J Fam Viol (2015) 30:579–590 DOI 10.1007/s10896-015-9689-7
ORIGINAL ARTICLE
Intimate Partner Violence and Homelessness as Mediators of the Effects of Cumulative Childhood Victimization Clusters on Adolescent Mothers’ Depression Symptoms Angie C. Kennedy & Deborah Bybee & Megan R. Greeson
Published online: 28 February 2015 # Springer Science+Business Media New York 2015
Abstract This study examines the relationships between cumulative victimization that began during childhood (witnessing intimate partner violence [IPV], physical abuse by a caregiver, and sexual victimization), adolescent IPV victimization, homelessness, and depression symptoms within a sample of 206 urban adolescent mothers. We used cluster analysis to identify five profiles of cumulative childhood victimization, and explored the extent to which adolescent IPV victimization and homelessness mediated the relationships between these clusters and participants’ current depression symptoms. Overall, we found that a significant portion of the effect of cumulative childhood victimization on current depression was explained by IPV victimization and homelessness, indicating that both are important mediators of childhood violence exposure within this high-risk sample of young women.
Keywords Cumulative victimization . Family violence . Sexual abuse . Pregnant and parenting adolescents . Mental health
An earlier version of this article was presented at the Society for Social Work and Research annual conference in January, 2014. A. C. Kennedy (*) School of Social Work, Michigan State University, 655 Auditorium Road, Baker Hall, Room 254, East Lansing, MI 48824, USA e-mail:
[email protected] D. Bybee Department of Psychology, Michigan State University, East Lansing, MI, USA M. R. Greeson Department of Psychology, DePaul University, Chicago, IL, USA
While researchers have studied individual types of family violence exposure among adolescents, including physical abuse by a caregiver, witnessing intimate partner violence (IPV), and sexual abuse, scholarly attention has recently turned to examining cumulative experiences with victimization. As the empirical evidence has accrued, it has become clear that different forms of family violence are likely to co-occur among youths. For example, Finkelhor and colleagues’ (2007) nationally representative study of cumulative victimization among children ages 2–17 found that, of those participants who had experienced any victimization within the past year, nearly threefourths (69 %) reported at least one additional concurrent type. In a similar study of adverse childhood experiences among over 8000 adult members of a health plan (the ACE study), co-occurring violence exposure was the norm: More than half (58 %) of those reporting witnessing IPV as children also experienced physical abuse, while 41 % of participants who had been sexually abused also reported physical abuse (Dong et al. 2004). Why are different forms of violence likely to co-occur among youths? Work in the area of cumulative adversity and stress points to the importance of social structural factors such as socioeconomic status (SES), race/ethnicity, gender, and age as heavily influencing a person’s vulnerability to victimization and other types of adversity such as chronic stressors and negative life events (Pearlin 1989; Pearlin et al. 2005; Turner and Lloyd 1995). Structural factors such as poverty or segregation into dangerous, poor communities can be powerful influences at the family level, shaping the family’s ability to provide safety, stability, and nurturance to its most vulnerable members (Turner et al. 2012). From this perspective, being poor, non-White, female, young, and urban confers heightened risk of exposure to multiple forms of violence which may occur concurrently as well as over time, as one type links to another in a chain of risks (Hatch 2005). As adversity and victimization experiences accumulate, they are
580
associated with a host of negative outcomes in adolescence, including childbearing, IPV victimization, homelessness, and mental health problems such as depression (Foshee et al. 2004; Haber and Toro 2004; Hillis et al. 2004; Kennedy et al. 2010; Turner et al. 2006). Indeed, pregnant and parenting adolescents appear to be disproportionately likely to report exposure to multiple forms of violence, with much higher rates of witnessing IPV, physical abuse by a caregiver, sexual abuse, and IPV victimization than those reported in national representative samples (Kennedy 2006; Klein 2005). Individual forms of victimization during youth are wellestablished as predictors of depression during adolescence and adulthood, particularly for female victims (Feiring et al. 2009; Fitzpatrick et al. 2005; McGee et al. 1997; Tandon and Solomon 2009; Turner et al. 2006). In part because they are likely to experience greater victimization and social disadvantage, adolescent mothers have higher rates of depression compared to both adolescents without children and adult women with children; using longitudinal data, researchers found that poverty in interaction with higher depressive symptoms predicted pregnancy, with increased depressive symptoms more likely to be sustained over time among adolescent childbearers (Mollborn and Morningstar 2009). Sustained depression during adolescence, in turn, has been linked to increased risk of major depression, suicidality, anxiety disorder, and educational underachievement by age 21 (Fergusson and Woodward 2002), adding further to the risks facing these young women as they transition to young adulthood. In studies examining adolescent mothers’ exposure to different types of victimization, experiencing sexual abuse and physical assault are especially strong predictors of depression symptoms (Mitchell et al. 2010; Osborne and Rhodes 2001). However, Finkelhor and colleagues’ (2007) work on youths’ violence exposure and mental health outcomes suggests that it is cumulative victimization, rather than experiences with individual types, that best predicts mental health problems. When Bpoly-victimization^ (defined as exposure to four or more types) was examined as a predictor in combination with single forms of violence exposure, they found that it either eliminated or greatly reduced the effects of individual types on depressive symptoms, suggesting the importance of examining violence exposure cumulatively. The relationship between cumulative victimization in childhood and subsequent IPV victimization in adolescence and adulthood has also been explored empirically. Using a nationally representative sample of adult women, researchers found that participants who reported both physical and sexual abuse during childhood were three times more likely to experience victimization by an intimate partner as an adult, compared to those with no exposure to childhood victimization (Desai et al. 2002). Work by Wolfe, Wekerle, et al. (1998, 2001) demonstrates a similar pattern for childhood victimization and later adolescent IPV victimization: A maltreatment history (a
J Fam Viol (2015) 30:579–590
combined measure of witnessing IPV and experiencing physical and sexual abuse) among girls was linked with increased vulnerability to partner victimization, mediated by feelings of interpersonal inadequacy and hostility in one study (Wolfe et al. 1998) and post-traumatic stress disorder in another (PTSD; Wekerle et al. 2001). The authors propose that childhood victimization leads to multiple interpersonal difficulties and traumatic effects, which in turn predict greater IPV victimization. In our recent work, we used clusters comprised of different forms of victimization (witnessing IPV, severe IPV effects such as seeing one’s mother injured, physical abuse in the family by a caregiver, and sexual victimization) to predict IPV victimization within a sample of female urban high school students. We found that the High Severe Family Violence cluster, made up of participants reporting high witnessed IPV and the severe effects of IPV, high physical abuse by a caregiver, and moderate sexual victimization, had a significantly higher rate of partner victimization in comparison to the Low All Violence cluster (Kennedy et al. 2012). Cumulative victimization within the family has been shown to be an important precursor to leaving home and experiencing homelessness among adolescents (Bao et al. 2000). Youths who are homeless have very high rates of prior family physical and sexual abuse: Ryan et al. (2000) found that over two-thirds of their sample of homeless adolescents had experienced victimization, with 30 % reporting both physical and sexual abuse. Participants who were exposed to both forms experienced their first abuse at a younger age as well as their latest abuse at an older age, and had been abused by more individuals. A recent study of homeless adolescents indicated even higher rates of cumulative victimization, with a majority (52 %) of the sample experiencing co-occurring physical abuse, sexual victimization, and emotional or psychological abuse; girls were much more likely to have been sexually victimized in comparison to boys (57 vs. 27 %; Taylor et al. 2008). Findings from a study of adolescent mothers who had experienced multiple forms of violence indicate that participants with a history of homelessness had higher rates of both witnessing IPV within the family and physical abuse by a caregiver, compared to participants with no history of homelessness (Kennedy 2007). Finally, partner violence victimization and homelessness have both been linked to adolescents’ depression symptoms. Teitelman et al. (2011) examined IPV victimization within a sample of urban adolescent women and found that 25 % reported experiencing concurrent threatening behavior, emotional abuse, and physical abuse, with higher rates of victimization associated with increased levels of depression symptoms. In two longitudinal studies of adolescent mothers’ IPV victimization and depression, higher or more severe IPV at baseline was associated with increased depression symptoms at 18 months (Sussex and Corcoran 2005) and at 4 and 8 years post-baseline (Lindhorst and Beadnell 2011), though in the
J Fam Viol (2015) 30:579–590
latter study these differences disappeared by 12 years postbaseline. Because these studies do not control for childhood cumulative victimization—which is linked to IPV victimization as detailed earlier—we are unable to disentangle the effects of earlier forms of victimization vs. partner violence on depression symptoms. With regards to the relationship between homelessness and depression, Whitbeck et al. (1999) developed and tested a risk amplification model that captures the paths to depression among homeless girls. For these youths, cumulative victimization within the family was positively associated with total time they had been on their own, affiliation with deviant peers, and victimization while on the street; street victimization, in turn, was the strongest predictor of depression symptoms. The authors argue that these cumulative negative experiences result in chains of risk that reinforce a negative sense of self and the untrustworthiness of others, thus contributing to depression and a sense of hopelessness (pp. 292–293). From this perspective, homelessness itself can be considered traumatic (Haber and Toro 2004). To summarize, researchers have begun to focus on cumulative victimization as the key to understanding poor mental health outcomes among adolescents, in contrast to examining one form of violence exposure in isolation. These cumulative experiences with victimization have been positively associated with depression, IPV victimization, and homelessness among youths; researchers typically examine the main effect of cumulative victimization on these outcomes by taking an additive approach (e.g., one type of victimization vs. two or three, with a higher number associated with poorer outcomes). Despite the contribution this approach has made, there is a need to explore different patterns or profiles of cumulative victimization so that we can better understand forms of victimization in combination with one another, and how these distinct patterns affect mental health outcomes. Additionally, it is important to go beyond direct effects models to examine mediating mechanisms that might link cumulative victimization to depression symptoms, particularly among high-risk populations such as adolescent mothers. In order to address these gaps, in the current study we explore the relationships between cumulative childhood victimization clusters and depression symptoms, as mediated by IPV victimization and a history of homelessness. To our knowledge this is the first time these variables have been examined in this way, despite evidence, summarized above, that suggests that IPV victimization and homelessness may play a key role in helping to explain the relationship between cumulative childhood victimization and depression symptoms among high-risk adolescent women. The following research questions were the focus of the study: 1) What are the associations between clusters of victimization that began by age 12 (witnessing IPV, physical abuse by a caregiver, and sexual victimization), IPV victimization, history of homelessness, and depression symptoms within this community sample of
581
adolescent mothers? 2) To what extent do IPV victimization and history of homelessness explain the relationship between patterns of cumulative victimization and symptoms of depression?
Method Research Design and Participants The study was a cross-sectional survey in which participants (N=206) completed a self-administered questionnaire that assessed their experiences with victimization during childhood and adolescence, history of homelessness, and mental health outcomes. In order to be eligible, participants were required to be between the ages of 16 and 21, be pregnant or have birthed a child, and read English. Recruitment took place at three urban sites in mid-Michigan: a prenatal clinic that serves low-income women operated by the county health department (69 %), a prenatal clinic that serves low-income women operated by a hospital (17 %), and a home-visiting program for adolescent mothers (14 %). At the first two sites, participants completed the informed consent process and the questionnaire at a private space at the clinic; at the third site, the home visitors passed out flyers to mothers, who then called one of the members of the research team to arrange a meeting (generally in participants’ homes) to complete the consent process and questionnaire. The university Institutional Review Board approved the study; participants were compensated $25. The sample (M age=19.34 years, SD=1.41) was predominately low-income, with nearly three-quarters (73 %) reporting Medicaid receipt within the past year. Half (50 %) of the participants were African American, with 28 % White, 15 % biracial or multi-ethnic, and 6 % Latina, reflecting the demographic profile of the recruitment site communities. Approximately 4 out of 10 (41 %) were living with family members, 34 % with a husband or male partner, and 22 % on their own. Most of the participants (72 %) were currently pregnant, with 56 % of the sample reporting they had at least one living child (M number of children=1.47, SD=.74, range of 1 to 5). Measures Childhood-Onset Victimization Variables Used for Clusters Witnessing IPV, Beginning Ages 0–12 Lifetime exposure to physical violence between adults within the home that began by age 12 was assessed using the physical assault sub-scale of the Revised Conflict Tactics Scale (CTS2; Straus et al. 1996). This sub-scale includes 12 items ranging from relatively less severe physical violence (e.g., seeing an adult slapped) to
582
relatively more severe (e.g., seeing an adult beaten up). Participants were asked if they had witnessed each of these 12 types, with yes coded 1, no coded 0; if participants endorsed at least one item, they were asked how old they were when they first witnessed IPV. For the current study, only witnessing that began by age 12 was included; a total score was derived by summing the 12 items. The CTS has been widely used to assess different forms of family violence among multiple ethnic/racial groups (Straus et al. 1996; Straus and Gelles 1990), including measuring children’s witnessing of IPV; it requires a 6th-grade reading ability to complete, making it appropriate for use with the adolescents participating in this study. The physical assault sub-scale was used by the first author in a previous study to assess witnessing IPV within a sample of adolescent mothers and demonstrated high internal consistency (Cronbach’s alpha=.91). For the 12-item witnessing IPV scale in the current study, alpha=.93. Physical Abuse by Caregiver, Beginning Ages 0– 12 Participants’ lifetime physical abuse by a parent or adult caregiver beginning by age 12 was assessed using the 12-item physical assault sub-scale of the CTS2 described above (Straus et al. 1996). Participants were asked if they had experienced each of the 12 types, with yes coded 1, no coded 0; if participants endorsed at least one item, they were asked how old they were when they first experienced it. Only physical abuse that began by age 12 was included; a total score was derived by summing the 12 items. The CTS has been widely used to measure familial physical abuse (Straus and Gelles 1990). The physical assault sub-scale was used in a previous study by the first author to assess physical abuse by a parent or adult caregiver within a sample of adolescent mothers, and demonstrated high internal consistency (alpha=.91). For the 12-item physical abuse by a caregiver scale in the current study, alpha=.92. Sexual Victimization, Beginning Ages 0–12 Participants’ lifetime sexual victimization beginning by age 12 was assessed using a modified version of Russell’s (1983) interview framework for child sexual abuse. The original framework consists of 20 yes or no items that tap various forms of unwanted, contact sexual victimization prior to the age of 18. The Russell framework has been widely used with adolescent and adult populations, including youths of color (e.g., Auslander et al. 2002). The modified version was made up of four items measuring different types of contact sexual victimization, including two items that assessed sexual victimization involving touching (e.g., BSomeone has made me touch their chest/breasts or genitals, or touch mine, when I didn’t want to^), one item on attempted rape (BSomeone has attempted to have sex with me [includes oral sex] when I didn’t want to^), and one item that assessed rape (BI have
J Fam Viol (2015) 30:579–590
had sex with someone [includes oral sex] when I didn’t want to, or because I was forced to^). Participants were asked if they had experienced each, with yes coded 1, no coded 0; if participants endorsed at least one item, they were asked how old they were when it first occurred. Only sexual victimization that began by age 12 was included; for the four items used in the current study, alpha=.84. For the current study, we computed a most severe sexual victimization variable in order to capture the severity of the sexual victimization experienced, with touching beginning by age 12 as most severe experienced coded as a 1 (reported by 2.4 % of the sample), attempted rape beginning by age 12 as most severe experienced coded as a 2 (reported by 12 %), and rape beginning by age 12 as most severe experienced coded as a 3 (reported by 7 %); this variable was used in the clusters profiles. Mediators IPV Victimization, Beginning Ages 13 and Up Participants’ physical victimization by an intimate partner beginning at age 13 or older was assessed using the 12-item physical assault sub-scale of the CTS2 described above (Straus et al. 1996). Participants were asked if they had experienced each of the 12 types, with yes coded 1, no coded 0; if participants endorsed at least one item, they were asked how old they were when they first experienced it. Only victimization that began at age 13 or older was included; a total score was derived by summing the 12 items. The CTS was developed to assess IPV behaviors and has been used extensively with a variety of populations. The physical assault sub-scale was used by the first author in a previous study to assess IPV victimization within a sample of adolescent mothers, and demonstrated high internal consistency (alpha=.92). For the 12-item IPV victimization scale in the current study, alpha=.91. History of Homelessness Participants were asked if they had ever spent the night sleeping or staying in a place that is not meant to be a home (e.g., a car or abandoned building or park) or staying with someone on their couch or floor, because they had nowhere else to go, with yes coded 1, no coded 0. Homelessness was operationalized this way to capture the unstable, frequently episodic housing patterns such as Bcouch surfing^ or Bhouse hopping^ that typify homelessness among adolescent girls and young women, particularly those who are parenting (Chicago Coalition for the Homeless 2011; Sanchez et al. 2006), and reflects the definition used by Ringwalt et al. (1998) in their national study of adolescent homelessness. Outcome Variable Current Depression Symptoms Participants’ current depression symptoms were assessed using the Center for Epidemiologic Studies Depression Scale (CES-D; Radloff
J Fam Viol (2015) 30:579–590
1977). The scale consists of 20 items that tap a range of feelings and behaviors participants might have experienced within the last 30 days (e.g., BI felt lonely,^ BI had crying spells^). It has strong test-retest reliability and internal consistency, and was able to discriminate between inpatient and general population samples, indicating good construct validity (Radloff 1977); it has been used widely to measure depressive symptomatology with a variety of populations, including adolescent mothers and other high-risk youths (Fitzpatrick et al. 2005; Mitchell et al. 2010). Response options were yes (coded 1) or no (coded 0); a total score was derived by summing the 20 items. Dichotomized response options on the CES-D have been used in previous research with participants in health care settings such as primary care clinics, and have demonstrated comparable internal consistency and validity (Irwin et al. 1999; Cheng and Chan 2005; Robison et al. 2002); however, a clinical cutoff score is not available for the dichotomized scale. In the current study, alpha=.85. Analytic Strategy We standardized the variables to minimize differential weighting due to different scales prior to cluster analysis. We used a two-stage approach to group cases into clusters, using the 206 cases with complete data on the three clusterdefining variables (witnessing IPV, physical abuse by a caregiver, and sexual victimization). First, we used Ward’s method of hierarchical agglomerative cluster analysis on squared Euclidean distances; inspection of the sequential changes in fusion coefficients suggested a five-cluster solution. Second, starting with the centroids from the initial solution, we used Kmeans cluster analysis to assign cases to the nearest cluster, yielding the smallest possible within-cluster variances. To examine the relationships between cumulative victimization clusters and depression symptoms, and determine to what extent IPV victimization and homelessness mediated those relationships, we used Mplus 6 (Muthén and Muthén 2010) to conduct path analysis and estimate the indirect, direct, and total effects. To accommodate the presence of both continuous and dichotomous mediating variables, we used robust weighted least squares estimates; we used bias-corrected bootstrap tests to assess the significance of indirect effects (Preacher and Hayes 2008). We dummy coded the cumulative victimization clusters; the LoAll cluster (low exposure to all three types of childhood victimization) was the reference category.
Results Descriptives and Bivariate Correlations Overall, lifetime victimization beginning in childhood was common, with nearly three-quarters (71.8 %) witnessing
583
IPV, over half (56.5 %) reporting physical abuse by a caregiver, and just over one in five (21.1 %) experiencing sexual victimization, all beginning before the age of 13. Over onethird (36.1 %) reported adolescent IPV victimization, and 40.1 % had experienced at least one night of homelessness. The average score on the CES-D was 9.90 (SD=4.17). The childhood victimization variables were positively associated with one another, with rs ranging from .21 to .58; IPV victimization, history of homelessness, and depression symptoms were positively correlated with each other, and with each of the childhood victimization variables (see Table 1 for descriptives and intercorrelations among primary variables). Cumulative Childhood Victimization Clusters Twenty-four participants (11.6 % of the total sample) made up the first cluster, which was characterized by high levels of exposure across all three forms of victimization beginning by age 12: high witnessing IPV, medium-high physical abuse by a caregiver, and high sexual victimization. We termed this cluster HiAll. The second cluster was comprised of 31 participants (15.0 % of the sample) with high levels of exposure to both witnessing IPV and physical abuse by a caregiver and very low sexual victimization; this was the HiFV cluster. The third cluster was made up of 50 participants (24.3 % of the sample) with high witnessing IPV, low physical abuse by a caregiver, and almost no sexual victimization; we termed this the HiWIPV cluster. The fourth cluster was comprised of 15 participants (7.3 % of the sample) with high sexual victimization, medium witnessing IPV, and low physical abuse by a caregiver; this was the HiSV cluster. Eighty-six participants (41.7 % of the sample) made up the fifth cluster, characterized by low exposure to witnessing IPV, physical abuse by a caregiver, and sexual victimization; this cluster was termed LoAll. See Fig. 1 for rates of different forms of childhood victimization by cluster. We examined cluster membership in relation to demographic variables, the mediators (IPV victimization and homelessness), and depression symptoms. There were no significant differences by cluster on age, race/ethnicity, age at first pregnancy, pregnant vs. parenting, current residence, or Medicaid receipt, nor were any of these demographic variables associated with depression symptoms. There were significant differences by cluster in mean level of IPV victimization, F(4, 204)=12.22, p =.000, prevalence of history of homelessness, χ2(4, N=206)=18.09, p=.001, and mean level of depression symptoms, F(4, 202) = 9.37, p = .000 (see Table 2). In general, HiAll and HiFV had significantly higher levels of IPV victimization in comparison to the other three clusters, while HiAll had a significantly higher rate of history of homelessness in comparison to the LoAll cluster. Finally, LoAll had a significantly lower level of depression symptoms in comparison to the other four clusters.
584 Table 1
J Fam Viol (2015) 30:579–590 Descriptives and intercorrelations among victimization variables used for clusters, mediators, and depression
Variable
Mean
SD
WIPV
PAC
SxVict
IPVvict
HxHmlss
Dep
WIPV PAC SxVict IPVvict HxHmlss Dep
5.08 2.97 .48 1.81 .40 9.90
4.16 3.59 .96 2.99 .49 4.17
1.00 .58*** .21** .35*** .18* .35***
1.00 .27*** .50*** .19** .34***
1.00 .14* .28*** .20**
1.00 .20** .32***
1.00 .31***
1.00
WIPV Witnessing intimate partner violence, PAC Physical abuse by a caregiver, SxVict Sexual victimization, IPVvict IPV victimization, HxHmlss History of homelessness, Dep Depression symptoms *p
