Journal of Adolescent Health 58 (2016) 208e214

www.jahonline.org

Original article

Polyvictimization and Youth Violence Exposure Across Contexts

Heather A. Turner, Ph.D. a, b, *, Anne Shattuck, Ph.D. a, David Finkelhor, Ph.D. a, b, and
Sherry Hamby, Ph.D. c
a
Crimes Against Children Research Center, University of New Hampshire, Durham, New Hampshire
b
Department of Sociology, University of New Hampshire, Durham, New Hampshire
c
Department of Psychology, Sewanee: The University of the South, Sewanee, Tennessee

Article history: Received March 31, 2015; Accepted September 24, 2015
Keywords: Polyvictimization; Violence; Trauma symptoms

A B S T R A C T
IMPLICATIONS AND
CONTRIBUTION
Purpose: The current research used latent class analysis to uncover groups of youth with specific
victimization profiles and identify factors that are associated with membership in each victimi-
Six groups of youth with
zation group. different victimization
Methods: This study used data from National Survey of Children Exposure to Violence II. Random profiles emerged. Groups
digit dialing and address-based sampling were used to obtain a nationally representative sample of who experienced victimi-
2,312 youth ages 10e17 years. Phone interviews, averaging 55 minutes in length, were conducted zations across different
with caregivers to obtain both consent and background information and then with youths environments generally
themselves. had more troubling pro-
Results: Six groups of youth emerged: (1) nonvictims (26.4%), (2) home victims (8.4%), (3) school files than groups victim-
victims (20.8%), (4) home and school victims (21.3%), (5) community victims (5.4%), and (6) pol- ized within single
yvictims (17.8%). Polyvictims were likely to have been victimized in multiple settings by multiple domains, like only school
perpetrators and experienced the most serious aggravating characteristics, including incidents or only home. Polyvictims,
involving a weapon, injury, or a sexual component. Youth in the polyvictim class experienced the with by far the most
highest number of different victimizations types in the past year and had the most problematic serious profile, appear to
profile in other ways, including greater likelihood of living in disordered communities, high have no “safe haven” for
probabilities of engaging in delinquency of all types, elevated lifetime adversity, low levels of positive development.
family support, and the highest trauma symptom scores. This group represents a
Conclusions: The study supports the contention that a core basis of the particularly damaging effects crucial target for
intervention.
of polyvictimization is the experience of victimization across multiple domains of the child’s life.
Ó 2016 Society for Adolescent Health and Medicine. All rights reserved.

Considerable research has documented high rates of exposure
to many different forms of victimization among children and
youth. For example, a 2011 national survey found that over 41% of
Conflicts of Interest: The authors have no conflicts of interest to disclose.
Disclaimer: Points of view or opinions in this document are those of the au-
thor(s) and do not necessarily represent the official position or policies of the US
Department of Justice.
* Address correspondence to: Heather A. Turner, Ph.D., Crimes Against Chil-
dren Research Center, University of New Hampshire, 125 McConnell Hall, 15
Academic Way, Suite 102C, Durham, NH 03824.
E-mail address: heather.turner@unh.edu (H.A. Turner).
all U.S. children ages 1 month to 17 years experienced a physical
assault in the past year, almost 14% experienced some form of
child maltreatment, and over 8% witnessed domestic violence [1].
There is also no doubt of the damaging consequences of the many
forms of child victimization on emotional, behavioral and devel-
opmental problems [2e5]. Much child victimization research,
however, has focused separately on particular types of child vic-
timizations in specific environments such as the home, in the case
of child maltreatment, or the school, in the case of bullying.
More recently, there has been a greater acknowledgement of
the importance of understanding the intersections of violence

1054-139X/Ó 2016 Society for Adolescent Health and Medicine. All rights reserved.
http://dx.doi.org/10.1016/j.jadohealth.2015.09.021
 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214
and victimization across different contexts and domains of
exposure. Research on polyvictimization, for example, has
highlighted how some youth are exposed to many different forms
of victimization, showing that over 48% of children and youth
experience two or more victimization types and over 15% are
exposed to six or more different types within a year [1]. Poly-
victimization has been found to be more highly related to
adverse child outcomes than experiencing a single, even serious
and repeated, type of victimization [6,7].
However, most studies to date have not addressed the di-
versity of victimized youth in ways that allow the researcher
to identify groups with particular victimizations patterns. For
example, it has been suggested that polyvictimization may be
especially damaging because youth who are exposed to many
different types of victimization are likely to experience serious
incidents involving multiple life contexts and many types of
perpetrators [7]. Yet, existing research has not directly specified
patterns of variation in the severity, location, and perpetrator
type among juvenile victims.
Youth with different victimization profiles likely also differ on
other important risk factors. For example, some groups of victims
may be particularly likely to reside in neighborhoods with high
levels of community disorder, places with rundown buildings,
graffiti, public drinking, vandalism, and crime [8e10]. Those
victimized largely by family members may most often live in
households characterized by high conflict and low support [11].
Youth victimized across different environments may be exposed
to particularly high levels of adversity [12], may be more likely to
engage in delinquency [13], and have greater risk of mental
health problems than single-domain victims [6]. Specifying how
youth with different clusters of victimization experiences also
differ in individual, family, and community factors will help to
increase our understanding of the sources of risk and resilience
in specific groups of youth victims and, ultimately, how partic-
ular vulnerabilities might be addressed.
Using a nationally representative sample of youth ages 10e17
years, the current research used latent class analysis (LCA) to
uncover groups of youth with specific victimization profiles and
describe their characteristics. LCA represents a uniquely power-
ful approach for identifying different profiles of youth victimi-
zation. Because latent groups in LCA are defined by combinations
of responses on the indicator variables, LCA makes it possible to
take into account multiple dimensions of victimization instead of
focusing only on single indicators such as number of different
types of victimization. LCA thus offers a “person-centered” rather
than a “variable-centered” approach to examining heterogeneity
among youth in terms of their experience of victimization
[14,15]. The primary objectives of this study are to (1) identify
underlying groups of adolescents based on their profiles of past
year victimization in terms of place, perpetrator, and aggravating
characteristics; (2) describe the prevalence of these latent groups
among 10- to 17-year-olds in the United States; and (3) using
posterior probabilities from the LCA, assign youth to their most
likely latent class of victimization and describe the characteris-
tics of each victimization group.
Methods
Participants
The National Survey of Children’s Exposure to Violence II
(NatSCEV II) was designed to obtain up-to-date incidence and
209
prevalence estimates of a wide range of childhood victimizations
and a variety of correlates. The larger survey consists of a national
sample of 4,503 children and youth ages 1 month to 17 years of in
2011; the present study focuses on the subsample of 2,312 chil-
dren and youth who were aged 10e17 years at the time of the
survey.
The primary foundation of the design was a nationwide
sampling frame of residential telephone numbers from which a
sample of telephone households was drawn by random digit
dialing. Two additional samples were drawn from sampling
frames chosen to represent the growing number of households
that rely entirely or mostly on cell phones: a small national
sample of cellular telephone numbers drawn from random digit
dialing methodology (N ¼ 31) and an address-based sample (N ¼
750). Approximately, one-half of the eligible households ob-
tained through address-based sample were cell phoneeonly
households and thus represented an effective way of including
households without landlines in our sample.
The cooperation and response rates averaged across the three
sampling frames were 60% and 40%, respectively, which are good
rates by current survey research standard [16e18]. Sample
weights were constructed to adjust for design effects of the
sampling procedure and for demographic differences between
the sample and the national population of children age <18 years.
Additional details regarding sampling frames, sample weighting,
and nonresponse analysis may be obtained from the authors.
Procedure
A short interview was conducted with an adult caregiver
(usually a parent) to obtain family demographic information.
One child was then randomly selected from all eligible children
living in a household. If the selected child was 10- to 17-years old
(the focal group for this study), the main telephone interview
was conducted with the child.
Respondents were promised complete confidentiality and
were paid $20 for their participation. The interviews, averaging
55 minutes in length, were conducted in either English or
Spanish. Respondents who disclosed a situation of serious threat
or ongoing victimization were recontacted by a clinical member
of the research team, trained in telephone crisis counseling,
whose responsibility was to stay in contact with the respondent
until the situation was appropriately addressed locally. All pro-
cedures were authorized by the Institutional Review Board of the
University of New Hampshire.
Measures
The survey instrument for NatSCEV II was a version of the
Juvenile Victimization Questionnaire which asks respondents
about their lifetime exposure to 51 specific types of crime and
violence including physical assault, property crime, sexual as-
sault, maltreatment, peer and sibling victimization, and wit-
nessing violence. A list of the victimization survey items has been
published elsewhere [1]. For each victimization item that youth
reported having experienced, a series of follow-up questions
were asked to gather additional information about the victimi-
zation incident including whether it took place in the past year;
where it took place; the age (juvenile or adult) and relationship
of the perpetrator; and whether any aggravating circumstances,
such as weapon, injury or sexual content accompanied the
victimization.
210 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214
Observed indicators of latent class
The variables used as indicators of latent class were chosen to
represent a diversity of possible contexts and characteristics of
youth victimization. We constructed a series of 10 dummy vari-
ables to reflect these incident characteristics and then used these
aspects of children’s past-year victimization to explore the latent
class structure of victimization in this sample. Variables were
coded 1 if the child experienced each characteristic and 0 if not.
Location. Four variables indicate whether youth experienced any
victimization in the past year in each of the following places:
home, school, the Internet, or some other location.
Perpetrator. Four dummy variables are used to indicate whether
the youth had any victimization by each of four possible perpe-
trator types: adult family perpetrator, juvenile family perpetrator
(siblings or other juvenile relatives), adult nonfamily perpetrator,
and juvenile nonfamily perpetrator.
Aggravating circumstances. To measure the seriousness of vic-
timizations, two variables indicate whether youth experienced
any incident that involved a weapon or injury and any incident
that had a sexual component (including an attempted or
completed sexual assault, flashing, verbal sexual harassment, or
unwanted “sex talk” on the Internet). Previous research has
found that weapon use, injury, and any sexual component are
significant aggravating characteristics in peer victimization in-
cidents [19].
Factors associated with latent class
The following variables pertaining to youth, their families,
and their communities were examined to describe the latent
classes in post hoc analyses.
Demographic characteristics. Gender is coded 1 for male and 0 for
female. Family structure of the youth’s household is coded into
four categories: two parents, one parent living with stepparent
or partner, single parent, or other adult caregiver. Age in years is a
continuous variable. Race/ethnicity is coded in 4 categories: white
non-Hispanic, black non-Hispanic, other race non-Hispanic, and
Hispanic any race.
Community disorder. The NatSCEV survey included nine items
asking about youths’ exposure to criminal activities, gangs, or
deteriorated physical conditions in their neighborhoods and
schools. Example of items include witnessing illicit drug sales,
seeing gang-related graffiti, and witnessing the police make an
arrest in one’s neighborhood. The mean number of items re-
ported in the past year by youth in this sample was .94. We
created a dummy variable coded 1 to indicate high community
disorder for youth who reported experiencing two or more of the
nine items in the past year.
Delinquency. Youth were asked about delinquent behaviors they
had engaged in during the past year. Based on factor analysis, we
selected items that represented three different categories of de-
linquent behaviors: violent delinquency (five items, e.g., carrying a
weapon, hitting, slapping, or pushing other kids); property de-
linquency (five items, e.g., take anything from another kids or a
teacher at school, take something from a store without paying for
it); and drug/minor delinquency (six items, e.g., cheat on school
tests, drink alcohol, smoke marijuana, abuse prescription drugs).
Three dummy variables were created, one for each type of de-
linquency, and coded 1 for youth who had engaged in any of the
activities of that type in the past year.
Lifetime nonvictimization adversity. Youth were asked about their
lifetime experience of 15 nonvictimization adverse events, such
as having a serious accident or illness, parent imprisonment,
parent unemployment, or the death of someone close. A lifetime
adversity score was created by summing the number of items the
youth reported (mean, 2.8; standard deviation [SD], 2.1; range,
0e12).
Family social support. Respondents were asked four questions
about perceived support from family (e.g., “I can talk about my
problems with my family”), slightly modified from the Multidi-
mensional Scale of Perceived Support [20]. Response options
were 1 “never,” 2 “seldom,” or 3 “often.” A sum score was created
from the four items (mean, 11.2; SD, 1.2; range, 3e12). A dummy
variable, low family social support, was coded 1 for youth who
scored in the roughly bottom fifth of respondents on the sum
score.
Number of past-year victimization types. We calculated the total
number of different victimization types (of 51 possible) that were
experienced by youth in the past year (range, 0e25; mean, 3.2;
SD, 3.5).
Trauma symptoms. Mental health status was measured through
the use of trauma symptom scores for the anger, depression,
anxiety, dissociation, and posttraumatic stress scales of the
Trauma Symptoms Checklist for Children (TSCC). For the purpose
of this study, the instrument was shortened to a total of 28 items.
Respondents were asked to indicate how often they had have
experienced each symptom within the past month on a four-
point scale from 1 (not at all) to 4 (very often). All item re-
sponses for the five scales together were summed to create an
aggregate trauma symptom score. The TSCC has shown very good
reliability and validity in both population-based and clinical
samples [21,22]. In this study, the a coefficient was .93 for the
TSCC (28 items). Scores were standardized for analyses.
Data analysis plan
Data analysis proceeded in several phases. First, descriptive
characteristics for the sample were calculated (Table 1). Next, we
fit a series of latent class models with 1e7 classes, using the 10
variables described previously pertaining to location, perpe-
trator, and aggravating characteristics. We evaluated the fit of
these seven models using the likelihood ratio test statistic, G2,
the Akaike information criterion (AIC), and the Bayesian infor-
mation criterion (BIC). A nonsignificant p value for the G2 statistic
is desired, but the G2 statistic loses accuracy when degrees of
freedom are large in a model with many parameters, as in the
case of our model [23,24]. Thus, we relied on the AIC and BIC test
statistics to judge model fit, with smaller values indicating better
model fit [23]. Identification of models with each specified
number of classes was checked by running the models with
1,000 randomly selected starting values and examining the
percentage of times that the model converged to the highest log-
likelihood value.
 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214 211

Table 1 Table 2
Sample characteristics (N ¼ 2,312) Fit statistics for different size latent class models of adolescent past year
victimization
Percent or
mean Number df G2a AIC BIC Entropy Solution %b
of classes
Demographic characteristics
Male (%) 51.7 1 1,013 6,253.3 6,273.3 6,330.8 1.00 100.0
Mean age in years (standard error) 13.6 (.07) 2 1,002 3,213.3 3,255.3 3,375.9 .92 100.0
Race/ethnicity (%) 3 991 2,400.9 2,464.9 2,648.8 .87 61.0
White, non-Hispanic 57.6 4 980 1,770.0 1,856.0 2,103.0 .85 96.0
Black, non-Hispanic 15.5 5 969 1,500.0 1,608.0 1,918.3 .88 60.0
Other race, non-Hispanic 7.7 6 958 1,252.6 1,382.6 1,756.1 .90 82.0
Hispanic, any race 19.1 7 947 1,111.31 1,263.31 1,700.00 .90 24.0
Family structure of household (%)
Two parents 52.0 Solution in bold is the selected model.
Parent and stepparent or partner 9.9 AIC ¼ Akaike information criterion; BIC ¼ Bayesian information criterion.
a
Single parent 32.8 p Values are not provided because the degrees of freedom are too large.
b
Other adult caregiver 5.3 Percentage of times the solution was selected in 1,000 sets of random
Past-year victimization characteristics starting values.
Location, any victimization (%)
At home 44.0
At school 54.3 identification and interpretability. The seven-class model had
At some other location 27.4 slightly lower AIC and BIC values, but this model was not well
On the Internet 11.6 identified, emerging as the maximum likelihood solution in only
Perpetrator, any victimization by (%)
24% of 1,000 sets of random starting values. The six-class model,
Adult family member 19.9
Juvenile family member 24.3 in contrast, was very well identified, with a clear modal solution
Adult nonfamily member 12.1 that emerged in 82% of random starts. Moreover, its latent classes
Juvenile nonfamily member 57.6 were more intuitively labeled and described than the seven-class
Victimization characteristics, any
model.
victimization involving (%)
Weapon or injury 22.4 The latent class prevalence’s and item response probabilities
Sexual component 13.3 for the six-class latent model are presented in Table 3. We have
Mean number of past year victimization types 3.2 named the six classes: nonvictims, home victims, school vic-
Percent of youth with seven or more past-year 14.6 tims, home and school victims, community victims, and poly-
victimization types
victims. In the discussion that follows, the small Greek letter
Other youth characteristics
Lives in high community disorder neighborhood (%) 25.4 rho (r) is used to refer to the estimated item response proba-
Delinquency past year, engaged in any (%) bilities within each class. To facilitate interpretation of the latent
Property delinquency 15.3 classes, item response probabilities of .50 are presented in
Violent delinquency 24.4
bold in Table 3.
Drugs/minor delinquency 31.1
Low family social support (%) 20.9
More than a quarter (26.4%) of 10- to 17-year-olds fell into
Mean lifetime adversity score (standard error) 2.8 (.07) the nonvictim class. These youth have a zero or near zero
probability of experiencing any victimization in any setting by
Means and percentages are weighted. N is unweighted.
any perpetrator in the past year. One-fifth of adolescents
(20.8%) qualified as school victims, with a near certain proba-
bility of experiencing at least one past-year victimization at
Once a final model was selected, we assigned individuals to school at the hands of a juvenile nonfamily peer but very low
the latent class for which they had the highest posterior proba- probabilities of being victimized in any other setting or of
bility of membership. The groups formed by this assignment experiencing the serious aggravating characteristics of weapon
were then compared on respondent characteristics using anal- use, injury, or sexual content. The home victims class (8.4% of
ysis of variance and chi-square tests. Descriptive analyses were youth) had a 1.0 probability of experiencing a victimization
conducted using Stata 13. Latent class models were fit using SAS at home, generally at the hands of juvenile family members
9.4 and the PROC LCA and LTA add-on software (version 1.3.1) (r ¼ .74) and, much less likely, at the hands of adult family
developed by The Methodology Center at Penn State University members (r ¼ .20). Members of this latent class had very
[25]. Sample weights were used in all analyses. low probabilities of experiencing any victimization in settings
outside the home. These youth might also well be described as
Results “sibling victims.”
Of youth, 21.3% fell into the home and school victim class,
Descriptive statistics for the overall sample, including socio- having a high probability of being victimized both at home (r ¼
demographic characteristics; victimization characteristics; and 1.0) and at school (r ¼ .78). Members of this class are most likely
the community, family, and individual characteristics of interest, to have been victimized by peers (r ¼.92) and siblings (r ¼.53).
are presented in Table 1. The probability of weapon or injury in a victimization (r ¼.39) is
also higher for this class than for the separate school and home
Latent class model victim classes; 5.4% of youth could be classified as community
victims meaning that they were very likely (r ¼ .99) to have been
Table 2 presents the fit statistics from the series of latent class victimized in a setting other than home, school, or the Internet.
models with 1e7 classes. We selected the six-class model as our The probability of victimization in these latter settings was low.
final model based on AIC and BIC values and on model Among possible perpetrators, the highest probability for the
212 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214

Table 3
Latent class prevalence and item response probabilities for a six-class model (N ¼ 2,312)

Latent class

Nonvictims Home victims School victims Home and school victims Community victims Polyvictims

Latent class prevalence 26.4% (1.42) 8.4% (.84) 20.8% (1.71) 21.3% (2.52) 5.4% (.80) 17.8% (1.99)
(standard error)
Probabililty of victimization characteristic (standard error)
Victimization at home .00 (.000) 1.00 (.001) .00 (.001) 1.00 (.001) .14 (.084) .76 (.076)
Victimization at school .01 (.020) .01 (.030) 1.00 (.000) .78 (.032) .00 (.015) .94 (.018)
Victimization at some other location .00 (.000) .05 (.025) .20 (.059) .27 (.041) .99 (.034) .67 (.058)
Victimization on the Internet .02 (.006) .02 (.015) .07 (.027) .03 (.029) .02 (.021) .49 (.079)
Adult family perpetrator .02 (.006) .20 (.041) .07 (.023) .32 (.037) .17 (.052) .48 (.060)
Juvenile family perpetrator .00 (.002) .74 (.046) .02 (.009) .53 (.036) .11 (.053) .32 (.051)
Adult nonfamily perpetrator .00 (.001) .02 (.016) .08 (.028) .16 (.025) .25 (.059) .31 (.047)
Juvenile nonfamily perpetrator .01 (.006) .00 (.003) .91 (.003) .92 (.030) .33 (.065) .98 (.012)
Victimization with injury or weapon .01 (.004) .17 (.032) .14 (.037) .39 (.038) .13 (.038) .50 (.060)
Victimization with a sexual component .01 (.005) .01 (.010) .04 (.024) .02 (.034) .06 (.029) .64 (.102)

Item probabilities of .50 are shown in bold to facilitate interpretation.

community victims class was to have had a juvenile nonfamily
perpetrator (r ¼.33) with adult nonfamily perpetrators being the
next most likely (r ¼ .25).
Finally, 17.8% of youth fell into the most highly victimized
class which we labeled polyvictims. Polyvictims are likely to
have been victimized in multiple settings (r ¼ .76 for home; .94
for school; .67 for elsewhere; and .49 for the Internet) and to
have experienced victimization at the hands of a number of
different people (r ¼ .48 for adult family member; .32 for sib-
ling; .98 for juvenile peer; and .31 for adult non family).
Members of the polyvictim class were also most likely to have
experienced serious aggravating characteristics including a .50
probability of having an incident that involved weapon or
caused injury and a .64 probability of an incident that included a
sexual component.
Individual, family, and community correlates of latent class
Table 4 presents the bivariate associations of demographic
characteristics, community disorder, past-year delinquency,
family support, lifetime adversity, trauma symptoms, and num-
ber of past-year victimization types with latent class member-
ship. A number of significant relationships are evident. Youth in
the polyvictim and community victim classes were the oldest on
average (mean of 14.7 and 13.9 years, respectively), while youth
victimized only at home were the youngest (12.9 years). Family

Table 4
Characteristics of latent class members (after-class assignment based on the highest posterior probability; N ¼ 2,312)

Latent class

Nonvictim; Home School Home and Community Polyvictims; p Value for

n ¼ 626 victims; victims; school victims; victims; n ¼ 115 n ¼ 354 c2 or F
n ¼ 220 n ¼ 509 n ¼ 488

Mean posterior probability (standard error) .998 (.000) .950 (.005) .943 (.005) .888 (.005) .924 (.014) .878 (.008)
Male (%) 48.3 56.4 53.8 61.1 66.4 34.0 .000
Age (mean) 13.4 12.9 13.7 13.0 13.9 14.7 .000
Family structure (%)
Two parents 57.4 66.0 54.2 45.2 53.0 41.5 .003
Parent and stepparent or 11.1 5.6 6.3 12.4 10.1 11.7
partner
Single parent 27.3 21.2 34.8 36.2 30.6 41.6
Other adult caregiver 4.1 7.3 4.7 6.2 6.3 5.3
Race/ethnicity (%)
White, non-Hispanic 60.1 58.9 59.6 53.4 58.9 55.8 .453
Black, non-Hispanic 10.8 18.5 17.4 17.3 21.7 14.8
Other race, non-Hispanic 6.7 8.0 6.8 8.2 8.9 9.5
Hispanic, any race 22.4 14.6 16.2 21.0 10.6 19.9
Lives in high community 7.0 14.6 21.8 32.0 19.1 49.4 .000
disorder neighborhood (%)
Delinquency past year, engaged in any (%)
Property delinquency 5.4 8.6 12.4 19.7 12.2 32.2 .000
Violent delinquency 6.1 13.2 15.5 33.2 20.0 47.5 .000
Drugs/minor delinquency 17.4 19.5 30.9 25.8 38.5 64.4 .000
Low family social support (%) 8.5 17.8 19.1 23.6 28.3 38.7 .000
Mean lifetime adversity score 1.9 2.0 2.6 3.2 2.4 4.5 .000
Mean trauma symptom scores (standardized) .61 .18 .13 .29 .12 .93 .000
Mean number of the past-year victimization types 0.2 2.1 2.5 4.7 1.9 8.1 .000

Chi-square used to test group differences in percentages. Analysis of variance used to test differences in means.
Percentages and means in table cells are weighted. N values and mean posterior probabilities are unweighted.
 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214
structure was related to group membership, with youth in two-
parent families underrepresented among polyvictims (41.5% of
polyvictims vs. 52.0% of the sample) and overrepresented among
home victims (66.0%). Youth residing in single-parent homes
made up 32.8% of the sample, but 41.6% of the polyvictim class
and only 21.2% of the home victim class. Black youth made up
21.7% of community victims and only 10.8% of nonvictims, but
the chi-square test of race by latent class was not significant.
Community disorder was strongly associated with group
membership. Only 7.0% of nonvictims lived in a high community
disorder neighborhood, but nearly half of polyvictims (49.4%)
did, followed by 32.0% of home and school victims, 21.8% of
school victims, and 19.1% of community victims. Polyvictims
were most likely to have been delinquent. Nearly one-third
(32.2%) had committed property delinquency acts, almost half
had engaged in some form of violent delinquency (47.5%) and
well over half in drug/minor delinquency (64.4%). Violent de-
linquency was also relatively high in the home and school victim
group (33.2%), and drug/minor delinquency was relatively high
among community victims (38.5%). Non-victims were substan-
tially less likely to have engaged in all forms of delinquency. More
than one-third of poly-victims (38.7%) reported low family sup-
port, whereas only 8.5% of nonvictims did. Polyvictims also had
the highest level of nonvictimization adversity, with a mean of
4.5 major lifetime events. Mean standardized trauma symptom
scores were highest among polyvictims (.93) followed by home
and school victims (.29), whereas nonvictims reported the lowest
level of distress (.61). Finally, mean number of past-year
victimization types was substantially higher among polyvictims
(8.1) than among the members of any other latent class. The
home and school victim group had the second highest mean
number of victimization types at 4.7.
An examination of all pairwise comparisons across latent
class membership (not shown) revealed a clear pattern whereby
polyvictims exhibited, by far, the greatest risk and worse out-
comes, followed by youth in the home-and-school latent class,
followed by each of the single-domain classes. Specific compar-
ison across the three single-domain classes showed the greatest
similarity between the school and community victims, signifi-
cantly differing only on the mean number of past-year victimi-
zation types (higher among the school victims). Home victims
were significantly younger and significantly less likely to be
engaged in minor or property delinquency than either the school
or community victims. Home victims also had significantly lower
lifetime adversity than both community and school victims
(possibly because of their younger age) and fewer past-year
victimization types relative to school victims.
Discussion
Using LCA, we uncovered six distinct groups of youths with
specific victimization profiles and identified factors associated
with membership in each victimization group. A group of poly-
victims emerged, representing 17.8% of all youth ages 10e17
years, who not only experienced many different forms of
victimization but who were also most likely to experience
serious incidents involving weapons, injury, and sexual content,
across multiple contexts (home, school, elsewhere, Internet), and
with multiple perpetrators. It is notable that this is the only
victim group with substantial probabilities of experiencing vic-
timizations perpetrated by adult family members and low levels
of family support. This suggests that caregiver dysfunction may
213
be importantly linked to a broad level of vulnerability, whereby
youth with such caregivers are also likely to be victimized in
nonfamily contexts and by other people. The greater probability
of living in disordered neighborhoods among polyvictims also
suggest the importance of the broader environmental context
and is consistent with research linking community disorder and
youth exposure to both family- and peer-perpetrated violence
[8,26]. Polyvictims also had the most troubling profile in other
ways, including an elevated exposure to nonvictimization
adversity, and high probabilities of engaging in delinquency of all
types. Property and violent delinquency were especially likely in
this group, suggesting strong links between outward-directed
antisocial behaviors and high levels of cross-context victimiza-
tion. This group also had, by far, the highest levels of trauma
symptoms.
Although all victim groups had significantly elevated scores
than nonvictims on the risk factors assessed, results suggest that
the youth victimized in single contexts have generally less-
serious risk profiles than victims in multiple contexts. Home-
only victims, younger on average and largely victimized by
siblings, had the least troubling victim profile; school-only vic-
tims had somewhat higher levels of drug and property de-
linquency but were similar to home victims on other risk factors.
Community victims had somewhat higher probability of being
victimized by unrelated adult perpetrators than either home or
school only victims and had lower family support and greater
drug delinquency, but similar trauma symptoms scores. In
contrast, the home and school victim group, although not nearly
as troubled as polyvictims, was substantially higher than all
single-domain victims on violent and property delinquency,
lifetime adversity, and trauma symptoms. This supports the
contention that a core basis of the particularly damaging effects
of polyvictimization is the experience of victimization across
multiple domains of the child’s life [7]. For such children,
victimization represents not just a set of adverse events but a life
condition where there is no “safe haven” that allows for positive
experiences and normal social development.
Although the cross-sectional design of this study does not
allow us to establish temporal ordering of victimization group
membership and other risk factors, it provides some clues con-
cerning potential pathways linked to different victimization
profiles. This research also confirms the importance of moving
toward a more holistic and child-center approach to evaluating
and responding to victimized youth, one that addresses risks in
all aspects of the child life.
Funding Sources
For the purposes of compliance with Section 507 of PL 104-
208 (the “Stevens Amendment”), readers are advised that
funding was derived from multiple sources. This project was
supported by federal grant No. 2010-JF-FX-0001, awarded by the
Office of Juvenile Justice and Delinquency Prevention, Office of
Justice Programs, U.S. Department of Justice.
References
[1] Finkelhor D, Turner HA, Shattuck AM, Hamby SL. Violence, crime, and
abuse exposure in a national sample of children and youth: An update.
JAMA Pediatr 2013;167:614e21.
[2] Bond L, Carlin JB, Thomas L, et al. Does bullying cause emotional problems?
A prospective study of young teenagers. BMJ 2001;323:480e4.
214 H.A. Turner et al. / Journal of Adolescent Health 58 (2016) 208e214
[3] Browne A, Finkelhor D. The impact of child sexual abuse: A review of the
research. Psychol Bull 1986;99:66e77.
[4] Richters JE, Martinez P. The NIMH community violence project: 1. Children
as victims of and witnesses to violence. Psychiatry 1993;56:7e21.
[5] Toth SL, Manly JT, Cicchetti D. Child maltreatment and vulnerability to
depression. Dev Psychopathol 1992;4:97e112.
[6] Finkelhor D, Ormrod RK, Turner HA. Poly-victimization: A neglected
component in child victimization. Child Abuse Negl 2007;31:7e26.
[7] Turner HA, Finkelhor D, Ormrod R. Poly-victimization in a national sample
of children and youth. Am J Prev Med 2010;38:323e30.
[8] Turner HA, Shattuck A, Hamby SL, Finkelhor D. Community disorder,
victimization exposure, and mental health in a national sample of youth.
J Health Soc Behav 2013;54:258e75.
[9] Geis KJ, Ross CE. A new look at urban alienation: The effect of neighborhood
disorder on perceived powerlessness. Soc Psychol Q 1998;61:232e46.
[10] Kim J, Conley ME. Neighborhood disorder and the sense of personal con-
trol: Which factors moderate the association? J Community Psychol 2011;
39:894e907.
[11] Turner HA, Finkelhor D, Ormrod R, et al. Family context, victimization, and
child trauma symptoms: Variations in safe, stable, and nurturing re-
lationships during early and middle childhood. Am J Orthop 2012;82:
209e19.
[12] Ford JD, Grasso DJ, Hawke J, Chapman JF. Poly-victimization among juve-
nile justice-involved youths. Child Abuse Negl 2013;37:788e800.
[13] Cuevas C, Finkelhor D, Turner HA, Ormrod RK. Juvenile delinquency and
victimization: A theoretical typology. J Interpers Violence 2007;22:
1581e602.
[14] Collins LM, Lanza ST. Latent class analysis and latent transition analysis.
Hoboken, NJ: John Wiley & Sons, Inc; 2010.
[15] Lanza ST, Rhoades BL. Latent class analysis: An alternative perspective on
subgroup analysis in prevention and treatment. Prev Sci 2011;14:11.
[16] Pew Research Center. Assessing the representativeness of public opinion
surveys. Washington, DC: Pew Research Center; 2012.
[17] Keeter S, Kennedy C, Dimock M, et al. Gauging the impact of growing
nonresponse on estimates from a national RDD telephone survey. Public
Opin Q 2006;70:759e79.
[18] Babbie E. The practice of social research. 11th edition. Belmont: CA:
Wadsworth; 2007.
[19] Turner HA, Finkelhor D, Shattuck A, et al. Beyond bullying: Aggravating
elements of peer victimization episodes. Sch Psychol Q 2015;30:366e84.
[20] Zimet GD, Dahlem NW, Zimet SG, Farley GK. The multidimensional scale of
perceived social support. J Pers Assess 1988;52:30e41.
[21] Briere J. Trauma symptoms checklist for children (TSCC): Professional
manual. Odessa, FL: Psychological Assessment Resources; 1996.
[22] Briere J, Johnson K, Bissada A, et al. The Trauma Symptom Checklist for
Young Children (TSCYC): Reliability and association with abuse exposure in
a multi-site study. Child Abuse Negl 2001;25:1001e14.
[23] Collins LM, Lanza ST. Latent class and latent transition analysis: With ap-
plications in the social, behavioral, and health sciences. Hoboken, NJ: John
Wiley & Sons; 2010.
[24] Collins LM, Fidler PL, Wugalter SE, Long JD. Goodness-of-fit testing for
latent class models. Multivariate Behav Res 1993;28:375e89.
[25] Lanza ST, Dziak JJ, Huang L, et al. Proc LCA & Proc LTA Users’ Guide (Version
1.3.1). University Park, PA: The Methodology Center, Penn State. 2014.
Available at: methodology.psu.edu. Accessed September 1, 2014.
[26] Curry AD, Latkin CA, Davey-Rothwell M. Pathways to depression: The
impact of neighborhood violent crime on inner-city residents in Baltimore,
Maryland, USA. Soc Sci Med 2008;67:23e30.