Foshee

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Preventive Medicine 39 (2004) 1007 – 1016

www.elsevier.com/locate/ypmed

Longitudinal predictors of serious physical and sexual dating violence


victimization during adolescence
Vangie Ann Foshee, Ph.D., a,* Thad Steven Benefield, M.S., b Susan T. Ennett, Ph.D., a
Karl E. Bauman, Ph.D., a and Chirayath Suchindran, Ph.D. b
a
Department of Health Behavior and Health Education, School of Public Health, University of North Carolina at Chapel Hill,
Chapel Hill, NC 27599-7400, USA
b
Department of Biostatistics, School of Public Health, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-7400, USA

Available online 11 June 2004

Abstract

Background. This study identifies potentially modifiable risk factors for the onset of and chronic victimization from serious physical and
sexual dating violence.
Methods. One thousand two hundred ninety-one 8th and 9th graders from a county in North Carolina were assessed annually for 5 and 4
years, respectively.
Results. For males, having been hit by an adult with the intention of harm, having low self-esteem, and having been in a physical fight
with a peer predicted onset of serious physical dating violence victimization. Those variables, plus having a friend who has been a victim of
dating violence, alcohol use, and being white, predicted chronic victimization for males. For females, onset of serious physical dating
violence victimization was predicted by having been hit by an adult; that variable, plus living in a single-parent household, predicted chronic
victimization from serious physical violence. Also for females, onset of sexual violence victimization was predicted by having a friend who
has been the victim of dating violence and being depressed; those variables and gender stereotyping predicted chronic victimization from
sexual dating violence.
Conclusions. The findings identify high-risk groups and risk factors to target for intervention and have implications for approaches to
delivering dating violence prevention programs.
D 2004 The Institute For Cancer Prevention and Elsevier Inc. All rights reserved.

Keywords: Adolescent dating violence; Domestic violence; Youth violence

Introduction guish causes from consequences of dating violence [3,11–


15]. Unless the risk factors cause dating violence, however, it
Between 8% and 20% of adolescents have been victims of is unlikely their alteration by programs will reduce violence.
serious physical dating violence [1,2]. Between 10% and Cohort data are required to infer that the risk factors cause
18% of female adolescents have been sexually victimized by victimization. Only one prior study examined predictors of
a dating partner [3 – 6]. The risk factors targeted by most adolescent dating violence victimization longitudinally, and
adolescent dating violence prevention programs [7 – 10]— it was limited to examining demographic predictors of mild
including alcohol use, adherence to traditional gender stereo- forms of dating violence over a 1-year period [20].
typing, depression, low self-esteem, inadequate social com- This study uses data from a cohort of adolescents
petencies, and acceptance of dating violence—have been assessed first in the 8th or 9th grades then annually for 5
identified from cross-sectional studies that cannot distin- or 4 years, respectively, to identify potentially modifiable
predictors of the onset of and chronic victimization from
serious physical and sexual dating violence. We examine
* Corresponding author. Department of Health Behavior and Health physical dating violence victimization separately for males
Education, School of Public Health, University of North Carolina at Chapel
Hill, 317 Rosenau Hall, Chapel Hill, NC 27599-7400. Fax: +1-919-966- and females because many of the correlates of physical
2921. dating violence victimization vary by gender [11,15– 17].
E-mail address: foshee@email.unc.edu (V.A. Foshee). We examine sexual dating violence victimization for

0091-7435/$ - see front matter D 2004 The Institute For Cancer Prevention and Elsevier Inc. All rights reserved.
doi:10.1016/j.ypmed.2004.04.014
1008 V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016

females only because females are much more likely than of age to be in 12th grade for a total of five and four follow-
males to be the victims of sexual dating violence [3,5]. up waves for the 8th and 9th grade cohorts, respectively. At
Findings concerning the onset of serious physical and sexual each wave, adolescents who had dropped out of school or
dating violence could guide development of primary pre- were absent on the data collection days were mailed a
vention interventions; identification of predictors of chronic questionnaire to complete and return. The University of
victimization could guide secondary prevention efforts. North Carolina School of Public Health Institutional Review
Social ecological models of human development provide Board on Research Involving Human Subjects approved the
the theoretical framework for identifying and examining study methodology.
predictors [18,19]. Social ecological models direct attention
to the role of both individual and environmental factors in Serious physical and sexual dating violence victimization
the development of individual behavior. Thus, it encourages Adolescents were asked at each wave if they had ever
consideration of the influences of multiple facets of one’s been on a date. Those who never dated skipped the ques-
life on individual behavior. We categorize the potential tions assessing dating violence victimization and were
predictors into six domains that are central to the lives of coded as non-victims at that wave. If they had ever been
adolescents: three assess social environmental predictors on a date, they were asked ‘‘During the last year, how many
(peer environment, family environment, and social norms) times has any person that you have been on a date with done
and three assess individual level predictors (personal com- the following things to you? Only include it when the dating
petencies, involvement in other problem behaviors, and partner did it to you first. In other words, do not count it if
demographic characteristics). The risk factors examined they did it to you in self-defense.’’ Eighteen behaviors were
within each domain are those found by cross-sectional listed that ranged from relatively mild (e.g., pushed,
studies to be associated with adolescent dating violence or grabbed, or shoved me) to more serious (e.g., assaulted
those targeted for change by current dating violence pre- me with a knife or gun). Response options ranged from 0 for
vention programs. ‘‘never’’ to 3 for ‘‘10 or more times.’’ Because our focus is
A public health approach to prevention calls for identi- predictors of serious physical dating violence victimization,
fying factors that predict problem behaviors then interven- we consider only the six acts indicating seriousness based
ing to eliminate or reduce those risk factors so the chain of on their likelihood of resulting in physical injury: ‘‘tried to
causation will be broken. Because we use a social ecological choke me’’, ‘‘burned me’’, ‘‘hit me with a fist’’, ‘‘hit me
framework to identify the predictors to be examined, our with something hard besides a fist’’, ‘‘beat me up’’, and
findings can potentially guide the development of interven- ‘‘assaulted me with a knife or gun’’. We defined sexual
tions that move beyond the typical individual-level ap- dating violence victimization by the acts ‘‘forced me to have
proach to dating violence prevention, to approaches that sex’’ and ‘‘forced me to do other sexual things that I did not
target change in systems, such as families, peer groups, and want to do’’.
societal norms.
Predictor variables
The predictor variables are described in Table 1. The
Methods mean, standard deviation, and range of values for each
variable, stratified by gender, are in Table 2.
Study design
Samples
The data are from a study that evaluated the effectiveness
of an adolescent dating violence prevention program, Safe Two subsamples were used. The first subsample includ-
Dates. The 14 public schools in a primarily rural North ed 1,291 adolescents who reported at baseline that they
Carolina county with students in the 8th or 9th grade were lived with a mother and had never been a victim or
stratified by grade (8th or 9th grade) and matched on school perpetrator of any of the 18 dating violence acts and
size. One member of each matched school-pair was ran- who completed at least one follow-up questionnaire. Ma-
domly assigned to treatment or control. In October 1994, ternal presence was required because two potential risk
before program implementation, baseline data (wave 1) factors were mother supervision and mother education.
were collected in schools with self-administered question- Limiting the sample to those who were not victims or
naires from 81% (n = 1,965) of the enrolled adolescents. perpetrators of dating violence at baseline was necessary to
Program activities occurred between November 1994 and control for the temporality of relationships. The second
March 1995 in the treatment schools and included a play subsample included 219 adolescents who reported at base-
delivered by peers, a 10-session curriculum taught in the line having already been victims of mild forms of dating
required health and physical education classes, and a poster violence (i.e., acts other than the six acts defined as serious
contest. For details on program content and evaluation physical violence and the two acts defined as sexual
results, see Foshee et al. [8,21,22,32]. Follow-up data were victimization). This sample represents adolescents who
collected each spring until adolescents were in 12th grade or are likely to be at increased risk for becoming victims of
V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016 1009

Table 1
Measures for the predictor variables
Predictor variable Number Response categories Item or example item
of items
(Alpha)a
Peer environment
Friend victim of dating 1 0 = no friends, 1 = >0 friends Number of friends who told the respondent that his or her
violence date had been violent to him or her.
Family environment
Witnessed domestic violence 1 1 = never to 4 = 10 or more times ‘‘How many times have you seen one of your parents hit
the other parent?’’
Hit by an adult 1 1 = never to 4 = 10 or more times ‘‘How often has an adult ever hit you with the purpose of
hurting you?’’b
Mother supervision 2 0 = never to 5 = almost always ‘‘When you are away from home and not at school or
work, does she [the mother figure] know where you are?’’
Social norms
Acceptability of 2 0 = strongly disagree to 3 = strongly agree ‘‘It is ok for a girl to hit a boy if he hit her first.’’
females hitting malesc
(Foshee et al. [23])
Acceptability of males hitting 6 (0.79) 0 = strongly disagree to 3 = strongly agree ‘‘It is ok for a boy to hit his girlfriend if she did something
femalesd (Foshee et al. [23]) to make him mad.’’
Gender stereotyping (Foshee 11 (0.67) 0 = strongly disagree to 3 = strongly agree ‘‘In general, the father should have greater authority than
et al. [23]) the mother in making decisions.’’
Personal competencies
Self-esteem (Rosenberg [24]) 10 (0.81) 0 = strongly disagree to 3 = strongly agree ‘‘I feel that I have a number of good qualities.’’
Conflict resolution skills 7 (0.78) 0 = never to 3 = very often ‘‘During the last 6 months, when you were angry at
(Foshee et al. [23]) someone, how often did you do or feel the following
things?’’ (e.g., I yelled and screamed insults at the person,
I made nasty comments about the person to others).
Depression (Kandel and 5 (0.86) 0 = never to 4 = all the time During the past 6 months, how much have you been bothered
Davis [25]) or troubled by: (e.g., feeling unhappy, sad, or depressed,
feeling hopeless about the future).
Other problem behavior
Physical fight with a peer 1 1 = no, 2 = yes ‘‘Have you ever been in a physical fight with someone the
same sex and about the same age as you?’’
Alcohol use (Youth Risk 1 1 = never to 8 = all 30 days ‘‘During the past 30 days, on how many days did you have at
Behavior Survey [26]) least one drink of alcohol (beer, liquor, or wine cooler)?
Demographics
Mother educationc 1 1 = did not graduate from high school –
to 6 = more than four years of collegee
Racee 1 0 = Non-Hispanic white 1 = non-white –
Religious Importance 1 0 = not at all important to 3 = very –
important
Single-parent household 1 0 = two biological (adoptive) parents, –
1 = single parent
Two-parents (non-biological) 1 0 = two biological (adoptive) parents, –
household 1 = two parents of
other than biological or adoptive.
Forced sex by non-date 1 1 = prior victimization from sexual violence –
from someone other than a date, 0 = no prior
sexual victimization from other than a date.
a
Chronbach’s alpha for scale measures formed by three or more items.
b
Child abuse-reporting laws preclude asking if a specific parent hit the adolescent.
c
Included in male models only.
d
Included in female models only.
e
Non-white was composed of the response categories black, Hispanic, Asian, American Indian, and others.

serious physical or sexual dating violence over the course in the 8th or 9th grade became a victim of serious physical
of adolescence. and sexual dating violence by the 12th grade. An assump-
tion of life table analyses is that missingness is not associ-
Statistical analysis strategy ated with study outcomes. To verify this assumption, we
conducted attrition analyses (n = 1,965). Baseline serious
Life table analyses of time to event were used in both physical and sexual dating violence victimization, when
subsamples to determine the probability that an adolescent controlling for all of the baseline predictor variables, did
1010 V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016

Table 2 event if they reported at any follow-up wave, being victim-


Distributions of predictor variables by gender
ized at least once by any of the six serious physical acts of
Predictor variable Males Females dating violence. They were coded as having a sexual dating
Mean STD Range Mean STD Range violence victimization event if they reported at any follow-
Peer environment up wave being victimized by at least one of the two sexual
Friend victim 0.11 0.31 0.00 – 1.00 0.17 0.38 0.00 – 1.00 dating violence acts. The second variable, time of event,
of dating indicated the wave of occurrence.
violence
Because of missing data, certain assumptions were made
Family environment
Witnessed 1.21 0.54 1.00 – 4.00 1.27 0.63 1.00 – 4.00 in creating those two variables. If data were missing before
domestic the wave when the adolescent reported being a victim, the
violence adolescent was coded as being a victim and the time to event
Hit by an 1.27 0.64 1.00 – 4.00 1.27 0.63 1.00 – 4.00 was the first non-missing wave. This is a potential misclas-
adult
sification problem only in that we do not know precisely
Mother 4.21 0.79 1.00 – 5.00 4.62 0.61 1.00 – 5.00
supervision when the event occurred. In our sample, there are six such
Social norms cases out of 1,291 when considering serious physical victim-
Acceptability 1.29 0.91 0.00 – 3.00 1.00 0.88 0.00 – 3.00 ization, and four such cases out of the 653 females when
of hitting considering sexual victimization. Misclassification is also
dates under
possible if a respondent is missing data on earlier but not
certain
circumstances later waves, and does not report victimization because the
Gender 0.83 0.45 0.00 – 2.30 0.58 0.39 0.00 – 2.09 individual may have been a victim at a missed wave. In our
stereotyping sample, there are 129 such cases out of 1,291 when consid-
Personal competencies ering serious physical dating violence, and 135 such cases out
Self-esteem 2.41 0.51 0.10 – 3.00 2.33 0.52 0.60 – 3.00
of 653 when considering sexual dating violence. Because of
Conflict 0.84 0.66 0.00 – 3.00 0.69 0.55 0.00 – 2.71
resolution the number of adolescents potentially misclassified in this
skills way, we conducted analyses both with and without the 129
Depression 1.34 0.95 0.00 – 4.00 1.86 0.93 0.00 – 4.00 and 135 adolescents. The predictors of onset of serious
Other problem behavior physical dating violence and sexual dating violence were
Physical fight 0.75 0.43 0.00 – 1.00 0.28 0.45 0.00 – 1.00
identical with and without these adolescents. All adolescents
with a peer
Alcohol use 1.97 1.29 1.00 – 8.00 1.70 0.93 1.00 – 7.00 are included in the analyses that are reported.
Demographics A Poisson model was used to assess baseline predictors
Mother 2.41 0.63 1.00 – 3.00 2.32 0.69 1.00 – 3.00 of chronic victimization from serious physical and sexual
education dating violence. The chronic serious physical victimization
Race 0.25 0.43 0.00 – 1.00 0.23 0.42 0.00 – 1.00
outcome variable was created by counting the number of
Religious 2.45 0.84 0.00 – 3.00 2.48 0.76 0.00 – 3.00
importance follow-up waves at which at least one serious physical
Single parent 0.15 0.36 0.00 – 1.00 0.13 0.33 0.00 – 1.00 victimization act occurred. Thus, adolescents in the 8th
household grade cohort could have values on this outcome ranging
Two-parents 0.21 0.41 0.00 – 1.00 0.21 0.41 0.00 – 1.00 from one to five and adolescents in the 9th grade cohort
(non-
could have values from one to four. Chronic victimization
biological)
Forced sex 0.02 0.13 0.00 – 1.00 0.07 0.26 0.00 – 1.00 by sexual dating violence was created in a parallel manner,
by non-date counting the number of waves at which at least one sexual
victimization event occurred. Using the Poisson model, we
investigated the relationship between the number of waves a
not predict dropout at any of the follow-up waves for either victimization event occurred and our covariates of interest,
males or females. controlling for the number of waves an individual had data.
Analyses for identifying predictors were conducted with Using the appropriate models, we conducted both bivar-
only the first subsample because the temporality of relation- iate and multivariate analyses. In the bivariate models, we
ships could be controlled in that subsample. A discrete regressed the dating violence outcome variable on one
proportional hazards model [27] was used to identify baseline predictor variable at a time. In the multivariate
baseline predictors of the onset of serious physical and models, we regressed the outcome variables on all 17
sexual dating violence victimization. For those analyses, it baseline predictor variables together. We include all varia-
was necessary to condense the data into two variables: one bles at once in the multivariate models because many of the
indicating the event that occurred and the other indicating associations observed in previous studies could have been
when the event occurred. The event could either be dropout confounded by uncontrolled variables related both to the
(in the statistical terminology of proportional hazards mod- risk factor and dating violence victimization. Because four
els, dropouts are called censored) or victimization. Adoles- of the risk factors examined were targeted by the Safe Dates
cents were coded as having a serious physical victimization program (i.e., the three social norms variables and conflict
V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016 1011

resolution skills), all bivariate and multivariate models that supervision, believing that it is acceptable for females to hit
included those four variables first test for their interaction their dating partners, holding traditional gender stereotypes,
with treatment condition. Nonsignificant interactions were having low self-esteem, having poor conflict resolution
dropped. All bivariate and multivariate models control for skills, having been in a physical fight with a peer, using
the main effect of treatment condition. We used the condi- alcohol, and having been forced to do something sexual by
tion index to test for collinearity in the multivariate models. someone other than a date (Table 3). The same baseline risk
A condition index with values over 100 may indicate factors predicted male chronic victimization.
substantial numerical instability and values over 30 may In multivariate models, when controlling for all other
indicate some cause for concern [28]. The condition index variables, being hit by an adult with the intention of harm,
for the complete set of covariates adjusting for the intercept having low self-esteem, and having been in a physical fight
was 2.51, suggesting the absence of collinearity. predicted onset of male serious physical dating violence
victimization (Table 4). The probability of victimization by
the 12th grade for a boy in the 8th or 9th grade who had
Results never been hit by an adult with the intention of harm, with
the highest possible self-esteem score, and who had never
Life table analyses results are presented in Fig. 1. The been in a physical fight is 0.06. In contrast, the probability
probability of becoming a victim of serious physical dating of victimization is 0.46 for a boy who has been hit by an
violence by the 12th grade for a non-victimized adolescent adult with the intention of harm 10 or more times, been in a
in the 8th or 9th grade was 0.21 (SE = 0.02) for males and physical fight, and had a self-esteem score one standard
0.16 (SE = 0.02) for females, and for females becoming a deviation below the sample mean. These same three varia-
victim of sexual dating violence was 0.21 (SE = 0.02). The bles, as well as having a friend who has been a victim,
probability of becoming a victim of serious physical dating drinking alcohol, and being Non-Hispanic white, also pre-
violence by the 12th grade for 8th and 9th graders who were dicted chronic victimization from serious physical dating
already victims of mild dating violence was 0.39 (SE = violence.
0.05) for males and 0.50 (SE = 0.06) for females, and for
females becoming a victim of sexual dating violence was Predictors of female onset of and chronic victimization from
0.28 (SE = 0.05). serious physical dating violence

Predictors of male onset of and chronic victimization from Significant bivariate predictors of female onset of serious
serious physical dating violence physical dating violence victimization were having a friend
who had been a victim of dating violence, being hit by an
Significant baseline bivariate predictors of male onset of adult with the intention of harm, lacking mother supervi-
serious physical dating violence victimization included sion, having been in a physical fight with a peer, and having
having a friend who had been a victim, having been hit been forced to do something sexual by someone other than a
by an adult with the intention of harm, lacking mother date (Table 5). Except for being forced to do something

Fig. 1. The probability of serious physical and sexual dating violence victimization by gender and wave in the two subsamples.
1012 V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016

Table 3 Predictors of female onset of and chronic victimization from


Bivariate predictors of male victimization from serious physical dating
sexual dating violence
violencea
Predictor variable Onset victimization Chronic victimization
Significant bivariate risk factors for becoming a victim
Hazard 95% CI b P value of sexual dating violence were having a friend who had
ratio
been a victim, having low self-esteem, being depressed,
Peer environment having a mother with a low level of education, and having
Friend victim 1.84* 1.09 – 3.10 0.72 0.001
been forced to have sex by someone other than a date
of dating violence
Family environment (Table 6). The same variables predicted chronic victimiza-
Witnessed domestic 0.95 0.65 – 1.40 0.04 0.83 tion from sexual dating violence. Other predictors of this
violence outcome were being hit by an adult with the intention of
Hit by an adult 1.33* 1.04 – 1.69 0.29 0.002 harm, and in the treatment group, holding traditional gender
Mother supervision 0.72** 0.58 – 0.90 0.37 0.0001
stereotypes.
Social norms
Acceptability of 1.37* 1.06 – 1.78 0.24 0.01 In multivariate analyses (Table 4), having a friend who
hitting dates has been a victim and being depressed predicted female
under certain onset of sexual dating violence victimization. The proba-
circumstances bility of being victimized by sexual dating violence by the
Gender 1.61* 1.07 – 2.43 0.43 0.02
12th grade for a girl in the 8th or 9th grade with no friends
stereotyping
Personal competencies who were victims of dating violence and with the lowest
Self-esteem 0.55*** 0.39 – 0.76 0.58 <0.0001 possible depression score is 0.10. In contrast, the proba-
Conflict resolution 1.47** 1.12 – 1.92 0.45 <0.0001 bility is 0.36 for a girl with a friend who has been
skills victimized and a depression score one standard deviation
Depression 1.10 0.90 – 1.36 0.12 0.20
above the mean sample score.
Other problem behavior
Physical fight with 2.22** 1.29 – 3.85 0.97 0.0001 In the multivariate models predicting chronic sexual
a peer victimization, the interaction between gender stereotyping
Alcohol use 1.26*** 1.12 – 1.42 0.28 <0.0001 and treatment condition was nonsignificant and therefore
Demographics dropped. Having a friend who was a victim of dating
Mother education 0.79 0.59 – 1.06 0.15 0.26
Race 0.87 0.55 – 1.37 0.17 0.39
Religious 0.95 0.75 – 1.20 0.04 0.72
Importance
Table 4
Single parent 1.13 0.68 – 1.88 0.01 0.98
Multivariate predictors (P < 0.05) by gender and outcomea
household
Two-parents 0.77 0.46 – 1.30 0.05 0.80 Predictor variable Onset victimizationb Chronic victimizationc
(non-biological) Hazard 95% CI b P value
Forced sex by 2.76* 1.01 – 7.50 0.89 0.05 Ratio
non-date
a
Male victimization of serious physical dating violence
Interactions between the independent variables theorized to be influenced Friend victim – – 0.58 0.02
by the intervention and treatment condition were examined and dropped if Hit by an adult 1.36* 1.03 – 1.79 0.30 0.01
nonsignificant (P > 0.05). All models control for the main effect of Self-esteem 0.62* 0.40 – 0.96 0.40 0.04
treatment condition. Been in a fight 1.90* 1.01 – 3.60 0.69 0.02
* P < 0.05. Alcohol use – – 0.17 0.01
** P < 0.01. Race – – 0.54 0.05
*** P < 0.001. Female victimization of serious physical dating violence
Hit by an adult 1.65*** 1.22 – 2.24 0.39 0.01
Single parent – – 0.72 0.01
sexual by someone other than a date, each of those variables household
also predicted chronic victimization. Additional predictors Female victimization of sexual dating violence
of chronic victimization were alcohol use and living in a Friend victim 1.80* 1.12 – 2.90 0.78 0.0001
Gender – – 0.56 0.02
single parent household. stereotyping
In multivariate analyses (Table 4), the only predictor of Depression 1.35* 1.05 – 1.74 0.29 0.01
female onset of serious physical victimization was having a
All models controlled for all 17 predictor variables and treatment
been hit by an adult with the intention of harm. Girls condition. Interactions between the independent variables theorized to be
without this risk factor have a probability of 0.12 of being influenced by the intervention and treatment condition were examined and
victimized by the 12th grade, whereas those who have been dropped if nonsignificant (P > 0.05). All models control for the main effect
hit by an adult with the intention of harm 10 or more times of treatment condition.
b
Time to event proportional hazards models.
have a probability of victimization of 0.43. That variable, c
Poisson models.
plus living in a single parent household, predicted female * P < 0.05.
chronic victimization. *** P < 0.001.
V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016 1013

Table 5 non-victimized peers to become victims of serious physical


Bivariate predictors of female victimization from serious physical dating
dating violence and 1.3 times more likely to become victims
violencea
of sexual dating violence. Another vulnerable group is
Predictor variable Onset victimization Chronic victimization
adolescents who have been victims of parental violence.
Hazard 95% CI b P value Being hit by an adult with the intention of harm was the
ratio
most consistent predictor regardless of gender or outcome,
Peer environment and parents were likely the perpetrators in most cases.
Friend victim of 1.79* 1.12 – 2.89 0.61 0.003
Another consistent predictor across gender and outcome
dating violence
Family environment was having a friend who had been a victim of dating
Witnessed domestic 1.08 0.79 – 1.47 0.07 0.64 violence. Because of our longitudinal design, we can infer
violence that adolescents were victimized after rather than before
Hit by an adult 1.67*** 1.34 – 2.09 0.44 <0.0001 their friend had been victimized, an important distinction
Mother supervision 0.73* 0.55 – 0.98 0.30 0.03
that could not be made in other studies [29]. Parents and
Social norms
Acceptability of 1.17 0.67 – 2.05 0.003 0.99
hitting dates
Table 6
under certain
Bivariate predictors of female victimization from sexual dating violencea
circumstances
Gender 0.78 0.45 – 1.36 0.30 0.22 Predictor variable Onset victimization Chronic victimization
stereotyping Hazard 95% CI b P value
Personal competencies ratio
Self-esteem 0.99 0.66 – 1.48 0.12 0.50
Conflict resolution 1.14 0.79 – 1.66 0.002 0.99 Peer environment
skills Friend victim of 1.77*** 1.16 – 2.69 0.74 <0.0001
Depression 1.10 0.87 – 1.37 0.11 0.26 dating violence
Other problem Family environment
behavior Witnessed domestic 1.08 0.82 – 1.43 0.06 0.67
Physical fight with 1.85** 1.21 – 2.83 0.58 0.002 violence
a peer Hit by an adult 1.16 0.90 – 1.50 0.26 0.01
Alcohol use 1.19 0.98 – 1.46 0.17 0.05 Mother supervision 0.91 0.68 – 1.23 0.003 0.98
Demographics Social norms
Mother education 0.98 0.72 – 1.33 0.10 0.46 Acceptability of 1.17 0.67 – 2.05 0.37 0.10
Race 1.01 0.62 – 1.66 0.002 0.99 hitting dates
Religious 0.86 0.66 – 1.12 0.20 0.07 under certain
importance circumstances
Single parent 1.66 0.96 – 2.86 0.73 0.001 Gender stereotyping 1.33 0.84 – 2.10 0.01 0.97
household Gender stereotyping * – – 0.52 0.03
Two-parents 1.15 0.70 – 1.88 0.04 0.87 treatment
(non-biological) Personal competencies
Forced sex by 2.06* 1.09 – 3.87 0.52 0.08 Self-esteem 0.66* 0.47 – 0.92 0.51 0.001
non-date Conflict resolution 1.14 0.82 – 1.58 0.14 0.35
a
skills
Interactions between the independent variables theorized to be influenced Depression 1.39** 1.14 – 1.69 0.31 0.0003
by the intervention and treatment condition were examined and dropped if Other problem behavior
nonsignificant (P > 0.05). All models control for the main effect of Physical fight 1.10 0.74 – 1.64 0.16 0.36
treatment condition. with a peer
* P < 0.05. Alcohol use 1.02 0.85 – 1.24 0.01 0.94
** P < 0.01. Demographics
*** P < 0.001. Mother education 0.72* 0.55 – 0.93 0.23 0.05
Race 1.01 0.65 – 1.57 0.03 0.89
Religious 0.92 0.73 – 1.16 0.10 0.32
violence, holding traditional gender stereotypes, and being importance
depressed were significant risk factors. Single parent 0.84 0.46 – 1.52 0.14 0.61
household
Two-parents 1.01 0.64 – 1.59 0.19 0.38
(non-biological)
Discussion Forced sex by 1.89* 1.06 – 3.37 0.63 0.01
non-date
a
Our findings suggest groups to target with interventions Interactions between the independent variables theorized to be influenced
to prevent serious physical and sexual dating violence, by the intervention and treatment condition were examined and dropped if
nonsignificant (P > 0.05). All models control for the main effect of
content for interventions, and approaches to delivering
treatment condition.
interventions. One group to target is young adolescents * P < 0.05.
who are already experiencing mild forms of dating violence. ** P < 0.01.
Those adolescents were 2.4 times more likely than their *** P < 0.001.
1014 V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016

practitioners need to assess the potential sources of risk to For example, one study found that endorsement of tradi-
adolescents who have friends who have been victimized by tional gender stereotypes predicted early sexual activity
dating partners. Perhaps having a peer group involved in [30]. An explanation for that finding was that girls with
dating violence exposes the adolescent to potential dating traditional stereotyping might be relatively submissive to the
partners who use violence against their partners, or the desires of boys, which may also apply to our findings.
adolescent and his or her friends are doing things and going Another study found that sexually active adolescent girls
places that puts them at risk for victimization. Finally, who endorsed traditional female stereotypes as opposed to
adolescents with multiple risk factors are at particularly those who did not endorse traditional female stereotypes
high risk. For example, a boy who has been hit by an adult were less likely to use contraceptives [31]. Activities that
10 or more times, has a self-esteem score two standard have been used in dating violence prevention programs to
deviations below the mean, and has been in a physical fight challenge traditional gender stereotypes have students ana-
with a peer has 2.7 times the risk of becoming a victim of lyze gender stereotypes represented in various media, ana-
serious physical dating violence than a boy with only the lyze their own gender stereotypes, and demonstrate the link
adult hit risk factor; 3.8 times the risk of a boy with only the between how having unfair expectations of a partner based
low self-esteem risk factor; and 3.4 times the risk of a boy on gender can lead to dating abuse [33]. Given the negative
with only the physical fight risk factor. impact on female adolescents of holding traditional gender
Risk factors to be targeted for change were identified, stereotypes and given the findings from numerous studies
thus providing guidance for developing the specific content that males who hold traditional gender stereotypes as
of both primary and secondary prevention interventions. For compared to those who do not are more likely to rape
boys, interventions designed to reduce peer fighting and [34 – 36], efforts should also be made to alter the represen-
increase self-esteem, in addition to attempts to decrease tation of traditional gender stereotypes in media targeted at
parental use of violence toward sons, could potentially adolescents.
prevent the onset of male serious physical dating violence In addition to the content of dating violence prevention
victimization. Addressing those same risk factors, plus interventions, our findings have implications for approaches
reducing alcohol use and assessing the potential risks to to delivering interventions. The risk factors varied by gender
young adolescent boys with friends who have been victims and outcome, suggesting the need for gender- and outcome-
of dating violence, could potentially lead to the prevention specific interventions, a departure from current practice.
of male chronic victimization. Poor conflict resolution skills Moreover, dating violence prevention programs are typical-
predicted male onset and chronic victimization in the ly delivered in schools, but our findings suggest that
bivariate models, and was marginally significant in the intervening with parents to change the family environment
multivariate model predicting male chronic victimization may also be an appropriate dating violence prevention
(P = 0.08). Thus, prevention programs targeted at boys approach. Based on our findings, appropriate content could
should consider incorporating content intended to improve include teaching parents nonviolent ways of responding to
conflict resolution skills. and disciplining their child, assessing the risks to their child
Programs for preventing sexual dating violence should due to his or her peer environment, and encouraging
consider components to identify and treat young adolescent parental monitoring. Additionally, interventions could teach
girls who are depressed. Depression was a risk factor for parents to be change agents by providing them with the
onset and chronic victimization from sexual dating vio- information they need to alter their adolescent’s norms
lence, and because of our longitudinal design, we can related to dating violence and gender-based expectations
conclude that the depression occurred before the sexual and teach their adolescent conflict resolution skills.
victimization, and that the significant associations were There are several potential limitations to this study. One
therefore not due to depression being a consequence of limitation is that we examined the risk factors after elim-
victimization. This fundamental distinction has not been inating a particularly high-risk group—those who reported
made in prior dating violence studies that relied on cross- victimization or perpetration at baseline. It is possible that
sectional designs. the risk factors for serious physical and sexual dating
Programs for preventing sexual dating violence should violence differ for adolescents who report early involve-
also consider including components designed to challenge ment in dating violence (i.e., before the 8th grade). This
traditional gender stereotypes. Girls who endorsed, as com- exclusion criterion was necessary, however, because our
pared to girls who did not endorse, traditional gender primary aim was disentangling the direction of relation-
stereotypes such as boys should be smarter than their girl- ships to distinguish causes from consequences of dating
friends, fathers should have greater authority than mothers violence.
in making family decisions, and it is more important for Other limitations relate to measurement. Several of our
boys than girls to do well in school were more likely to be variables were measured with one item and therefore they
chronically victimized from sexual dating violence. The may not have captured the complexity of some concepts.
negative effects on adolescent girls of endorsing traditional For example, religiosity has been defined along many
gender stereotypes have also been found in other studies. dimensions including frequency of participation in religious
V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016 1015

activities, religious denomination, feelings of spirituality, tion is about the same for males and females, which is
religious commitment, religious importance, and the amount consistent with almost all other adolescent dating violence
of influence religion has on one’s life and decisions [37]. We studies [3,7,11,20].
tapped only one of these dimensions, religious importance. Finally, the study was conducted in a primarily rural
However, all of our one-item constructs, except witnessing county. Given the high percentage of adolescents in the
domestic violence and religious importance, predicted at study county who completed baseline questionnaires (81%),
least one form of dating violence in bivariate analyses. the sample should approximate the characteristics of 8th and
We did not have measures of context that would have been 9th graders in the study county. These findings can be
helpful to more fully understanding the dynamics of dating generalized with some confidence to other rural counties
violence. For example, we were not able to distinguish with similar demographic characteristics. At the time of the
heterosexual from same sex dating violence encounters, but study, relative to the United States as a whole, the county
the predictors of serious physical and sexual dating violence had an overrepresentation of minority residents (20%),
could vary depending on the gender of one’s partner. Pre- lower income households (40% with less than US$10,000
dictors of serious physical and sexual dating violence could annual income), and individuals with limited education
also vary by the dynamics of a particular violent encounter. (53% of people more than 25 years of age have less than
For example, the predictors of sexual dating violence victim- a high school education). The confidence with which we can
ization could vary depending on whether the sexual violence generalize these findings to adolescents across the United
was due to psychological coercion or physical force. We did States is unknown. There are no longitudinal studies on
not have measures that allowed us to distinguish sexual adolescent dating violence for comparison that were con-
violence that resulted from psychological coercion from that ducted with a nationally representative probability sample.
resulting from physical force. Future studies on the etiology Conducting such studies should be a goal of future adoles-
of adolescent dating abuse victimization and perpetration cent dating abuse research.
should consider measuring more contextual information The study has many strengths. It is the first study to
about relationships and the abusive events. examine longitudinal predictors of onset and chronic
Consistent with social ecology models of human devel- victimization from serious physical and sexual dating
opment, we assessed predictors at both the individual and violence. The probability of being victimized over a 5-
environmental level. We found significant predictors at both year period of adolescence is determined as opposed to 1
levels and thus our findings can guide individual-level year of adolescence, which is the time frame considered in
interventions, for example, by focusing on changing indi- most other studies. Although previous dating violence
vidual attitudes and perceptions, and environmental-level studies have used multivariate modeling, none has incor-
interventions, for example, by changing the family and peer porated all of the variables considered in this study. Thus,
environments and altering the norms of society related to many of the associations observed in previous studies
dating abuse and gender-based expectations. Social ecolog- could have been confounded by variables related both to
ical models encourage consideration of the influence of the risk factor and dating violence. We controlled for
even higher-level or macro-level factors on individual confounding by examined the influence of each risk factor,
behavior. Macro-level factors that could influence the etiol- controlling for all other risk factors. Other strengths are
ogy of dating abuse include neighborhood characteristics, that the response rate was high, the sample size was large,
school characteristics, poverty, and life event stress. Unfor- and the sample came from the general population as
tunately, we did not have measures for those constructs. At opposed to selected groups such as adolescents using
this point, no studies have examined the effects of macro- dating violence-related services, or those in the judicial
level factors on the initiation of adolescent dating abuse, but system, thus increasing the generalizability of study
that should be a goal for future studies. results. Consistent with social ecology models of human
Another potential measurement limitation is our reliance development, the risk factors came from multiple domains
on self-reports of adolescent dating violence. However, self- of influence and thus the findings have implications for
reports are the only reasonable options for obtaining indi- developing interventions at multiple levels including indi-
vidual measures of adolescent dating abuse. Few incidents vidual, family, peer, and societal levels.
of adolescent dating violence are witnessed by others,
adolescents typically do not tell anyone about violent
incidents [8], and victims and perpetrators rarely appear in Acknowledgments
the law enforcement system. Evidence for the validity of our
dating violence measure comes from the similarity of our This publication was supported by grant Number U81/
baseline prevalence of serious physical forms of dating CCU409964-06 from the Centers for Disease Control and
violence victimization to those from the South Carolina Prevention. Its contents are solely the responsibility of the
[1] and the Massachusetts [2] Youth Risk Behavior Surveys. authors and do not necessarily represent the official views of
Other evidence supporting the validity of our dating vio- the funding agency. This project was initiated by and the
lence scale is our finding that the probability of victimiza- data were analyzed by the investigators.
1016 V.A. Foshee et al. / Preventive Medicine 39 (2004) 1007–1016

References [20] Halpern CT, Oslak SG, Young ML, Martin SL, Kupper LL. Partner
violence among adolescents in opposite-sex romantic relationships:
[1] Coker AL, McKeown RE, Sanderson M, Davis KE, Valois RF, Hueb- findings from the national longitudinal study of adolescent health. Am
ner ES. Severe dating violence and quality of life among South Car- J Public Health 2001;91(10):1679 – 85.
olina high school students. Am J Prev Med 2000;19(4):220 – 7. [21] Foshee VA, Linder GF, Bauman KE, Langwick SA, Arriaga XB, Heath
[2] Silverman JG, Raj A, Mucci LA, Hathaway JE. Dating violence JL, et al. The safe dates project: theoretical basis, evaluation design, and
against adolescent girls and associated substance use, unhealthy selected baseline findings. Am J Prev Med 1996;12(Suppl. 5):39 – 47.
weight control, sexual risk behavior, pregnancy, and suicidality. [22] Foshee VA, Bauman KE, Greene WF, Koch GG, Linder GF, Mac-
JAMA 2001;286(5):572 – 9. Dougall JE. The Safe Dates program: one-year follow-up results. Am
[3] O’Keefe M, Treister L. Victims of dating violence among high school J Public Health 2000;90(10):1619 – 22.
students. Are the predictors different for males and females? Violence [23] Foshee VA, Bauman KE, Linder F. Family violence and the perpe-
Against Women 1998;4(2):195 – 223. tration of adolescent dating violence: examining social learning and
[4] Bergman L. Dating violence among high school students. Soc Work social control processes. J Marriage Fam 1999;61:331 – 42.
92;37(1):21 – 27. [24] Rosenberg M. Society and adolescent self-image. Princeton, NJ:
[5] Foshee VA. Gender differences in adolescent dating abuse prevalence, Princeton Univ. Press; 1965.
types, and injuries. Health Educ Res 1996;11(3):275 – 86. [25] Kandel DB, Davies M. Epidemiology of depressive mood in ado-
[6] Molidor C, Tolman RM. Gender and contextual factors in adolescent lescents: an empirical study. Arch Gen Psychiatry 1983;39(10):
dating violence. Violence Against Women 1998;4(2):180 – 94. 1205 – 12.
[7] Avery-Leaf S, Cascardi M, O’Leary KD, Cano A. Efficacy of a dating [26] 2003 Youth Risk Behavior Survey High School questionnaire. Avail-
violence prevention program on attitudes justifying aggression. J able at: http://www.cdc.gov/nccdphp/dash/yrbs/2003/questionnaire.
Adolesc Health 1997;21(1):11 – 7. htm. Accessed December 20, 2002.
[8] Foshee VA, Bauman KE, Arriaga XB, Helms RW, Koch GG, Linder [27] Cox DR. Regression models and life-tables. J R Stat Soc, Ser [with
GF. An evaluation of safe dates, an adolescent dating violence pre- Discussion] 1972;34(Series B):187 – 220.
vention program. Am J Public Health 1998;88(1):45 – 50. [28] Belsley DA, Kuh E, Welsch RE. Regression diagnostics: identifying
[9] Jaffe PG, Sudermann M, Reitzel D, Killip SM. An evaluation of a influential data and sources of collinearity. New York: John Wiley and
secondary school primary prevention program on violence in intimate Sons, Inc.; 1980.
relationships. Violence Vict 1992;7(2):129 – 46. [29] Gwartney-Gibbs PA, Stockard J, Bohmer S. Learning courtship ag-
[10] Macgowan MJ. An evaluation of a dating violence prevention pro- gression: the influence of parents, peers, and personal experiences.
gram for middle school students. Violence Vict 1997;12(3):223 – 35. Fam Relat 1987;36:276 – 82.
[11] Malik S, Sorenson SB, Aneshensel CS. Community and dating vio- [30] Foshee VA, Bauman KE. Gender stereotyping and adolescent sexual
lence among adolescents: perpetration and victimization. J Adolesc behavior: a test of temporal order. J Appl Soc Psychol 1992;22:
Health 1997;21(5):291 – 302. 1561 – 79.
[12] Schwartz IL. Sexual violence against women: prevalence, conse- [31] Resnick MD, Blum RW. Development and personalogical correlates
quences, societal factors, and prevention. Am J Prev Med 1991;7(6): of adolescent sexual behavior and outcomes. Int J Adolesc Med
363 – 73. Health 1985;1(3 – 4):293 – 313.
[13] Jezl DR, Molidor C, Wright RL. Physical, sexual and psychological [32] Foshee VA, Bauman KE, Ennett ST, Linder GF, Benefield T,
abuse in high school dating relationships: prevalence rates and self- Suchindran C. Assessing the long-term effects of the Safe Dates
esteem issues. Child Adolesc Soc Work J 1996;13(1):69 – 87. program and a booster in preventing and reducing adolescent dating
[14] Magdol L, Moffitt TE, Caspi A, Newman DL, Fagan J, Silva PA. violence victimization and perpetration. American Journal of Public
Gender differences in partner violence in a birth cohort of 21-year-olds: Health (in press).
bridging the gap between clinical and epidemiological approaches. [33] Foshee VA, Langwick S. Safe Dates: an adolescent dating abuse
J Consult Clin Psychol 1997;65(1):68 – 78. prevention curriculum. Center City, MN: Hazelden Publishing and
[15] O’Keefe M. Predictors of dating violence among high school stu- Educational Services; 2004.
dents. J Interpers Violence 1997;12(4):546 – 68. [34] Check J, Malamuth NM. Sex role stereotyping and reactions to depic-
[16] Foshee VA, Linder F, MacDougall JE, Bangdiwala S. Gender differ- tions of stranger versus acquaintance rape. Pers Soc Psychol 1983;45:
ences in the longitudinal predictors of adolescent dating violence. 344 – 56.
Prev Med 2001;32(2):128 – 41. [35] Burt MR. Cultural myths and support for rape. Pers Soc Psychol
[17] Schwartz M, O’Leary SG, Kendziora KT. Dating aggression among 1980;38:217 – 30.
high school students. Violence Vict 1997;12(4):295 – 305. [36] Koss M, Leonard H, Beazley D, Oros C. Nonstranger sexual aggres-
[18] Brofenbrenner U. The ecology of human development. Cambridge, sion: a discriminant analysis of the psychological characteristics of
MA: Harvard Univ. Press; 1979. undetected offenders. Sex Roles 1985;12(9/10):981 – 92.
[19] McLeroy KR, Bibeau D, Steckler A, Glanz K. An ecological per- [37] Foshee V, Hollinger B. Maternal religiosity, adolescent social
spective on health promotion programs. Health Educ Q 1988;15(4): bonding, and adolescent alcohol use. J Early Adolesc 1996;16(4):
351 – 77. 451 – 67.

You might also like