to detect a 15% difference between groups (α = 0.10; 2-tailed). Cases were more likely to have seen someone shot or attacked with a knife. There were no significant differences between cases and controls on weapon-carrying, weapon access, gang involvement (≤3%), police involvement, history of being threatened, violent injury or death of family or friends, or witnessing someone threatened, mugged, or sexually assaulted.
This case-control study assessed risk and protective factors for adolescent assault injury and potential locations of contact to reach at-risk adolescents. We found the prevalence of physical fighting was high in all groups. However, living apart from parents, past fights, past fight injuries, and weapon injuries may be markers for assault injury. We also found that primary care health providers have access to at-risk teens for clinical risk assessment and intervention. Locations for intervention with at-risk youth could include primary care providers, the ED, and group homes.
Potential limitations include validity of self-report data and 2 modes of interviewing, bias related to convenience sampling, nonparticipation, exclusion criteria, and control group selection issues. The validity of self-report data is a persistent concern in behavioral research. There was face validity and internal consistency in participant responses in extensive pretesting and study interviewing. Research on the use of the Walkman audiotape and DigitGrabber interviewing has suggested greater confidentiality and validity of these modes of interviewing.35,36 Others have found that the more efficient telephone interviewing is as accurate as face-to-face interviews.37,38 It is possible that there existed bias in convenience sampling and nonparticipation; however, our refusal rate was low and there were few differences noted between those who did or did not enroll as shown in Table 2.
Uninjured adolescents with sexual complaints were excluded because of our interest in exploring the association of assault injury and risk behavior including sexual risk behavior. Exclusion of these patients in the control group may have resulted in a lower risk uninjured control group. Nonetheless, there were surprisingly few factors that differentiated the case and control groups. This exclusion may have overestimated the specific association between assault injury and sexual risk behavior that did reach statistical significance (ever had sex OR: 2.00; 95% CI: 1.08, 3.72; lifetime sexual partners OR: 2.13; 95% CI: 1.15, 3.94). However, there were statistically significant differences in sexual behavior in the assault-injured group compared with the unintentional control group as well (pregnancy: OR: 2.71; 95% CI: 1.14, 6.46; lifetime sexual partners: OR: 1.93; 95% CI: 1.01, 3.68).
Other limitations include the fact that the 2 chosen control groups may not have been appropriate for assessment of risk and protective factors and locations of contact with youth. It has been postulated that injured youth with intentional and unintentional injuries share behavioral risk and protective factors that lead to injury. If this is true, we may have overmatched. This study was able to test this hypothesis of shared risk and protective factors and confirms similar profiles on many factors with few exceptions.
Our second control group, uninjured patients, may not have been representative of the community, but representative of the community that uses EDs for primary care. It has been reported that adolescents who utilize EDs for acute illness may be a high-risk group without a regular source of health care or without insurance.39,40 Though it is possible that they are a high-risk group, in our study the noninjured group were just as likely as the injured group to have a regular health care provider and have had a regular checkup in the last year.
Finally, the study was conducted in a high-risk community,3 and the results may not be generalizable to other regions.
One objective of this study was to assess locations for intervention with assault-injured youth. The American Academy of Pediatrics and others have suggested that primary care providers take an active role in violence risk assessment.25,26 For this strategy to be effective, it is critical to know whether assault-injured youth utilize primary care providers for risk assessment and intervention. We found that the majority of assault-injured youth and controls had a regular health care provider and had had a checkup in the past year. Thus, health care providers in primary care and in the ED may have opportunities for intervention. Group homes and schools are other locations of potential contact.
Although it is possible that the majority of assault-injured youth may be reached in primary care or in schools, in our study a notable proportion of youth did not have a regular health care provider (14%) and were not enrolled in school (9%). Though this was not significantly different across cases and controls, it highlights the need to explore other venues to reach these potentially high-risk adolescents and young adults.
The unique combinations of risk and protective factors of the injured individual may influence the likelihood of future violence and injury. Our study found that group home residents, living apart from parents, history of many fights, fight or weapon injuries, or witnessing weapon injuries were all possible risk factors for assault injury presenting to an ED. Few studies have had the benefit of control groups to assess the usefulness of these risk factors for screening. One longitudinal study followed a sample of youth over time (median: 5.2 years) and found that school status, drug use, and fighting history were most predictive of violence-related injuries.41 Our study confirms the association of fighting history with assault injury, but in our population, associations with school status and drug and alcohol use were inconsistent at best.
We queried a variety of social and community factors, behavioral factors, and violence factors that have been discussed in the literature. Surprisingly, only a small number of risk factors and no protective factors tested consistently differentiated assault-injured youth from controls. There could be many explanations for this finding including inappropriate control groups as discussed above, lack of sensitivity of our measures to detect differences (especially for protective factors), and the high risk of the community studied. Our case and control populations may have been from communities of such high risk and high violence exposure that predictive factors previously reported in lower-risk communities were not applicable. The high levels of violence exposure in both the cases and controls suggest that many youth were at-risk. The risk and protective factors that we tested were only those of individual risk behavior and exposure. Other social and community factors such as school performance and connectedness, family relations, and community connectedness were not examined and may offer greater insight into risk for assault injury.
Although there were inconsistent differences between cases and the 2 control groups there is a suggestion that a higher proportion of assault-injured youth were involved in other high-risk behavior including trying cigarettes, marijuana, and increased sexual activity. This is consistent with literature and theory on problem behavior. More than 2 decades ago Jessor and Jessor28 articulated the theory that a cluster of co-varying problem behaviors contribute to and comprise a pattern of risk and this has been confirmed by the Centers for Disease Control and Prevention’s Youth Risk Behavior Survey42 and the World Health Organization’s survey Health Behavior in School Children.43 Jessor44 points out that a number of structural factors (biological characteristics, personality, and social environment) are related and interact with adolescent risk behaviors in both risk-enhancing and protective ways. This study suggests that fighting behavior and assault injury may be part of this constellation of problem behaviors.
Some predictors of assault injury included previous number of lifetime physical fights, fights in the last year, and fight injuries. Unfortunately, among cases the wording of the fight questions did not clearly exclude the fight that brought them into the ED, thus limiting the predictive value of this screening question. However, after conservatively eliminating the current fight or injury among cases, there remained a significant difference between cases and controls on history of a weapon injury and fight-related injury (noninjured controls OR: 2.00; 95% CI: 0.60, 6.64). The vast majority of youth in all groups had been in at least 2 physical fights in their lifetime (92% assault-injured, 78% unintentionally injured, 71% noninjured). In the past year, most had had at least 1 fight (80% assault-injured, 55% unintentionally injured, 46% noninjured). These data support claims that fighting is normative in some communities. Interventions need to focus not only on decreasing numbers of fights, but also increasing ways to save face and disengage from fights and decreasing injury potential. Although fighting may be normative, our study suggests that an increased number and severity of past fights are associated with increased risk for injury.
A strong risk factor for assault injury was previous weapon injury and witnessing weapon injury. In our assault-injured group, 37% had been injured by a weapon. Twenty-nine percent of the assault-injured youth stated that they had been injured with a weapon in the past 12 months which was 8.5 to 9.5 times higher than the 2 control groups (unintentional control group OR: 9.50; 95% CI: 3.39, 26.60; noninjured control group OR: 8.50; 95% CI: 3.02, 23.95). Even subtracting those with weapon injuries that brought them into the study, 16% of the assault-injured group had had a previous weapon injury in the last year, 2.4 to 3.5 times higher than controls (unintentional OR: 3.51; 95% CI: 1.37–9.29; noninjured OR: 2.39; 95% CI: 1.13, 5.06). Interestingly, ability to get a gun, access to a weapon in the home, and report of weapon-carrying were not significantly different among the groups. Clearly, decreasing injury potential by decreasing access to and use of weapons are critical for prevention. Focus on individuals with previous weapon injury is warranted and screening for this risk in health care settings may assist in management.
Violence risk screening protocols have been developed for use in primary care. For example, the FISTS pneumonic developed by Sege45 assesses risk using (F) previous fights, (I) previous injuries, (S) history of sexual violence, (T) threats of violence, and (S) self-defense strategies including weapon access. Our data support a limited number of the components of this pneumonic in this population (previous fights and injuries) and suggest that in high-risk populations, screening for risk may be difficult and a universal prevention strategy may be needed. To guide prevention, knowledge of the strength of predictive factors is critical in choosing a targeted or universal approach. Finally, identifying settings of contact with at-risk adolescents who may or may not attend school is another crucial factor in designing programs. This study suggests that primary care sites, EDs, and group homes may be locations of contact for intervention.
This project was supported by the Robert Wood Johnson Foundation Generalist Faculty Scholars Program (Dr Cheng), the Centers for Disease Control and Prevention R49/CCR331657 and the Maternal and Child Health Bureau (Title V, Social Security Act), Health Resources and Services Administration, Department of Health and Human Services, R40MC00174.
- Received July 9, 2002.
- Accepted November 12, 2002.
- Address correspondence to Tina L. Cheng, MD, MPH, 9616 Accord Dr, Potomac, MD. E-mail:
Dr Cheng’s current affiliation is: Division of General Pediatrics and Adolescent Medicine, Johns Hopkins University, Baltimore, Maryland.
- ↵National Center for Health Statistics. Prevention Profile. Health, United States, 1989. Hyattsville, MD: US Dept of Health and Human Services; 1990. DHHS Publ. No. PHS 90-1232
- ↵Institute of Medicine. Reducing the Burden of Injury Washington, DC: National Academy Press; 1999:42–43
- Alexander CS, Ensminger ME, Somerfield MR, et al. Behavioral risk factors for injury among rural adolescents. Am J Epidemiol.1992;136 :673
- ↵Irwin CE, Cataldo MF, Matheny AP, Peterson L. Health consequences of behaviors: injury as a model. Pediatrics.1992;90 :798– 807
- Marshall WN, Bowen K, Aldous MB. Adolescent deaths in relation to past delinquency and history of abuse or neglect in the family. Ambulatory Pediatric Association Abstracts.1998:98
- Redeker NS, Smeltzer SC, Kirkpatrick J, Parchment S. Risk factors of adolescent and young adult trauma victims. Am J Crit Care.1995;4 :370– 378
- Herrenkohl TI, Maguin E, Hill KG, Hawkins JD, Abbott RD, Catalano RF. Developmental risk factors for youth violence. J Adolesc Health.2000;3 :176– 186
- ↵Ellliott DS, Huizinga D, Menard S. Multiple Problem Youth: Delinquency, Substance Use, and Mental Health Problems. New York, NY: Sprinnger-Verlag; 1989
- ↵Office of the Surgeon General. Youth Violence: A Report of the Surgeon General. Released January 17, 2001. Available at: http://www.surgeongeneral.gov/library/youthviolence/report.html. Accessed June 19, 2002
- ↵Huesmann LR, Guerra NG. Children’s normative beliefs about aggression and aggressive behavior. J Pers Soc Psychol.1997;2 :408– 419
- ↵American Academy of Pediatrics, Task Force on Adolescent Assault Victim Needs. Adolescent assault victim needs: a review of issues and a model protocol. Pediatrics.1996;98 :991– 1001
- ↵American Academy of Pediatrics, Task Force on Violence. The role of the pediatrician in youth violence prevention in clinical practice and at the community level. Pediatrics.1999;103 :173– 181
- ↵Cheng TL, Wright JL, Fields CB, Brenner RA, Ricardo IB, Scheidt PC. Urban adolescent views of the causes of teen violence [abstract]. 1997 APHA Annual Meeting Abstracts. Ambulatory Child Health.1997;3 :223– 224
- ↵Jessor R, Jessor SL. Problem Behavior and Psychosocial Development: A Longitudinal Study of Youth. New York, NY: Academic Press; 1977
- ↵Office of the Surgeon General. Youth Violence: A Report of the Surgeon General. Released January 17, 2001. Available at: http://www.surgeongeneral.gov/library/youthviolence/report.html. Accessed June 19, 2002
- ↵Brener ND, Collins JL, Kann L, Warren CW, Williams BI. Reliability of the Youth Risk Behavior Survey Questionnaire. Am J Epidemiol.1995;141 :575– 580
- ↵Ziv A, Boulet JR, Slap GB. Emergency department utilization by adolescents in the United Sates. Pediatrics.1998;101 :987– 994
- ↵Overpeck MO, Scheidt PC, Rouse BA et al. Activities and locations associated with adolescent nonfatal injuries in the United States, 1996 [abstract]. Poster presented at the World Conference on Injury Prevention; May 17–20, 1998; Amsterdam, the Netherlands
- ↵Alpert E, Sege R, Bradshaw Y. Interpersonal violence and the evaluation of physicians. Acad Med.1997;72 :S42– S50
- Copyright © 2003 by the American Academy of Pediatrics