Depressive Symptomatology as a Predictor of Exposure to Intimate Partner Violence Among US Female Adolescents and Young Adults

Objective  To examine whether depressive symptomatology is predictive of subsequent intimate partner violence victimization among a national prospective sample of female adolescents and young adults.

Design  Home interview data from 2 waves of the school-based National Longitudinal Study of Adolescent Health were used to examine whether baseline depressive symptomatology was associated with increased risk of past-year exposure to physical abuse by a current partner at 5-year follow-up, controlling for age, race/ethnicity, parental education, retrospective childhood physical and sexual abuse, and baseline dating violence and forced sex.

Setting  A stratified random sample of 80 US high schools and 52 middle schools.

Participants  All young women (n = 1659) were in a current opposite-sex relationship at follow-up.

Main Exposure  Baseline past-week depressive symptomatology, measured as both a dichotomous and continuous variable.

Main Outcome Measures  Past-year exposure to mild and moderate to severe physical abuse by a current intimate partner.

Results  In adjusted models with dichotomous depressive symptoms, high baseline symptom levels were associated with 1.86 times the odds of subsequent exposure to moderate to severe partner violence (95% confidence interval, 1.05-3.29). In adjusted models with continuous depressive symptoms, an increase of 1 SD in baseline symptom levels was associated with a 24% increase in odds of exposure to mild partner violence and a 24% increase in the odds of exposure to moderate to severe partner violence.

Conclusions  The findings of this study suggest that depressive symptomatology among girls during adolescence is associated with increased risk of subsequent exposure to physical partner violence. Prevention, identification, and treatment of depressive symptomatology among adolescent girls may help to reduce the likelihood of subsequent victimization.

While depression is understood to be a common outcome of exposure to intimate partner violence (IPV) among female adolescents and young adults,¹ little research has examined the reverse relationship—the extent to which depressive symptomatology in adolescence may be associated with risk of exposure to future abuse by an intimate partner, independent of factors such as prior dating violence or childhood abuse. Increased understanding of the depression-IPV relationship may contribute to prevention efforts by helping to identify girls at risk of future IPV exposure.

Lifetime prevalence of major depressive disorder and minor depression among girls aged 15 to 24 years have been estimated to be 21% and 10%, respectively.² The time of transition from adolescence to adulthood is a time of peak depression onset.³ At the same time, 8.8% of female high school students in the national 2003 Youth Risk Behavior Surveillance System survey reported past-year experience of physical abuse by a dating partner.⁴ Estimates of dating violence victimization among girls in smaller studies range from 8% to 57%; studies vary in the period assessed.⁵ Nationally, 22% of women 18 years and older report experiencing physical IPV at some point in their lifetime.⁶

Depressive symptomatology, dysthymia, and major depression during adolescence have been found to be associated with a broad range of functional impairments and adverse outcomes in later adolescence and young adulthood, including diminished self-esteem, self-mastery and life satisfaction, internalizing behavior problems, interpersonal difficulties, increased need for social support, career dissatisfaction,⁷ divorce,^8-11 impaired work and social functioning, lower social class, increased work absence, unemployment,¹² diminished educational attainment,^9,11,12 female teenage childbearing,¹⁰ and decreased likelihood of on-time marriage.¹³ In addition to predicting adverse social outcomes, diagnosed depression in adolescence is associated with subsequent substance abuse¹⁴ and is a strong and specific predictor of depression in early adulthood^7,12,15-18 as well as suicide and suicide attempts.^12,19 One study found that depression at age 18 years was associated with impaired psychological and social functioning at age 21 years regardless of depression status at age 21 years, suggesting that depression during the time of transition to adulthood is associated with continued psychosocial vulnerability and poorer functioning in young adulthood regardless of depression status in young adulthood¹⁹; other studies support the finding of vulnerability beyond the duration of the depressive episode.¹⁴

Few prospective studies have examined IPV exposure as a potential outcome related to adolescent depressive symptomatology, and findings among these studies have been mixed. A 3-year study of a high-risk sample found that women's baseline depressive symptomatology was predictive of their male partners' subsequent perpetration of psychological but not physical abuse.²⁰ A 6-year study of male and female high school students found that depressive symptomatology in adolescence was not associated with subsequent physical aggression in dating relationships or marriage (perpetration and victimization were not distinguished).²¹ A retrospective national study of married or cohabiting adults found that premarital depression was not associated with subsequent IPV victimization among women.²² A 5-year study of female high school students found that the proportion of days spent in depressive episodes was positively associated with psychological and physical abuse by boyfriends over the follow-up period; temporal order of depressive episodes and abuse could not be ascertained.¹⁶

Given the high prevalence of depressive symptomatology and IPV exposure among female youth in the United States, as well as research demonstrating many adverse psychosocial outcomes for adolescents with depressive symptomatology, this study examined whether depressive symptomatology among adolescent girls is predictive of subsequent IPV exposure in late adolescence and young adulthood. Analyses were conducted with prospective, population-based data from the National Longitudinal Study of Adolescent Health (Add Health).²³

The Add Health study is a longitudinal study of the determinants and contexts of youth health and risk behaviors. In summary, a stratified random sample of 80 US high schools and 52 middle schools was selected with probability of selection proportional to school size. Stratification was with respect to region of country, urbanicity, school size, school type, and race/ethnic composition. Wave 1 structured, closed-ended home interviews were conducted with youth in grades 7 through 12 in 1995; 82% of wave 1 home interview subjects with sampling weights had a parental figure who participated in a wave 1 parent interview. Wave 2 home interview data were collected in 1996 (n with sampling weights = 13 570) and wave 3 data were collected in 2001 and 2002 (n with sampling weights = 10 828). Written informed consent was provided by youth participants as well as a parental figure at waves 1 and 2 and by youth participants at wave 3. Approval for the full Add Health study was granted by the institutional review board of the University of North Carolina at Chapel Hill School of Public Health; approval for the present data analysis was granted by the Human Subjects Committee of the Harvard School of Public Health (Boston, Mass).

The study sample was restricted to the 1659 girls in the contractual use data set²⁴ who participated in home interviews at waves 1, 2, and 3; had sample weights; and were selected for the couples sample of the Add Health study. One half of all wave 3–eligible subjects were randomly selected to be considered for the wave 3 couples sample. By design, the couples sample includes one-third dating couples, one-third cohabiting couples, and one-third married couples. Criteria for inclusion in the couples sample were (1) involvement in a current relationship; (2) relationship duration of 3 months or more; (3) opposite-sex relationship; and (4) partner 18 years or older. The sample for the present study was also restricted to subjects who had nonmissing data regarding IPV outcomes and who had 3 or fewer missing items on the wave 2 depressive symptoms scale. Three subjects who reported that their partner had physically abused them more than 1 year prior to the wave 3 interview, but not within the past year, were also excluded so as to maintain temporal clarity between wave 2 depressive symptoms and IPV exposure (Figure).

Self-reported past-week depressive symptomatology was assessed both as a dichotomous and continuous variable at wave 2, with a 19-item, modified Center for Epidemiologic Studies Scale (CES-D) used in the Add Health survey.²⁵ The original 20-item scale has been widely used as a measure of depressive symptomatology in epidemiologic research and as an initial screening tool for clinical depression in community-based samples. In the modified Add Health scale, 2 original CES-D items were removed (“I had crying spells” and “My sleep was restless”) and 1 item was added (“I felt that life was not worth living”). Two additional items were rephrased. Item scores in the modified scale correspond to symptom frequency and range from 0 (“never or rarely”) to 3 (“most or all of the time”); scale scores range from 0 to 56. Internal consistency reliability in the present sample was 0.89, similar to that found for the CES-D in other adolescent samples.^26-28

A CES-D cutoff score of 24 has been found to maximize sensitivity (0.84) and specificity (0.75) for Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition²⁹ major depressive disorder or dysthymia for adolescent girls.²⁷ Proportionally adapted to the 19-item scale, the cutoff score used in this study for above-threshold or “high” depressive symptomatology was 23.^30,31 Secondary analyses were also performed with the continuous scale for greater statistical power.

Age. This was a continuous variable for age at wave 2.

Race/Ethnicity. Four categories were used: non-Hispanic white, non-Hispanic black, Hispanic, and other. Youth reporting more than 1 race/ethnicity were assigned to the category with which they indicated they most identify.

Wave 1 Parental Education. This was a dichotomous variable indicating whether the highest educational attainment among resident parental figures at wave 1 was less than high school. Data were obtained from the parent interview.

Childhood Experience of Physical and/or Sexual Abuse. At wave 3, subjects were asked 2 self-administered questions regarding the time before they entered the sixth grade: (1) “How often had your parents or other adult caregivers slapped, hit, or kicked you?” and (2) “How often had one of your parents or other adult caregivers touched you in a sexual way, forced you to touch him or her in a sexual way, or forced you to have sexual relations?” A dichotomous variable was created reflecting subject report of being slapped, hit, or kicked more than 5 times and/or report of any occurrence of sexual abuse.

Dating Violence and/or Forced Sex Between Waves 1 and 2. At wave 2, subjects were asked whether any partner they had had since wave 1 ever threatened them with violence, pushed or shoved them, or thrown something at them that could hurt. Subjects who indicated at wave 2 that they had ever had sexual intercourse were also asked whether they were ever physically forced to have sexual intercourse against their will since the wave 1 interview. These data were collected with audio computer-assisted self-interview.^32,33 A dichotomous variable was created reflecting 1 or more affirmative responses to these items.

Mild Violence. At wave 3, subjects were asked 1 self-administered question regarding how often in the past year their current partner had threatened them with violence, pushed or shoved them, or thrown something at them that could hurt. A dichotomous variable was created reflecting any past-year experience of this outcome.

Moderate to Severe Physical Violence. At wave 3, subjects were asked 1 self-administered question regarding how often in the past year their current partner had slapped, hit, or kicked them and another question regarding how often in the past year they had had an “injury, such as a sprain, bruise, or cut” because of a fight with their current partner. A dichotomous variable was created reflecting experience of either or both of these outcomes.

All analyses were conducted with SAS version 8 (SAS Institute Inc, Cary, NC). Logistic regression models were estimated using the 2 dichotomous violence variables as dependent variables. The SAS proc GENMOD function (generalized estimating equations) was used for all regressions, with specification of an independent working correlation structure. The use of generalized estimating equations accounts for data clustering to provide robust standard error estimation.

Multivariate models first adjusted for control variables of age, race/ethnicity, and parental education (model 1) and then additionally adjusted for covariates of childhood physical/sexual abuse and dating violence/forced sex (model 2). All baseline data for these analyses were collected at wave 2, except for wave 1 parental education and wave 3 retrospective childhood abuse.

Potential interactions between depressive symptomatology and each of the 2 prior abuse covariates were examined separately in unadjusted and adjusted models predicting mild and moderate to severe abuse outcomes, with both the dichotomous and continuous depressive symptoms variables. These interaction terms were insignificant in all model specifications and were not included in final models.

In addition, 2 variables reflecting wave 2 alcohol abuse comorbidity (frequent drinking and heavy drinking) were considered for inclusion as covariates; neither variable was included in final models because they were not associated in bivariate or multivariate models with either IPV outcome, and their addition to the adjusted models did not alter the effect size or significance of depressive symptoms. This may have been owing to insufficient statistical power.

Variables reflecting subjects' relationship status were also not included in final models because relationship status was viewed as a potential mediator between depressive symptoms and IPV exposure. Inclusion of relationship status variables in preliminary analyses, however, produced no difference in the effect size or significance level of depressive symptomatology with regard to either IPV outcome.

All analyses used adjusted grand sample weights to account for wave 3 nonresponse and subgroup oversampling. For analyses examining the outcome of mild IPV, a subsample of 1632 subjects was used; 27 subjects who reported exposure to moderate to severe but not mild violence at wave 3 were excluded. For analyses examining moderate to severe IPV, a subsample of 1525 subjects was used; 134 subjects who experienced mild but not moderate to severe violence were excluded. For the 15 subjects who were missing 3 or fewer items on the wave 2 depressive symptoms scale, the mean of the remaining items was imputed for each missing item.

The sample of girls was 68.5% non-Hispanic white, 15.6% non-Hispanic black, 10.6% Hispanic, and 5.3% other race/ethnicities (Table 1). The mean ages at baseline and follow-up were 15.9 years and 21.3 years, respectively; the wave 3 age range was 18 to 27 years. Ten percent of subjects came from homes where the highest parental education level was less than 12th grade. The mean baseline depressive symptoms scale score was 11.4, and 10.2% of subjects had above-threshold symptom scores. At wave 3, 25.7% of subjects were married, 24.9% were cohabiting, and 49.4% were considered to be dating. The mean number of years between wave 2 and 3 interviews was 5.4. One percent of subjects were married at both waves 2 and 3.

Twenty-eight percent of subjects with high depressive symptom levels at baseline reported any IPV or injury at follow-up as compared with 17.5% of subjects with lower symptom levels (odds ratio [OR], 1.84 [95% confidence interval (CI), 1.20-2.84]) (Table 2).

In bivariate analyses with the dichotomous depressive symptoms variable, girls with high baseline symptom levels had 1.66 times the odds of subsequent exposure to mild IPV and 2.46 times the odds of exposure to moderate to severe IPV as compared with girls with lower baseline symptom levels (results not shown). In fully adjusted analyses, high depressive symptomatology was not significantly associated with likelihood of exposure to mild IPV (P = .11) (Table 3); girls with high depressive symptoms had 1.86 times the odds of subsequent exposure to moderate to severe IPV (95% CI, 1.05-3.29) (Table 4).

In bivariate analyses with continuous depressive symptomatology, a 1-point increase in baseline symptoms was associated with a 3% increase in the odds of subsequent exposure to mild IPV (95% CI, 1.01-1.05) and a 4% increase in the odds of exposure to moderate to severe IPV (95% CI, 1.02-1.06). In fully adjusted analyses, each 1-point increase in depressive symptomatology was associated with a 3% increase in the odds of exposure to mild IPV (95% CI, 1.01-1.05) and a 3% increase in the odds of exposure to moderate to severe IPV (95% CI, 1.01-1.05) (tables available on request).

Regarding covariates and controls, adjusted analyses demonstrated that childhood physical/sexual abuse and baseline dating violence/forced sex were each associated with approximately 2 times the odds of exposure to moderate to severe IPV (Table 4). Additional analyses found that the lack of association observed between the abuse covariates and mild IPV in the fully adjusted model did not seem to be due to a mediating effect of depressive symptomatology.³⁴

Findings from this prospective study suggest that depressive symptomatology among adolescent girls is associated with increased risk of subsequent physical abuse by an intimate partner in late adolescence and young adulthood, independent of childhood physical or sexual abuse as well as prior experience of dating violence or forced sex. Twenty-eight percent of girls with high levels of baseline depressive symptomatology, corresponding to an elevated likelihood of mood disorder, reported some form of past-year physical IPV victimization 5 years later. Girls with high baseline depressive symptoms had 1.86 times the odds of girls with lower symptom levels of subsequent exposure to moderate to severe IPV, and victimization risk was also found to increase with increasing levels of continuous depressive symptomatology. The finding that childhood abuse and dating violence/forced sex were associated with subsequent IPV risk also provides prospective evidence that prior abuse is predictive of revictimization among young women.

The time of transition from adolescence to adulthood is one of increasing independence from parents, strengthening self-concept, and critical decision making regarding education, occupation, and intimate relationships, all of which have a salient impact on the long-term social trajectory of the individual.^16,19 Depression that occurs during this time of transition may have a particularly significant impact on maturational processes and choices made, generating a trajectory of risk for continued psychosocial impairment and psychopathology.^{11,15,16,35,36}

Regarding social pathways that may be affected by depression in adolescence, studies of peer group homophily and assortative mating have found that adolescents and adults with depressive symptoms or disorders often associate, in friendship,^37,38 dating,²⁰ and marriage,³⁹ with others who have similar levels of symptomatology. In turn, depressive symptoms and disorders among men have been found to be positively associated with subsequent psychological²⁰ and/or physical²² aggression toward an intimate partner.

Depressive symptomatology is also associated with a range of adolescent risk behaviors including substance use and abuse,^40-42 and depressed youth are more likely to associate with deviant or risky peer groups¹⁸ from which partner selection is likely to occur. Alcohol and drug use, in turn, have been associated in some studies with males' physical abuse perpetration toward their dating partners,⁴³ cohabiting partners, and wives.⁴⁴ Depressive symptomatology has also been associated with antisocial behavior among young men^20,45; delinquency⁴⁶ and antisocial behavior^20,21 among male adolescents are associated with increased likelihood of physical IPV perpetration. In sum, depressive symptomatology may elevate girls' risk of future IPV exposure by way of its contribution to risky peer group affiliations and assortative partnering with high-risk mates.

In addition to differential selection into abusive relationships, young women with a history of depressive symptomatology may be less likely to leave abusive relationships (or take a longer time to leave) for various reasons. First, insofar as depression in adolescence is predictive of outcomes such as teenage childbearing, diminished educational attainment, and unemployment, it may be associated with increased economic dependency on future partners and thus decreased likelihood of relationship exit.^47,48 Second, increasing social isolation of the woman from family and friends is a common feature of abusive relationships, and diminished social support is another obstacle to relationship exit.^49,50 Depressed youth often have impaired social relationships and diminished support from family members and peers,^11,35,51 such that young women with former or current depression may enter intimate relationships from a starting point of greater isolation than those with no former or current depression. This diminished potential pool of external support may contribute to an increased likelihood of not leaving, or returning to, an abusive partner. Girls with a history of depression and diminished social support may also place an elevated premium on intimacy and maintenance of a relationship with a romantic partner and be more willing to tolerate acts of violence from an otherwise seemingly secure partner.⁴⁸

Third, depression is a primary psychological response to abuse, and Barnett⁴⁸ notes that depression or other mental health difficulties due to IPV may impair women's ability to mobilize to exit their relationships. Women who have a history of vulnerability to depression may have an increased likelihood of responding to the stressor of IPV with depression or other debilitation, contributing to a decreased likelihood of relationship exit. Further research should examine these and other potential mechanisms.

This study has several limitations. First, the adapted CES-D scale is not a diagnostic instrument. The wave 2 measurement of past-week depressive symptomatology may also not fully reflect the severity, chronicity, or recurrence of symptomatology over the period between waves 2 and 3, which may be relevant to likelihood of future IPV exposure. However, depressive symptoms and disorders in adolescence are predictive of subsequent disorder in young adulthood, and a number of studies have noted the stability of depressed mood among adolescents, although a consistent pattern of findings has not yet emerged for girls.^11,52-54 Second, incomplete measurement of confounding factors may have contributed to the observation of an association between depressive symptomatology and IPV exposure. The childhood abuse, dating violence, and sexual coercion covariates do not account for all relevant time in respondents' lives, and the wave 2 sexual coercion item focuses only on forced intercourse. The Add Health study also does not include a measure of exposure to interparental violence, which has been found to be predictive of women's subsequent depressive symptomatology⁵⁵ and abuse by an intimate partner.⁵⁶ Third, IPV is frequently underreported by survivors,^57,58 which may have led to bias in results. Fourth, this study did not address perpetration; this relationship should be examined with surveys that distinguish between violence committed in self-defense or retaliation and self-initiated acts.⁵⁹ Finally, the generalizability of the study findings is limited, given the specific nature of the sample.

Overall, this study provides evidence that elevated levels of depressive symptomatology among adolescent girls are associated with an increased vulnerability to subsequent IPV exposure in late adolescence and young adulthood. However, the question of whether depressive symptomatology plays an independent role in shaping the trajectory toward future IPV exposure among young women, or whether it is simply a marker for IPV risk, is not fully resolved with the present study and warrants further investigation. If depressive symptomatology among adolescent girls becomes recognized as an independent predictor of subsequent IPV exposure, this will be one more reason to increase efforts in the prevention, identification, and treatment of depressive symptomatology among adolescent girls. If depressive symptomatology comes to be understood as a marker for risk of IPV exposure, it may be helpful in identifying high-risk girls for targeted intervention efforts.

Correspondence: Jocelyn A. Lehrer, ScD, 4601 25th St, 7, San Francisco, CA 94114 (jlehrer1@gmail.com).

Accepted for Publication: September 21, 2005.

Funding/Support: This research uses data from Add Health, a program project designed by J. Richard Udry, PhD, Peter S. Bearman, PhD, and Kathleen Mullan Harris, PhD, and funded by grant P01-HD31921 from the National Institute of Child Health and Human Development, Rockville, Md, with cooperative funding from 17 other agencies. Partial funding for research came from the National Institute of Mental Health Academic Fellowship on Mental Health and Developmental Disorders.

Role of the Sponsor: The funders were not involved in the design, conduct, analysis, or interpretation of the data or in preparation, review, or approval of the manuscript.

Additional Information: Persons interested in obtaining data files from Add Health should contact Add Health, Carolina Population Center, 123 W Franklin St, Chapel Hill, NC 27516-2524 (addhealth@unc.edu).

Acknowledgment: Special acknowledgment is due Ronald R. Rindfuss, PhD, and Barbara Entwisle, PhD, for assistance in the Add Health Study original design.