- Email: [email protected]
Partner violence among homeless young adults: measurement issues and associations
JOURNAL OF ADOLESCENT HEALTH 2002;30:355–363
ORIGINAL ARTICLE
Partner Violence Among Homeless Young Adults: Measurement Issues and Associations NEIL W. BORIS, M.D., SHERRYL SCOTT HELLER, Ph.D., TONYA SHEPERD, M.D., AND CHARLES H. ZEANAH, M.D.
Purpose: The primary goal of this study was to test the reliability of the Partner Violence Interview and examine validity by measuring differential correlates of partner violence. Methods: Sixty young adults (30 males and 30 females) housed in an urban shelter participated in this study. All participants were between the ages of 18 and 21 years and the majority were African-American. The participants were administered two measures of partner violence exposure, one measure of community violence exposure and one measure of depression. A random selection of 30 of the participants was retested after 1 month. Results: As predicted, current and past partner violence was common in this sample, with over 70% endorsing a history of physical violence. The Partner Violence Interview (PVI) had adequate retest reliability (Pearson r for two PVI scales ⴝ .7 and .85) and internal consistency (KR-20 for each scale ⴝ .78 to .93). Preliminary evidence of convergent validity was suggested by the fact that the PVI lifetime partner violence scale was significantly correlated with a physical violence scale from a second measure (the Conflict Tactics Scale; r ⴝ .596, p < .001). Violence in past relationships, as opposed to current relationships, was associated with both lifetime community violence exposure and current level of depression. Conclusions: The Partner Violence Interview is a reliable, comprehensive instrument suited to high-risk pop-
From the Departments of Psychiatry and Neurology, Tulane University School of Medicine (N.W.B., S.S.H., C.H.Z.) and Community Health Sciences (N.W.B.), Tulane University School of Public Health and Tropical Medicine, New Orleans, Louisiana; and Department of Obstetrics and Gynecology, Louisiana State University School of Medicine, New Orleans, Louisiana (T.S.). Address correspondence to: Neil W. Boris, M.D., Tulane University School of Public Health and Tropical Medicine, Suite 2301, Box SL-29, 1440 Canal Street, New Orleans, Louisiana 70112. E-mail: [email protected]. Manuscript accepted September 1, 2001.
ulations. Homeless young adults commonly experience severe partner violence, and preventive intervention is clearly indicated for this group. © Society for Adolescent Medicine, 2002 KEY WORDS:
Adolescents Homeless Partner violence Reliability Validity Young adults
Epidemiologic data on violent crime in the United States suggest that by the mid-1990s adolescents were exposed to the highest overall rates of lethal and nonlethal violence of any age group. Furthermore, rates of violence exposure are rising for adolescents faster than for any other age bracket [1]. Urban, minority adolescents and young adults, in particular, are likely to have been exposed to high levels of community violence [2]. Surveys of the effects of community violence in adolescence have established associations with serious psychological sequelae [3,4]. However, the relationships between varying types of violence exposure in adolescence, such as maltreatment by parents and violence in intimate partner relationships, are less well studied [5]. Furthermore, the complex processes that convert some children who have been victimized into adolescents and adults who subsequently victimize others have not yet been established, despite the considerable public health implications [5– 8]. Homeless young adults are a subpopulation
© Society for Adolescent Medicine, 2002 Published by Elsevier Science Inc., 655 Avenue of the Americas, New York, NY 10010
1054-139X/02/$–see front matter PII S1054-139X(01)00410-4
356
BORIS ET AL.
known to be particularly likely both to have been victimized and to be victimizers. Most of the available data on this subgroup of young adults come from urban drop-in centers or shelters set up to serve this population. The fact that many of these centers are located in or near neighborhoods with high rates of violence may account for the extremely high rates of exposure to community violence reported by homeless adolescents and young adults [9]. For many homeless young adults, the violence associated with living on the streets or in shelters is consistent with previous violence exposure. For instance, interviews with parents or caretakers of homeless young adults who subsequently ran away have validated teens’ reports of high levels of family violence and abuse in the home [10]. Furthermore, a careful retrospective review found that child physical abuse is a potent predictor of adult homelessness [11], although it is as yet unclear whether child abuse, considered longitudinally with potential moderating variables, is a causative factor for homelessness at any age. On the other hand, it is clear that homeless young adults may also be perpetrators of violence; high rates of aggression and conduct problems have been documented in several cross-sectional studies [10,12,13]. Furthermore, a large systematic study of homeless adults revealed that many had long histories of victimizing aggressive behaviors, often beginning before they became homeless. In this sample, victimization also was very common and rates of posttraumatic symptoms were high [14]. Finally, a longitudinal study of runaway youth seen at a child guidance clinic revealed eventual rates of antisocial personality disorder that were higher than rates for adolescents from the clinic who were not runaways [15]. There is a clear link between homelessness in adolescence and violent behavior. It is not yet clear how frequently the cycle of violence impacting homeless adolescents and young adults plays out in intimate relationships. In fact, there is little in the literature documenting the degree of violence in intimate relationships among homeless young adults, although the problem of dating violence spans the adolescent years and is generally common [8,16]. It is clear that the rate of risky sexual behavior among homeless adolescents is high [17] as are pregnancy rates [18]. It is also clear that rates of substance abuse is high among homeless adolescents and young adults [19] and that substance abuse is associated with partner violence [20,21], although the relationship between substance abuse and partner violence among young adults appears complex [22].
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
Measurement Issues Measuring exposure to, and experience of, violence in any group is a complex task. Several issues are of particular importance in approaching partner violence in subgroups such as homeless young adults. There is no accepted definition of what particular experiences constitute a case of partner violence. In fact, measures that rely on an individual’s subjective experience may not correlate highly with an objective judge’s rating of violence reports [23]. For instance, even women admitted to shelters for battered women may fail to meet differing cutoffs for physical abuse [24]. Furthermore, global ratings focused on presence or absence of violent experiences may not capture significant events that are nonetheless traumatizing. Finally, the relationship among different scales or subscales focusing on varying types of violent experiences may not be related. Another problem is that most measures focus exclusively on victimization rather than on mixed perpetration and experience of violence. There is evidence that for high-risk adolescents and young adults in particular, bidirectional violence may be common [25], and it is only more recent measures that assess this phenomenon. There is a longstanding question as to whether anonymous questionnaires allow for more forthright responses on the part of subjects who may be embarrassed or uncomfortable in reporting on their experiences. Although this question is not easily resolved, potential problems with reading comprehension among homeless adolescents make an interview format especially important for gathering reliable data in this and comparable risk groups. One factor related to the limited data on partner violence in high-risk populations may be the lack of a validated and standardized measure of partner violence suitable for use with this high-risk population [23]. An important, although understudied issue regarding partner violence among young adults has to do with gender differences in perpetration. Although there is considerable evidence among adults that partner violence is primarily a problem of coercive control of women by men, there appears to be more heterogeneity in adolescence [26]. A key factor is the reliability of the data because getting solid data on perpetration may be difficult. Similarly, the potential association among depression, partner violence, and homelessness in adult women has not been comprehensively studied in younger homeless populations. It remains unclear how commonly men are the victims of violence in
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
intimate relationships and whether the association between violence and homelessness in adult women, whether moderated by depression or not, applies to adolescent or young adult males or females. Furthermore, it is unknown whether scales that capture verbal aggression and reasoning or subjective selfratings of violence give a different perspective than scales that simply capture whether certain events have ever happened [27,28]. The data presented in this paper are from a larger study of the relationship among attachment history, childhood experiences, and a number of variables on the development of intimate relationships in homeless adolescents in which a simple 26-item measure called the Partner Violence Interview (PVI) was pilot-tested [29]. One primary goal in this study was to test whether the PVI would serve as a reliable and valid measure of violence exposure in intimate relationships in a homeless adolescent population. Further, we hypothesized that exposure to community violence would correlate with exposure to partner violence. Finally, we hypothesized that males would score significantly higher than females on the partner violence scales, despite existing evidence of patterns of bidirectional violence in adolescence.
Methods Subjects Sixty homeless young adults were recruited over a 3-month period from a private nonprofit shelter/ transitional housing facility in a medium-sized southern city. Demographic data on the sample are summarized in Table 1. Although the shelter housed some teens under 18 years, only 18- to 21-year-olds were recruited for this study because of the need to seek parental permission for those under 18 years. Consequently, the mean age for this sample was 19 years. At this particular shelter, most of the adolescents and young adults were housed in an “acute unit” in which they were allowed to stay for 60 days. Although we did not systematically gather data on how long each subject had been homeless, the majority of the sample was recruited from this unit. As noted in Table 1, most of the sample was African-American, single, and, at the time of the study, had failed to complete high school. Of the sample, 42% had at least one child. Males and females did not differ significantly on any of the demographic variables.
357
Table 1. Demographic Data for Study Participants (n ⫽ 60) Variable Gender Male Female Ethnicity African-American White Hispanic Native American Marital status Single Married Divorced/separated Education Not enter high school Completed high school Number of children None One Two Three
Percent (%) 50 50 73 17 5 2 90 3 7 4 37 58 25 15 2
Procedure Once recruited, the third author (T.S.) administered all measures to each participant in an interview format (to control for differences in reading comprehension level). These interviews were conducted at the shelter and lasted approximately 2 h. Half of the participants (15 males and 15 females) were randomly selected to complete the PVI and Conflict Tactics Scale (CTS) a second time 4 weeks after the initial interview. The group that participated in the test-retest administration did not differ significantly on any of the demographic variables from the group that participated in the first administration only. Measures Partner Violence Interview. This 26-item structured interview, adapted from the CTS [28], asks subjects to report on the various levels of violence that they have experienced or perpetrated in their lifetime. The measure focuses on and expands the analysis of physical (including sexual) violence. It takes approximately 25 minutes to complete. The PVI begins with behaviors common in abusive relationships, such as preventing an individual from seeing certain people, verbally pressuring one’s partner for sex, or driving recklessly to scare the partner. Items progress to cover incidences in which physical force is applied from physically forcing sex on one’s partner to striking, stabbing, or shooting them. Unlike the CTS,
358
BORIS ET AL.
PVI items are directed at three separate areas: violence inflicted by current partners (current partner scale), violence inflicted by any previous partners (ex-partner scale), and violence inflicted by the subject toward a partner (toward partner scale, collected at time 2 only). The CTS has subscales for verbal aggression and reasoning as well as actual physical violence, while other clinical scales, such as the PVI, concentrate on direct violent interchanges [27]. The PVI forgoes items pertaining to verbal aggression and reasoning, although the questions that make up these scales capture the use of belittling and rationalization in relationships, which may proceed to the kind of coercion that is common in violent relationships. In place of these items, a broader spectrum of physical violence items was added to the PVI so that specific types of violent interchanges that are not included in the CTS are not missed [23]. The PVI begins with warm-up questions (e.g., “Has your partner yelled or screamed at you?”) of a nonphysical nature (which are not scored) and ends with 14 questions that range from pushing and shoving to stabbings and shootings. Respondents are also asked whether their child witnessed violence between the subject and current or ex-partners. Each item is responded to as “not ever having experienced or engaged in” (0) or as “having experienced or engaged in the item listed” (1). Items are listed in increasing order of severity. The overall score in each of the three categories is obtained by summing positive responses. As will be reported, various cutoffs can be used, based on a priori set points for severity. Two additional scales are reported on at the end of the interview. The first asks the respondent to judge the level of effect that witnessing violence has had on their child or children, ranging from 0 ⫽ “no effect/no exposure” to 4 ⫽ “severe effect” (e.g., significant long-term reaction). The second asks the respondent to report on lasting physical effects of experienced violence, ranging from no lasting effect to permanent physical damage (e.g., hearing loss, scarring, etc.) Conflict Tactics Scale. This 20-item self-report survey [28] measures the degree of conflict present in intimate relationships over the past year. Scores are computed for three categories: Reasoning, Verbal Aggression, and Violence. For each item, subjects are asked to rate how often an item of conflict occurs in their relationship during the average year (the possible answers are: 0 ⫽ “never”; 1 ⫽ “once a year”; 2 ⫽ “two or three times a year”; 3 ⫽ “often but less than once a month”; 4 ⫽ “about once a month”; 5 ⫽
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
“more than once a month”). The Reasoning subscale measures the use of rational discussion, argument, and reasoning. The Verbal Aggression subscale measures the use of verbal and nonverbal acts and/or threats between partners. The Violence subscale measures the use of physical force between partners. These subscales are designed to capture the progression in relationships from the use of reasoning to verbal aggression and coercion to controlling the partner with physical force. Previous research with the CTS has demonstrated sufficient internal consistency and validity, and it is clear that relationships may include incidences of, for example, patterns of attempts at reasoning with a partner mixed with the use physical force to control the partner [28]. Beck Depression Inventory (BDI). This is a widely used, reliable, and extensively validated self-report measure of depressed mood [30] that has been studied with adults and adolescents. This questionnaire’s 21 items require respondents to rate the frequency with which they experienced certain feelings or behaviors during the past week (e.g., feeling sad, blaming oneself for mistakes, crying, difficulty making decisions, and lack of appetite). All items are answered on a 4-point scale (range 0 to 3) with a higher score indicating more symptomatology. Subjects may endorse one or more answers for each item. A summed score using research-based cutoffs is correlated with the diagnosis of Major Depression as diagnosed by clinician interview. In this study, the BDI was administered at Time 1 only. Survey of Children’s Exposure to Community Violence. This 51-item questionnaire [31] assesses an individual’s lifetime exposure to 20 forms of community violence. Items are answered in a true-false format and consist of statements regarding being a victim of, witnessing, or hearing about violent activities. Violence exposure via media, however, is explicitly excluded. A summed continuous score is used as an indicator of total community violence exposure. This measure was administered at Time 1 only. This article will report data on the reliability of the PVI scales through three methods: (a) a Pearson correlation will be calculated with the data collected at two assessments conducted 1 month apart, (b) a repeated-measures analysis between the two administrations will be computed, and (c) the internal consistency of each PVI scale at each time period will be examined. Convergent validity will be assessed by calculating correlations between a second, and
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
Table 2. Prevalence of Violence on the Three PVI Scales for Items 13 to 26: Percentage of Respondents Endorsing One or More Items. PVI Category Time 1 Current partner Ex-partner Toward partner Time 2 Current partner Ex-partner Toward partner
Items 13–26 % (n)
Items 14 –26 % (n)
Severity Score % (n)
63 (22) 78 (47)
60 (21) 73 (44)
54 (19) 72 (43)
27 (07) 73 (22) 53 (16)
27 (07) 73 (22) 57 (17)
19 (05) 66 (20) 47 (14)
* PVI ⫽ Partner Violence Interview.
related, partner violence measure. Further, we will present the data from the partner violence measures as they relate to community violence exposure and level of depression. From the perspective of convergent validity, we hypothesized that there would be a significant relationship between partner violence and depression for females but not for males.
Results We will begin by presenting data on the incidence and severity of partner violence in this sample, using the PVI. Following that, we will discuss the properties of the PVI, including data on test-retest reliability and internal consistency. We then present evidence of convergent validity by presenting data on how the PVI relates to the CTS. Finally, we will detail the associations among partner violence, community violence, and depression for our sample and explore the issue of gender differences across these measures. Incidence and Severity of Partner Violence This group of 60 homeless young adults reported a very high rate of violence in their intimate relationships. On average, these subjects reported being involved in two violent intimate relationships, with a range of 0 to 20. Because there is no accepted definition of what defines violence exposure, we varied our cutoffs [23]. Of the sample, 44 of 60 (73%) endorsed at least one item reflecting physical violence inflicted by an ex-partner (Table 2 for prevalence across all three scales and both time periods). At a minimum, subjects had to endorse being “pushed, carried, restrained, grabbed, or shoved” by a partner to be included in this group. A slightly stricter cutoff, namely upping the minimum criteria
359
to being “slapped or spanked” (with hostile intent), did not change this percentage. Although shelter rules and personnel discouraged intimate relationships, 35 (58%) of the subjects reported having a current partner; of these, 21 (60%) endorsed being “pushed, carried, restrained, grabbed, or shoved” by that partner (Table 2). All but one of the subjects who reported this level of violence also endorsed being “slapped or spanked” by their partner. Given that there was little difference based on these two cutoffs for physical violence and because being “pushed, carried, restrained, grabbed, or shoved” does reflect some degree of hands-on assault, this cutoff was used in all remaining analyses. We also investigated whether there was a subsample of this cohort who denied engaging in partner violence. (No one in the sample denied having at least one intimate partner in his or her life.) At Time 1, 6 of 60 subjects (10%) reported no violence with either ex-partners or their current partner. Similarly, 3 of 30 subjects (10%) retested a month later were consistent in reporting no violence with any partner. We used two methods to estimate severity of violence: (a) we summed the more severe physical violence items, and (b) we used the subject’s reports of injury resulting from physical attack by their partner, which ranged from 0 (did not sustain an injury) to 4 (sustained permanent damage). The items summed into a single score reflecting severe violence involved extreme, forceful, and life-threatening levels of violence (e.g., “burned you,” “repeatedly kicked or hit you/beat you up,” “used a gun or knife on you”). This type of more severe violence was common, with 43 of 60 (72%) reporting at least one of these items in past relationships and 19 of the 35 (54%) who had a current partner experiencing one or more of these items with their current partner. At Time 2, the levels of severe violence with an expartner remained high: 20 of 30 subjects (67.1%) endorsed at least one severe physical violence item. However, reported violence with the current partner dropped because only 19% (5 of 26 subjects) endorsed one or more of these same items. When subjects were asked about the worst injuries they had sustained as a result of partner violence using the subjective 4-point severity scale, the median score was 1.00 (a score of 1 is equivalent to “mild injury” with “a small bruise, scratch, or swelling” given as examples). However, 11 of 60 (18%) of the sample reported experiencing “moderate injury” (for example, “fractures, minor burns, cuts, and large bruises”) and a further 6 of 60 (10%) reported sus-
360
BORIS ET AL.
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
Table 3. Retest Reliability for the Partner Violence Interview (PVI) Scales (n ⫽ 30) PVI Category
Pearson r
Significance Level
Current partner Ex-partner
0.85 0.70
p ⬍ .001 p ⬍ .001
taining a “severe injury” (e.g., major wounds, severe bleeding, loss of consciousness). One subject reported sustaining “permanent damage” from their injuries (“visual loss, hearing loss, disfigurement, and chronic pain” were used as examples for this category). We then looked at whether severity of violence (e.g., endorsing one or more severe physical violence items) was correlated with the subject’s rating of their worst injury. Those subjects who endorsed severe violence with ex-partners also endorsed more severe injury using the 4-point scale (Spearman’s rho ⫽ .557, p ⬍ .001). Injury severity was not related to reported violence with current partners or with violence perpetrated by the respondent toward their partner. Retest Reliability Both the Current Partner and Ex-partner PVI scales demonstrated a significant and moderate to high correlation between the two administration periods, which were 1 month apart with 30 randomly selected subjects from the original group (Table 3). (Toward Partner data were available only for Time 2.) The repeated-measures analysis of variance was computed to examine differences between administration at Time 1 and administration at Time 2 for two of the three PVI scales. As expected, the Ex-partner scale score did not differ between the two time administrations (F ⫽ (1, 29) .438, p ⬎ .05). However, scale scores reflecting violent acts by the current partner toward the respondent did differ significantly between the first and second administration (F (1, 12) ⫽ 5.73, p ⬍ .05). The mean score for Current Partner violence was 4.00 (SD ⫽ 4.36) at the first administration; this decreased to 2.38 (SD ⫽ 3.70) at the second administration. Only 13 subjects reported having a current partner at both time periods, and it is unclear whether the subjects who reported a current partner at Time 1 and Time 2 were reporting on the same partner. Internal Consistency Test-retest reliability has been criticized as a method of reliability because the experience of the first
Table 4. Internal Consistency (KR-20) for the Three Partner Violence Interview (PVI) Scales Time 1 PVI Category Current partner Ex-partner Toward partner
Time 2
Items 1–26
Items 13–26
Items 1–26
Items 13–26
.91 .89
.89 .88
.95 .92 .83
.95 .91 .80
testing may influence the second testing [32]. For this and other reasons (e.g., engaging in intimate relationships was discouraged by shelter staff), we calculated the internal consistency for each PVI scale at each time period using the Kuder-Richardson formula 20 (KR-20) designed for data involving dichotomous items. For all three scales, the KR-20 ranged from .78 to .93 (Table 4). Convergent Validity A first test of convergent validity was investigated by examining correlations between PVI categories and the three CTS subscale scores. Because subjects tended to report less current partner violence from Time 1 to Time 2, both time points were included for analysis. The CTS Violence subscale correlated significantly with the PVI Ex-partner subscale at both time points and with the PVI Toward partner scale at Time 2 only (Table 5). There were no relationships between any CTS scale and the PVI Current Partner scale. Ex-partner also correlated with Verbal Aggression at Time 1 but not Time 2. The relationship among scores on the Community Violence measure and partner violence (as measured by the PVI) was also investigated, using the entire sample at Time 1. The Ex-partner scale correlated with lifetime Community Violence exposure (r ⫽ Table 5. Correlations Between the Partner Violence Interview (PVI) Categories and Conflict Tactics Scale (CTS) at Administration 1 and 2 PVI Category
CTS Reasoning
Time 1 Current partner Ex-partner Time 2 Current partner Ex-partner Toward partner * p ⬍ .001; ** p ⬍ .01.
⫺.193 .140 .246 ⫺.092 .198
CTS Violence
CTS Verbal Aggression
.203
.066
.596*
.363**
.168 .444** .475**
.020 .157 .287
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
.310, p ⬍ .05). However, violence with current partners was not significantly correlated with Community Violence exposure. Finally, the relationship between the BDI and the two PVI scales was examined. Again, the data from Time 1 (when the full sample was tested) were used. The Ex-partner scale score did not correlate significantly with the BDI (r ⫽ .15). However, the BDI did correlate significantly with the Current Partner score (r ⫽ .457, p ⬍ .01).
Gender Differences Gender differences were examined on the Partner Violence measure, the Community Violence scale, and the BDI. There were no gender differences on any of the three PVI scales (“current partner,” “expartner,” or “toward partner”) or BDI scores. Further, the number of violent partners for each subject differed by neither gender nor severity of each subject’s most serious injury. Those subjects who were mothers reported no more severe effect of witnessing partner violence on their children than those subjects who were fathers. The only scale in which significant differences were found was the Community Violence scale (F ⫽ 16.79. p ⬍ .001), with males reporting a higher level of exposure to community violence.
Discussion This study confirms that homeless young adults experience significant amounts of violence, both in their communities and in intimate relationships. In fact, physical violence between homeless young adults and their partners appears normative. For this sample, setting a cut point for physical violence was not difficult. A large majority reported experiencing violence, and these experiences were likely to include being beaten, burned, or knifed; more than one-third of the sample reported moderate or severe injury from partner violence. Furthermore, for many in this sample, physical violence with partners recurs. As a group, more than one violent relationship was the norm. The data also suggest that the PVI appears wellsuited for populations of this sort. For data on both past and current relationships, test-retest stability across a 1-month time span was high. Internal consistency across all three scales was adequate. Furthermore, the measure is easy to administer in interview format and covers a broader range of physical
361
violence experiences than the CTS. Reading comprehension is, of course, less of a concern with a questionnaire that is designed to be used in interview format. The PVI also allows for two ways of estimating severity: namely using different cutoffs and having the subject rate the severity of any sequelae they may have experienced. In this cohort, subjects who reported very severe violence in relationships reported correspondingly high injury scores. Asking subjects to identify violent acts that they perpetrated toward their partners is important. Although verifying these data by comparing reports from both partners would strengthen these data [33], it is clear from this study that this group of young adults was willing to admit perpetrating violence in an interview session. There is a tradition of using anonymous scales to gather data on partner violence, although it is not clear whether there is a greater tendency to give socially desirable answers even when anonymity is maintained. The PVI also includes questions about whether the subject’s child(ren) has witnessed any instance of partner violence. In this study, young adults freely admitted perpetrating violence in relationships and, in some cases, stated that their children witnessed instances of partner violence. The negative developmental impact of witnessing domestic violence is currently being documented, although it is already clear that serious negative developmental and behavioral consequences are associated with witnessing partner violence [34,35]. A study that used the PVI to assess partner violence revealed an association between witnessing partner violence in infancy and disorganized infant attachment, which is a known risk factor for later psychopathology [36]. This study gives only a preliminary and incomplete view of convergent validity. Comparing reports of violence using the PVI with reports using the CTS does suggest that the CTS is tapping other constructs besides physical aggression, and the PVI is not designed to do this. On the other hand, the PVI extends the number of violent acts covered, which gives a more complete picture of violence exposure. The PVI also gauges aggressive acts by the subject themselves and documents exposure of young children to this violence. Selection of which instrument to use in research will depend on the research questions at hand. Unlike with adults, in whom coercion and control by the male of the female seems common, bidirectional violence appears to be common among high-risk young adults. The fact that the PVI and CTS violence subscale correlate suggests that each measure is tapping the same general construct; this is
362
BORIS ET AL.
not a surprise, given the derivation of the PVI. On the other hand, the degree of correlation is only moderate, suggesting that each instrument may provide unique data about physical violence in relationships [37]. The CTS has been extensively used with adults, and it may be that the PVI is better suited for high-risk samples of adolescents and young adults. However, the road to documenting convergent validity is a long one, and further study of the PVI is indicated by using other related measures in other samples. We began this process in this study by investigating how the PVI relates to other experiences of interest. Predictably, there was a significant correlation between partner violence with current partners and level of depressive symptoms for both males and females. However, past violence did not correlate with depression scores. Other studies have shown that even low-level violence in relationships may be related to depressive symptoms (at least for women); the data are less clear for adolescents and young adults [38]. It remains unclear both what level of violence relates with depressive symptoms and how long those symptoms, given intervening variables, might last. Certainly, in the assessment of depression in young adults, inquiring about violence exposure in general, and partner violence in particular, is warranted given these findings. In fact, a growing body of literature suggests that a broader evaluation, including assessment for posttraumatic stress disorder, is warranted for high-risk adolescents and young adults such as those residing in inner cities [3,4]. Still, we were encouraged that past exposure to partner violence was not related to current symptoms of depression, suggesting some resiliency even in this extremely high-risk group. Another interesting relationship concerns community violence and partner violence. In this case, current partner violence did not correlate with lifetime community violence experiences. However, expartner violence did. Furthermore, although on average, males reported more exposure to community violence, partner violence was as commonly reported by males as by females. Surely, the mechanisms with which violent experiences beget other violent experiences require more study [7]. This study was conducted with a high-risk group, and an association between community violence and partner violence might be predicted by social learning theory. If etiologic links between one type of violent experience and another exist, it is perhaps longitudinal studies of groups like homeless young adults that will provide the necessary data.
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
On the other hand, one of the limitations of this study is the high-risk status of this group. These data may not be generalizable to young adults at less risk. Furthermore, some findings may have been influenced by the fact that this sample was housed in a shelter. Data on current partner violence may not reflect that which might be observed in another setting, given that the shelter staff actively discouraged intimate relationships. (Nevertheless, more than half the sample reported having a current partner at both time points.) It would have been interesting to have a comparison group of urban young adults who are not homeless. The current partner repeated-measures data, in particular, may have been influenced by shelter rules. Although there was a strong correlation between reported current partner violence at the two time points, there was also a significant drop-off in reported acts of violence with current partners across a month’s time. We do not know whether the subjects who reported current partner violence at Time 2 had the same partners at Time 1. Of course, this decline in current partner violence did not occur because the sample stopped having relationships. In fact, unstable relationships with high turnover rates appear to be the norm for highrisk urban adolescents and young adults [40]. It may be that the general shelter structure and/or specific unmeasured interventions account for the differences in current partner scores across time. Characterizing and measuring what prevents partner violence in adolescence is a critical research goal, although few studies are available as yet [39,40]. Given the severity and pervasiveness of partner violence among homeless young adults, preventing ongoing partner violence is clearly warranted. In fact, partner violence appears to be a key precipitating factor pushing housed adult women to the streets [41], a pattern that may begin in adolescence. Our data suggest that both male and female homeless young adults are at high risk for injury at the hands of their partners. Prevention programs might decrease injury and also keeping already at-risk children from witnessing ongoing partner violence. We thank Maudelle Cade, Jackie Harris, and the dedicated staff of Covenant House of New Orleans. We are especially indebted to the young adults who shared their experiences with us.
References 1. Rachuba L, Stanton B, Howard D. Violent crime in the United States. An epidemiologic profile. Arch Pediatr Adolesc Med 1995;149:953– 60.
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
2. DuRant R, Cadenhead C, Pendergrast RA, et al. Factors associated with the use of violence among urban black adolescents. Am J Public Health 1994;84:612–7. 3. Giancona RM, Reinherz HZ, Silverman AB, et al. Traumas and posttraumatic stress disorder in a community population of older adolescents. J Am Acad Child Adolesc Psychiatry 1995; 34:1369 – 80. 4. Horowitz K, Weine S, Jekel J. PTSD symptoms in urban adolescent girls: Compounded community trauma. J Am Acad Child Adolesc Psychiatry 1995;34:1353– 61. 5. Wolfe DA, Wekerle C, Reitzel-Jaff, D, et al. Factors associated with abusive relationships among maltreated and nonmaltreated youth. Dev Psychopathol 1998;10:61– 85. 6. Cicchetti D, Toth SL. A developmental psychopathology perspective on child abuse and neglect. J Am Acad Child Adolesc Psychiatry 1995;34:541– 65. 7. Follette VM, Polusny MA, Bechtle AE, et al. Cumulative trauma: The impact of child sexual abuse, adult sexual assault, and spouse abuse. J Trauma Stress 1996;9:25–35. 8. Malik S, Sorenson SB, Aneshensel CS. Community and dating violence among adolescents: Perpetration and victimization J Adolesc Health 1997;21:291–302. 9. Kipke MD, Simon TR, Montgomery SB, et al. Homeless youth and their exposure to and involvement in violence while living on the streets. J Adolesc Health 1997;20:360 –7. 10. Whitbeck LB, Hoyt DR, Ackley KA. Families of homeless and runaway adolescents: A comparison of parent/caretaker and adolescent perspectives on parenting, family violence, and adolescent conduct. Child Abuse Negl 1997;2:517–28. 11. Herman DB, Susser ES, Streuning EL, et al. Adverse childhood experiences: Are they risk factors for adult homelessness? Am J Public Health 1997;87:249 –55. 12. Feitel B, Margetson N, Chamas J. Psychosocial background and behavioral and emotional disorders of homeless and runaway youth. Hosp Community Psychiatry 1992;43:155–9. 13. Booth R, Zhang Y. Severe aggression and related conduct problems among runaway and homeless adolescents. Psychiatr Serv 1996;47:75– 80. 14. North CS, Smith ES, Spitznagel EL. Violence and the homeless: An epidemiologic study of victimization and aggression. J Trauma Stress 1994;7:95–110. 15. Robins LN, O’Neal P. The adult prognosis for runaway children. Am J Orthopsychiatry 1959;29:752– 61. 16. Hamberger LK, Ambuel B. Dating violence. Pediatr Clin North Am 1998;45:381–90. 17. Bailey SL, Camlin CS, Ennett ST. Substance use and risky sexual behavior among homeless and runaway youth. J Adolesc Health 1998;23:378 – 88. 18. Greene JM, Ringwalt CL. Pregnancy among three national samples of runaway and homeless youth. J Adolesc Health 1998;23:370 –7. 19. Greene JM, Ennett ST, Ringwalt CL. Substance use among runaway and homeless youth in three national samples. Am J Public Health 1997;87:229 –35. 20. Amaro H, Fried LE, Cabral H, et al. Violence during pregnancy and substance abuse. Am J Public Health 1990;80:575–9. 21. Coker AL, Smith PH, McKeown RE, et al. Frequency and correlates of partner violence by type: Physical, sexual and psychological battering. Am J Public Health 2000;90:553–9.
363
22. Kreiter SR, Krowchuck DP, Woods CR, et al. Gender differences in risk behaviors among adolescents who experience date violence. Pediatrics 1999;104:1286 –92. 23. Hamby SL, Poindexter VC, Gray-Little B. Four measures of partner violence: construct similarity and classification differences. J Marr Fam 1996;58:127–39. 24. Attala JM, Hudson WW, McSweeney M. A partial validation of two short-form partner abuse screens. Women Health 1994;21:125–39. 25. Chase KA, Treboux D, O’Leary KD, et al. Specificity of dating aggression and its justification among adolescents. J Abnorm Child Psychol 1998;26:467–73. 26. Martin B. The transmission of relationship difficulties from one generation to the next. J Youth Violence 1990;19:181–99. 27. Furbee PM, Sikora R, Williams JM, et al. Comparison of domestic violence screening methods: A pilot study. Ann Emerg Med 1998;31:495–501. 28. Straus MA. The Conflict Tactics Scale and its critics: An evaluation of new data on validity and reliability. In: Straus MA, Gelles RJ (eds). Physical Violence in American Families: Risk Factors and Adaptations to Violence in 8,145 Families. New Brunswick, NJ: Transaction, 1990:49 –73. 29. Heller SS, Zeanah CH, Hawkins T. Predictors of domestic violence in homeless young adults. Presented at the Meeting for the Society for Research in Child Development, Washington, DC, April 1997. 30. Beck AT, Ward CH, Mendelson M, et al. An inventory for measuring depression. Gen Psychiatry 1961;4:561–71. 31. Richters JE, Saltzman W. Survey of Children’s Exposure to Community Violence: Parent Report. National Institute of Mental Health, 1990. 32. Carmine EG, Zeller RA. Reliability and Validity Assessment. Sage University Paper series on Quantitative Applications in the Social Sciences, series no. 17. Beverly Hills, CA: Sage, 1979. 33. Bohannon JR, Dosser DA Jr, Lindley SE. Using couple data to determine domestic violence rates: An attempt to replicate previous work. Violence Victim 1995;10:133– 41. 34. Suderman M, Jaffe P. Children and youth who witness violence. In: Wolfe DA, McMahon RJ, Peters R DeV (eds). Child Abuse: New Directions in Prevention and Treatment Across the Lifespan. Thousand Oaks, CA: Sage, 1997:55–78. 35. Fantuzzo J, Baruch R, Bercama A, et al. Domestic violence and children: Prevalence and risk in five major U.S. cities. J Am Acad Child Adolesc Psychiatry 1997;36:116 –22. 36. Zeanah CH, Danis B, Hirschberg L, et al. Disorganized attachment associated with partner violence: A research note. Inf Mental Health J 1999;20:77– 86. 37. Streiner DL. A checklist for evaluating the usefulness of rating scales. Can J Psychiatry 1993;38:140 – 8. 38. Singer MI, Anglin TM, Song LY, et al. Adolescents’ exposure to violence and associated symptoms of psychological trauma. JAMA 1995;273:477– 82. 39. Wekerle C, Wolfe DA. The role of child maltreatment and attachment style in adolescent relationship violence. Dev Psychopathol 1999;10:571– 86. 40. Foshee VA, Bauman KE, Arriaga XB, et al. An evaluation of safe dates, and adolescent dating violence prevention program. Am J Public Health 1998;88:45–50. 41. Mullins G. The battered woman and homelessness. J Law Policy 1994;1:237–55.
ORIGINAL ARTICLE
Partner Violence Among Homeless Young Adults: Measurement Issues and Associations NEIL W. BORIS, M.D., SHERRYL SCOTT HELLER, Ph.D., TONYA SHEPERD, M.D., AND CHARLES H. ZEANAH, M.D.
Purpose: The primary goal of this study was to test the reliability of the Partner Violence Interview and examine validity by measuring differential correlates of partner violence. Methods: Sixty young adults (30 males and 30 females) housed in an urban shelter participated in this study. All participants were between the ages of 18 and 21 years and the majority were African-American. The participants were administered two measures of partner violence exposure, one measure of community violence exposure and one measure of depression. A random selection of 30 of the participants was retested after 1 month. Results: As predicted, current and past partner violence was common in this sample, with over 70% endorsing a history of physical violence. The Partner Violence Interview (PVI) had adequate retest reliability (Pearson r for two PVI scales ⴝ .7 and .85) and internal consistency (KR-20 for each scale ⴝ .78 to .93). Preliminary evidence of convergent validity was suggested by the fact that the PVI lifetime partner violence scale was significantly correlated with a physical violence scale from a second measure (the Conflict Tactics Scale; r ⴝ .596, p < .001). Violence in past relationships, as opposed to current relationships, was associated with both lifetime community violence exposure and current level of depression. Conclusions: The Partner Violence Interview is a reliable, comprehensive instrument suited to high-risk pop-
From the Departments of Psychiatry and Neurology, Tulane University School of Medicine (N.W.B., S.S.H., C.H.Z.) and Community Health Sciences (N.W.B.), Tulane University School of Public Health and Tropical Medicine, New Orleans, Louisiana; and Department of Obstetrics and Gynecology, Louisiana State University School of Medicine, New Orleans, Louisiana (T.S.). Address correspondence to: Neil W. Boris, M.D., Tulane University School of Public Health and Tropical Medicine, Suite 2301, Box SL-29, 1440 Canal Street, New Orleans, Louisiana 70112. E-mail: [email protected]. Manuscript accepted September 1, 2001.
ulations. Homeless young adults commonly experience severe partner violence, and preventive intervention is clearly indicated for this group. © Society for Adolescent Medicine, 2002 KEY WORDS:
Adolescents Homeless Partner violence Reliability Validity Young adults
Epidemiologic data on violent crime in the United States suggest that by the mid-1990s adolescents were exposed to the highest overall rates of lethal and nonlethal violence of any age group. Furthermore, rates of violence exposure are rising for adolescents faster than for any other age bracket [1]. Urban, minority adolescents and young adults, in particular, are likely to have been exposed to high levels of community violence [2]. Surveys of the effects of community violence in adolescence have established associations with serious psychological sequelae [3,4]. However, the relationships between varying types of violence exposure in adolescence, such as maltreatment by parents and violence in intimate partner relationships, are less well studied [5]. Furthermore, the complex processes that convert some children who have been victimized into adolescents and adults who subsequently victimize others have not yet been established, despite the considerable public health implications [5– 8]. Homeless young adults are a subpopulation
© Society for Adolescent Medicine, 2002 Published by Elsevier Science Inc., 655 Avenue of the Americas, New York, NY 10010
1054-139X/02/$–see front matter PII S1054-139X(01)00410-4
356
BORIS ET AL.
known to be particularly likely both to have been victimized and to be victimizers. Most of the available data on this subgroup of young adults come from urban drop-in centers or shelters set up to serve this population. The fact that many of these centers are located in or near neighborhoods with high rates of violence may account for the extremely high rates of exposure to community violence reported by homeless adolescents and young adults [9]. For many homeless young adults, the violence associated with living on the streets or in shelters is consistent with previous violence exposure. For instance, interviews with parents or caretakers of homeless young adults who subsequently ran away have validated teens’ reports of high levels of family violence and abuse in the home [10]. Furthermore, a careful retrospective review found that child physical abuse is a potent predictor of adult homelessness [11], although it is as yet unclear whether child abuse, considered longitudinally with potential moderating variables, is a causative factor for homelessness at any age. On the other hand, it is clear that homeless young adults may also be perpetrators of violence; high rates of aggression and conduct problems have been documented in several cross-sectional studies [10,12,13]. Furthermore, a large systematic study of homeless adults revealed that many had long histories of victimizing aggressive behaviors, often beginning before they became homeless. In this sample, victimization also was very common and rates of posttraumatic symptoms were high [14]. Finally, a longitudinal study of runaway youth seen at a child guidance clinic revealed eventual rates of antisocial personality disorder that were higher than rates for adolescents from the clinic who were not runaways [15]. There is a clear link between homelessness in adolescence and violent behavior. It is not yet clear how frequently the cycle of violence impacting homeless adolescents and young adults plays out in intimate relationships. In fact, there is little in the literature documenting the degree of violence in intimate relationships among homeless young adults, although the problem of dating violence spans the adolescent years and is generally common [8,16]. It is clear that the rate of risky sexual behavior among homeless adolescents is high [17] as are pregnancy rates [18]. It is also clear that rates of substance abuse is high among homeless adolescents and young adults [19] and that substance abuse is associated with partner violence [20,21], although the relationship between substance abuse and partner violence among young adults appears complex [22].
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
Measurement Issues Measuring exposure to, and experience of, violence in any group is a complex task. Several issues are of particular importance in approaching partner violence in subgroups such as homeless young adults. There is no accepted definition of what particular experiences constitute a case of partner violence. In fact, measures that rely on an individual’s subjective experience may not correlate highly with an objective judge’s rating of violence reports [23]. For instance, even women admitted to shelters for battered women may fail to meet differing cutoffs for physical abuse [24]. Furthermore, global ratings focused on presence or absence of violent experiences may not capture significant events that are nonetheless traumatizing. Finally, the relationship among different scales or subscales focusing on varying types of violent experiences may not be related. Another problem is that most measures focus exclusively on victimization rather than on mixed perpetration and experience of violence. There is evidence that for high-risk adolescents and young adults in particular, bidirectional violence may be common [25], and it is only more recent measures that assess this phenomenon. There is a longstanding question as to whether anonymous questionnaires allow for more forthright responses on the part of subjects who may be embarrassed or uncomfortable in reporting on their experiences. Although this question is not easily resolved, potential problems with reading comprehension among homeless adolescents make an interview format especially important for gathering reliable data in this and comparable risk groups. One factor related to the limited data on partner violence in high-risk populations may be the lack of a validated and standardized measure of partner violence suitable for use with this high-risk population [23]. An important, although understudied issue regarding partner violence among young adults has to do with gender differences in perpetration. Although there is considerable evidence among adults that partner violence is primarily a problem of coercive control of women by men, there appears to be more heterogeneity in adolescence [26]. A key factor is the reliability of the data because getting solid data on perpetration may be difficult. Similarly, the potential association among depression, partner violence, and homelessness in adult women has not been comprehensively studied in younger homeless populations. It remains unclear how commonly men are the victims of violence in
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
intimate relationships and whether the association between violence and homelessness in adult women, whether moderated by depression or not, applies to adolescent or young adult males or females. Furthermore, it is unknown whether scales that capture verbal aggression and reasoning or subjective selfratings of violence give a different perspective than scales that simply capture whether certain events have ever happened [27,28]. The data presented in this paper are from a larger study of the relationship among attachment history, childhood experiences, and a number of variables on the development of intimate relationships in homeless adolescents in which a simple 26-item measure called the Partner Violence Interview (PVI) was pilot-tested [29]. One primary goal in this study was to test whether the PVI would serve as a reliable and valid measure of violence exposure in intimate relationships in a homeless adolescent population. Further, we hypothesized that exposure to community violence would correlate with exposure to partner violence. Finally, we hypothesized that males would score significantly higher than females on the partner violence scales, despite existing evidence of patterns of bidirectional violence in adolescence.
Methods Subjects Sixty homeless young adults were recruited over a 3-month period from a private nonprofit shelter/ transitional housing facility in a medium-sized southern city. Demographic data on the sample are summarized in Table 1. Although the shelter housed some teens under 18 years, only 18- to 21-year-olds were recruited for this study because of the need to seek parental permission for those under 18 years. Consequently, the mean age for this sample was 19 years. At this particular shelter, most of the adolescents and young adults were housed in an “acute unit” in which they were allowed to stay for 60 days. Although we did not systematically gather data on how long each subject had been homeless, the majority of the sample was recruited from this unit. As noted in Table 1, most of the sample was African-American, single, and, at the time of the study, had failed to complete high school. Of the sample, 42% had at least one child. Males and females did not differ significantly on any of the demographic variables.
357
Table 1. Demographic Data for Study Participants (n ⫽ 60) Variable Gender Male Female Ethnicity African-American White Hispanic Native American Marital status Single Married Divorced/separated Education Not enter high school Completed high school Number of children None One Two Three
Percent (%) 50 50 73 17 5 2 90 3 7 4 37 58 25 15 2
Procedure Once recruited, the third author (T.S.) administered all measures to each participant in an interview format (to control for differences in reading comprehension level). These interviews were conducted at the shelter and lasted approximately 2 h. Half of the participants (15 males and 15 females) were randomly selected to complete the PVI and Conflict Tactics Scale (CTS) a second time 4 weeks after the initial interview. The group that participated in the test-retest administration did not differ significantly on any of the demographic variables from the group that participated in the first administration only. Measures Partner Violence Interview. This 26-item structured interview, adapted from the CTS [28], asks subjects to report on the various levels of violence that they have experienced or perpetrated in their lifetime. The measure focuses on and expands the analysis of physical (including sexual) violence. It takes approximately 25 minutes to complete. The PVI begins with behaviors common in abusive relationships, such as preventing an individual from seeing certain people, verbally pressuring one’s partner for sex, or driving recklessly to scare the partner. Items progress to cover incidences in which physical force is applied from physically forcing sex on one’s partner to striking, stabbing, or shooting them. Unlike the CTS,
358
BORIS ET AL.
PVI items are directed at three separate areas: violence inflicted by current partners (current partner scale), violence inflicted by any previous partners (ex-partner scale), and violence inflicted by the subject toward a partner (toward partner scale, collected at time 2 only). The CTS has subscales for verbal aggression and reasoning as well as actual physical violence, while other clinical scales, such as the PVI, concentrate on direct violent interchanges [27]. The PVI forgoes items pertaining to verbal aggression and reasoning, although the questions that make up these scales capture the use of belittling and rationalization in relationships, which may proceed to the kind of coercion that is common in violent relationships. In place of these items, a broader spectrum of physical violence items was added to the PVI so that specific types of violent interchanges that are not included in the CTS are not missed [23]. The PVI begins with warm-up questions (e.g., “Has your partner yelled or screamed at you?”) of a nonphysical nature (which are not scored) and ends with 14 questions that range from pushing and shoving to stabbings and shootings. Respondents are also asked whether their child witnessed violence between the subject and current or ex-partners. Each item is responded to as “not ever having experienced or engaged in” (0) or as “having experienced or engaged in the item listed” (1). Items are listed in increasing order of severity. The overall score in each of the three categories is obtained by summing positive responses. As will be reported, various cutoffs can be used, based on a priori set points for severity. Two additional scales are reported on at the end of the interview. The first asks the respondent to judge the level of effect that witnessing violence has had on their child or children, ranging from 0 ⫽ “no effect/no exposure” to 4 ⫽ “severe effect” (e.g., significant long-term reaction). The second asks the respondent to report on lasting physical effects of experienced violence, ranging from no lasting effect to permanent physical damage (e.g., hearing loss, scarring, etc.) Conflict Tactics Scale. This 20-item self-report survey [28] measures the degree of conflict present in intimate relationships over the past year. Scores are computed for three categories: Reasoning, Verbal Aggression, and Violence. For each item, subjects are asked to rate how often an item of conflict occurs in their relationship during the average year (the possible answers are: 0 ⫽ “never”; 1 ⫽ “once a year”; 2 ⫽ “two or three times a year”; 3 ⫽ “often but less than once a month”; 4 ⫽ “about once a month”; 5 ⫽
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
“more than once a month”). The Reasoning subscale measures the use of rational discussion, argument, and reasoning. The Verbal Aggression subscale measures the use of verbal and nonverbal acts and/or threats between partners. The Violence subscale measures the use of physical force between partners. These subscales are designed to capture the progression in relationships from the use of reasoning to verbal aggression and coercion to controlling the partner with physical force. Previous research with the CTS has demonstrated sufficient internal consistency and validity, and it is clear that relationships may include incidences of, for example, patterns of attempts at reasoning with a partner mixed with the use physical force to control the partner [28]. Beck Depression Inventory (BDI). This is a widely used, reliable, and extensively validated self-report measure of depressed mood [30] that has been studied with adults and adolescents. This questionnaire’s 21 items require respondents to rate the frequency with which they experienced certain feelings or behaviors during the past week (e.g., feeling sad, blaming oneself for mistakes, crying, difficulty making decisions, and lack of appetite). All items are answered on a 4-point scale (range 0 to 3) with a higher score indicating more symptomatology. Subjects may endorse one or more answers for each item. A summed score using research-based cutoffs is correlated with the diagnosis of Major Depression as diagnosed by clinician interview. In this study, the BDI was administered at Time 1 only. Survey of Children’s Exposure to Community Violence. This 51-item questionnaire [31] assesses an individual’s lifetime exposure to 20 forms of community violence. Items are answered in a true-false format and consist of statements regarding being a victim of, witnessing, or hearing about violent activities. Violence exposure via media, however, is explicitly excluded. A summed continuous score is used as an indicator of total community violence exposure. This measure was administered at Time 1 only. This article will report data on the reliability of the PVI scales through three methods: (a) a Pearson correlation will be calculated with the data collected at two assessments conducted 1 month apart, (b) a repeated-measures analysis between the two administrations will be computed, and (c) the internal consistency of each PVI scale at each time period will be examined. Convergent validity will be assessed by calculating correlations between a second, and
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
Table 2. Prevalence of Violence on the Three PVI Scales for Items 13 to 26: Percentage of Respondents Endorsing One or More Items. PVI Category Time 1 Current partner Ex-partner Toward partner Time 2 Current partner Ex-partner Toward partner
Items 13–26 % (n)
Items 14 –26 % (n)
Severity Score % (n)
63 (22) 78 (47)
60 (21) 73 (44)
54 (19) 72 (43)
27 (07) 73 (22) 53 (16)
27 (07) 73 (22) 57 (17)
19 (05) 66 (20) 47 (14)
* PVI ⫽ Partner Violence Interview.
related, partner violence measure. Further, we will present the data from the partner violence measures as they relate to community violence exposure and level of depression. From the perspective of convergent validity, we hypothesized that there would be a significant relationship between partner violence and depression for females but not for males.
Results We will begin by presenting data on the incidence and severity of partner violence in this sample, using the PVI. Following that, we will discuss the properties of the PVI, including data on test-retest reliability and internal consistency. We then present evidence of convergent validity by presenting data on how the PVI relates to the CTS. Finally, we will detail the associations among partner violence, community violence, and depression for our sample and explore the issue of gender differences across these measures. Incidence and Severity of Partner Violence This group of 60 homeless young adults reported a very high rate of violence in their intimate relationships. On average, these subjects reported being involved in two violent intimate relationships, with a range of 0 to 20. Because there is no accepted definition of what defines violence exposure, we varied our cutoffs [23]. Of the sample, 44 of 60 (73%) endorsed at least one item reflecting physical violence inflicted by an ex-partner (Table 2 for prevalence across all three scales and both time periods). At a minimum, subjects had to endorse being “pushed, carried, restrained, grabbed, or shoved” by a partner to be included in this group. A slightly stricter cutoff, namely upping the minimum criteria
359
to being “slapped or spanked” (with hostile intent), did not change this percentage. Although shelter rules and personnel discouraged intimate relationships, 35 (58%) of the subjects reported having a current partner; of these, 21 (60%) endorsed being “pushed, carried, restrained, grabbed, or shoved” by that partner (Table 2). All but one of the subjects who reported this level of violence also endorsed being “slapped or spanked” by their partner. Given that there was little difference based on these two cutoffs for physical violence and because being “pushed, carried, restrained, grabbed, or shoved” does reflect some degree of hands-on assault, this cutoff was used in all remaining analyses. We also investigated whether there was a subsample of this cohort who denied engaging in partner violence. (No one in the sample denied having at least one intimate partner in his or her life.) At Time 1, 6 of 60 subjects (10%) reported no violence with either ex-partners or their current partner. Similarly, 3 of 30 subjects (10%) retested a month later were consistent in reporting no violence with any partner. We used two methods to estimate severity of violence: (a) we summed the more severe physical violence items, and (b) we used the subject’s reports of injury resulting from physical attack by their partner, which ranged from 0 (did not sustain an injury) to 4 (sustained permanent damage). The items summed into a single score reflecting severe violence involved extreme, forceful, and life-threatening levels of violence (e.g., “burned you,” “repeatedly kicked or hit you/beat you up,” “used a gun or knife on you”). This type of more severe violence was common, with 43 of 60 (72%) reporting at least one of these items in past relationships and 19 of the 35 (54%) who had a current partner experiencing one or more of these items with their current partner. At Time 2, the levels of severe violence with an expartner remained high: 20 of 30 subjects (67.1%) endorsed at least one severe physical violence item. However, reported violence with the current partner dropped because only 19% (5 of 26 subjects) endorsed one or more of these same items. When subjects were asked about the worst injuries they had sustained as a result of partner violence using the subjective 4-point severity scale, the median score was 1.00 (a score of 1 is equivalent to “mild injury” with “a small bruise, scratch, or swelling” given as examples). However, 11 of 60 (18%) of the sample reported experiencing “moderate injury” (for example, “fractures, minor burns, cuts, and large bruises”) and a further 6 of 60 (10%) reported sus-
360
BORIS ET AL.
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
Table 3. Retest Reliability for the Partner Violence Interview (PVI) Scales (n ⫽ 30) PVI Category
Pearson r
Significance Level
Current partner Ex-partner
0.85 0.70
p ⬍ .001 p ⬍ .001
taining a “severe injury” (e.g., major wounds, severe bleeding, loss of consciousness). One subject reported sustaining “permanent damage” from their injuries (“visual loss, hearing loss, disfigurement, and chronic pain” were used as examples for this category). We then looked at whether severity of violence (e.g., endorsing one or more severe physical violence items) was correlated with the subject’s rating of their worst injury. Those subjects who endorsed severe violence with ex-partners also endorsed more severe injury using the 4-point scale (Spearman’s rho ⫽ .557, p ⬍ .001). Injury severity was not related to reported violence with current partners or with violence perpetrated by the respondent toward their partner. Retest Reliability Both the Current Partner and Ex-partner PVI scales demonstrated a significant and moderate to high correlation between the two administration periods, which were 1 month apart with 30 randomly selected subjects from the original group (Table 3). (Toward Partner data were available only for Time 2.) The repeated-measures analysis of variance was computed to examine differences between administration at Time 1 and administration at Time 2 for two of the three PVI scales. As expected, the Ex-partner scale score did not differ between the two time administrations (F ⫽ (1, 29) .438, p ⬎ .05). However, scale scores reflecting violent acts by the current partner toward the respondent did differ significantly between the first and second administration (F (1, 12) ⫽ 5.73, p ⬍ .05). The mean score for Current Partner violence was 4.00 (SD ⫽ 4.36) at the first administration; this decreased to 2.38 (SD ⫽ 3.70) at the second administration. Only 13 subjects reported having a current partner at both time periods, and it is unclear whether the subjects who reported a current partner at Time 1 and Time 2 were reporting on the same partner. Internal Consistency Test-retest reliability has been criticized as a method of reliability because the experience of the first
Table 4. Internal Consistency (KR-20) for the Three Partner Violence Interview (PVI) Scales Time 1 PVI Category Current partner Ex-partner Toward partner
Time 2
Items 1–26
Items 13–26
Items 1–26
Items 13–26
.91 .89
.89 .88
.95 .92 .83
.95 .91 .80
testing may influence the second testing [32]. For this and other reasons (e.g., engaging in intimate relationships was discouraged by shelter staff), we calculated the internal consistency for each PVI scale at each time period using the Kuder-Richardson formula 20 (KR-20) designed for data involving dichotomous items. For all three scales, the KR-20 ranged from .78 to .93 (Table 4). Convergent Validity A first test of convergent validity was investigated by examining correlations between PVI categories and the three CTS subscale scores. Because subjects tended to report less current partner violence from Time 1 to Time 2, both time points were included for analysis. The CTS Violence subscale correlated significantly with the PVI Ex-partner subscale at both time points and with the PVI Toward partner scale at Time 2 only (Table 5). There were no relationships between any CTS scale and the PVI Current Partner scale. Ex-partner also correlated with Verbal Aggression at Time 1 but not Time 2. The relationship among scores on the Community Violence measure and partner violence (as measured by the PVI) was also investigated, using the entire sample at Time 1. The Ex-partner scale correlated with lifetime Community Violence exposure (r ⫽ Table 5. Correlations Between the Partner Violence Interview (PVI) Categories and Conflict Tactics Scale (CTS) at Administration 1 and 2 PVI Category
CTS Reasoning
Time 1 Current partner Ex-partner Time 2 Current partner Ex-partner Toward partner * p ⬍ .001; ** p ⬍ .01.
⫺.193 .140 .246 ⫺.092 .198
CTS Violence
CTS Verbal Aggression
.203
.066
.596*
.363**
.168 .444** .475**
.020 .157 .287
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
.310, p ⬍ .05). However, violence with current partners was not significantly correlated with Community Violence exposure. Finally, the relationship between the BDI and the two PVI scales was examined. Again, the data from Time 1 (when the full sample was tested) were used. The Ex-partner scale score did not correlate significantly with the BDI (r ⫽ .15). However, the BDI did correlate significantly with the Current Partner score (r ⫽ .457, p ⬍ .01).
Gender Differences Gender differences were examined on the Partner Violence measure, the Community Violence scale, and the BDI. There were no gender differences on any of the three PVI scales (“current partner,” “expartner,” or “toward partner”) or BDI scores. Further, the number of violent partners for each subject differed by neither gender nor severity of each subject’s most serious injury. Those subjects who were mothers reported no more severe effect of witnessing partner violence on their children than those subjects who were fathers. The only scale in which significant differences were found was the Community Violence scale (F ⫽ 16.79. p ⬍ .001), with males reporting a higher level of exposure to community violence.
Discussion This study confirms that homeless young adults experience significant amounts of violence, both in their communities and in intimate relationships. In fact, physical violence between homeless young adults and their partners appears normative. For this sample, setting a cut point for physical violence was not difficult. A large majority reported experiencing violence, and these experiences were likely to include being beaten, burned, or knifed; more than one-third of the sample reported moderate or severe injury from partner violence. Furthermore, for many in this sample, physical violence with partners recurs. As a group, more than one violent relationship was the norm. The data also suggest that the PVI appears wellsuited for populations of this sort. For data on both past and current relationships, test-retest stability across a 1-month time span was high. Internal consistency across all three scales was adequate. Furthermore, the measure is easy to administer in interview format and covers a broader range of physical
361
violence experiences than the CTS. Reading comprehension is, of course, less of a concern with a questionnaire that is designed to be used in interview format. The PVI also allows for two ways of estimating severity: namely using different cutoffs and having the subject rate the severity of any sequelae they may have experienced. In this cohort, subjects who reported very severe violence in relationships reported correspondingly high injury scores. Asking subjects to identify violent acts that they perpetrated toward their partners is important. Although verifying these data by comparing reports from both partners would strengthen these data [33], it is clear from this study that this group of young adults was willing to admit perpetrating violence in an interview session. There is a tradition of using anonymous scales to gather data on partner violence, although it is not clear whether there is a greater tendency to give socially desirable answers even when anonymity is maintained. The PVI also includes questions about whether the subject’s child(ren) has witnessed any instance of partner violence. In this study, young adults freely admitted perpetrating violence in relationships and, in some cases, stated that their children witnessed instances of partner violence. The negative developmental impact of witnessing domestic violence is currently being documented, although it is already clear that serious negative developmental and behavioral consequences are associated with witnessing partner violence [34,35]. A study that used the PVI to assess partner violence revealed an association between witnessing partner violence in infancy and disorganized infant attachment, which is a known risk factor for later psychopathology [36]. This study gives only a preliminary and incomplete view of convergent validity. Comparing reports of violence using the PVI with reports using the CTS does suggest that the CTS is tapping other constructs besides physical aggression, and the PVI is not designed to do this. On the other hand, the PVI extends the number of violent acts covered, which gives a more complete picture of violence exposure. The PVI also gauges aggressive acts by the subject themselves and documents exposure of young children to this violence. Selection of which instrument to use in research will depend on the research questions at hand. Unlike with adults, in whom coercion and control by the male of the female seems common, bidirectional violence appears to be common among high-risk young adults. The fact that the PVI and CTS violence subscale correlate suggests that each measure is tapping the same general construct; this is
362
BORIS ET AL.
not a surprise, given the derivation of the PVI. On the other hand, the degree of correlation is only moderate, suggesting that each instrument may provide unique data about physical violence in relationships [37]. The CTS has been extensively used with adults, and it may be that the PVI is better suited for high-risk samples of adolescents and young adults. However, the road to documenting convergent validity is a long one, and further study of the PVI is indicated by using other related measures in other samples. We began this process in this study by investigating how the PVI relates to other experiences of interest. Predictably, there was a significant correlation between partner violence with current partners and level of depressive symptoms for both males and females. However, past violence did not correlate with depression scores. Other studies have shown that even low-level violence in relationships may be related to depressive symptoms (at least for women); the data are less clear for adolescents and young adults [38]. It remains unclear both what level of violence relates with depressive symptoms and how long those symptoms, given intervening variables, might last. Certainly, in the assessment of depression in young adults, inquiring about violence exposure in general, and partner violence in particular, is warranted given these findings. In fact, a growing body of literature suggests that a broader evaluation, including assessment for posttraumatic stress disorder, is warranted for high-risk adolescents and young adults such as those residing in inner cities [3,4]. Still, we were encouraged that past exposure to partner violence was not related to current symptoms of depression, suggesting some resiliency even in this extremely high-risk group. Another interesting relationship concerns community violence and partner violence. In this case, current partner violence did not correlate with lifetime community violence experiences. However, expartner violence did. Furthermore, although on average, males reported more exposure to community violence, partner violence was as commonly reported by males as by females. Surely, the mechanisms with which violent experiences beget other violent experiences require more study [7]. This study was conducted with a high-risk group, and an association between community violence and partner violence might be predicted by social learning theory. If etiologic links between one type of violent experience and another exist, it is perhaps longitudinal studies of groups like homeless young adults that will provide the necessary data.
JOURNAL OF ADOLESCENT HEALTH Vol. 30, No. 5
On the other hand, one of the limitations of this study is the high-risk status of this group. These data may not be generalizable to young adults at less risk. Furthermore, some findings may have been influenced by the fact that this sample was housed in a shelter. Data on current partner violence may not reflect that which might be observed in another setting, given that the shelter staff actively discouraged intimate relationships. (Nevertheless, more than half the sample reported having a current partner at both time points.) It would have been interesting to have a comparison group of urban young adults who are not homeless. The current partner repeated-measures data, in particular, may have been influenced by shelter rules. Although there was a strong correlation between reported current partner violence at the two time points, there was also a significant drop-off in reported acts of violence with current partners across a month’s time. We do not know whether the subjects who reported current partner violence at Time 2 had the same partners at Time 1. Of course, this decline in current partner violence did not occur because the sample stopped having relationships. In fact, unstable relationships with high turnover rates appear to be the norm for highrisk urban adolescents and young adults [40]. It may be that the general shelter structure and/or specific unmeasured interventions account for the differences in current partner scores across time. Characterizing and measuring what prevents partner violence in adolescence is a critical research goal, although few studies are available as yet [39,40]. Given the severity and pervasiveness of partner violence among homeless young adults, preventing ongoing partner violence is clearly warranted. In fact, partner violence appears to be a key precipitating factor pushing housed adult women to the streets [41], a pattern that may begin in adolescence. Our data suggest that both male and female homeless young adults are at high risk for injury at the hands of their partners. Prevention programs might decrease injury and also keeping already at-risk children from witnessing ongoing partner violence. We thank Maudelle Cade, Jackie Harris, and the dedicated staff of Covenant House of New Orleans. We are especially indebted to the young adults who shared their experiences with us.
References 1. Rachuba L, Stanton B, Howard D. Violent crime in the United States. An epidemiologic profile. Arch Pediatr Adolesc Med 1995;149:953– 60.
May 2002
PARTNER VIOLENCE AMONG HOMELESS YOUNG ADULTS
2. DuRant R, Cadenhead C, Pendergrast RA, et al. Factors associated with the use of violence among urban black adolescents. Am J Public Health 1994;84:612–7. 3. Giancona RM, Reinherz HZ, Silverman AB, et al. Traumas and posttraumatic stress disorder in a community population of older adolescents. J Am Acad Child Adolesc Psychiatry 1995; 34:1369 – 80. 4. Horowitz K, Weine S, Jekel J. PTSD symptoms in urban adolescent girls: Compounded community trauma. J Am Acad Child Adolesc Psychiatry 1995;34:1353– 61. 5. Wolfe DA, Wekerle C, Reitzel-Jaff, D, et al. Factors associated with abusive relationships among maltreated and nonmaltreated youth. Dev Psychopathol 1998;10:61– 85. 6. Cicchetti D, Toth SL. A developmental psychopathology perspective on child abuse and neglect. J Am Acad Child Adolesc Psychiatry 1995;34:541– 65. 7. Follette VM, Polusny MA, Bechtle AE, et al. Cumulative trauma: The impact of child sexual abuse, adult sexual assault, and spouse abuse. J Trauma Stress 1996;9:25–35. 8. Malik S, Sorenson SB, Aneshensel CS. Community and dating violence among adolescents: Perpetration and victimization J Adolesc Health 1997;21:291–302. 9. Kipke MD, Simon TR, Montgomery SB, et al. Homeless youth and their exposure to and involvement in violence while living on the streets. J Adolesc Health 1997;20:360 –7. 10. Whitbeck LB, Hoyt DR, Ackley KA. Families of homeless and runaway adolescents: A comparison of parent/caretaker and adolescent perspectives on parenting, family violence, and adolescent conduct. Child Abuse Negl 1997;2:517–28. 11. Herman DB, Susser ES, Streuning EL, et al. Adverse childhood experiences: Are they risk factors for adult homelessness? Am J Public Health 1997;87:249 –55. 12. Feitel B, Margetson N, Chamas J. Psychosocial background and behavioral and emotional disorders of homeless and runaway youth. Hosp Community Psychiatry 1992;43:155–9. 13. Booth R, Zhang Y. Severe aggression and related conduct problems among runaway and homeless adolescents. Psychiatr Serv 1996;47:75– 80. 14. North CS, Smith ES, Spitznagel EL. Violence and the homeless: An epidemiologic study of victimization and aggression. J Trauma Stress 1994;7:95–110. 15. Robins LN, O’Neal P. The adult prognosis for runaway children. Am J Orthopsychiatry 1959;29:752– 61. 16. Hamberger LK, Ambuel B. Dating violence. Pediatr Clin North Am 1998;45:381–90. 17. Bailey SL, Camlin CS, Ennett ST. Substance use and risky sexual behavior among homeless and runaway youth. J Adolesc Health 1998;23:378 – 88. 18. Greene JM, Ringwalt CL. Pregnancy among three national samples of runaway and homeless youth. J Adolesc Health 1998;23:370 –7. 19. Greene JM, Ennett ST, Ringwalt CL. Substance use among runaway and homeless youth in three national samples. Am J Public Health 1997;87:229 –35. 20. Amaro H, Fried LE, Cabral H, et al. Violence during pregnancy and substance abuse. Am J Public Health 1990;80:575–9. 21. Coker AL, Smith PH, McKeown RE, et al. Frequency and correlates of partner violence by type: Physical, sexual and psychological battering. Am J Public Health 2000;90:553–9.
363
22. Kreiter SR, Krowchuck DP, Woods CR, et al. Gender differences in risk behaviors among adolescents who experience date violence. Pediatrics 1999;104:1286 –92. 23. Hamby SL, Poindexter VC, Gray-Little B. Four measures of partner violence: construct similarity and classification differences. J Marr Fam 1996;58:127–39. 24. Attala JM, Hudson WW, McSweeney M. A partial validation of two short-form partner abuse screens. Women Health 1994;21:125–39. 25. Chase KA, Treboux D, O’Leary KD, et al. Specificity of dating aggression and its justification among adolescents. J Abnorm Child Psychol 1998;26:467–73. 26. Martin B. The transmission of relationship difficulties from one generation to the next. J Youth Violence 1990;19:181–99. 27. Furbee PM, Sikora R, Williams JM, et al. Comparison of domestic violence screening methods: A pilot study. Ann Emerg Med 1998;31:495–501. 28. Straus MA. The Conflict Tactics Scale and its critics: An evaluation of new data on validity and reliability. In: Straus MA, Gelles RJ (eds). Physical Violence in American Families: Risk Factors and Adaptations to Violence in 8,145 Families. New Brunswick, NJ: Transaction, 1990:49 –73. 29. Heller SS, Zeanah CH, Hawkins T. Predictors of domestic violence in homeless young adults. Presented at the Meeting for the Society for Research in Child Development, Washington, DC, April 1997. 30. Beck AT, Ward CH, Mendelson M, et al. An inventory for measuring depression. Gen Psychiatry 1961;4:561–71. 31. Richters JE, Saltzman W. Survey of Children’s Exposure to Community Violence: Parent Report. National Institute of Mental Health, 1990. 32. Carmine EG, Zeller RA. Reliability and Validity Assessment. Sage University Paper series on Quantitative Applications in the Social Sciences, series no. 17. Beverly Hills, CA: Sage, 1979. 33. Bohannon JR, Dosser DA Jr, Lindley SE. Using couple data to determine domestic violence rates: An attempt to replicate previous work. Violence Victim 1995;10:133– 41. 34. Suderman M, Jaffe P. Children and youth who witness violence. In: Wolfe DA, McMahon RJ, Peters R DeV (eds). Child Abuse: New Directions in Prevention and Treatment Across the Lifespan. Thousand Oaks, CA: Sage, 1997:55–78. 35. Fantuzzo J, Baruch R, Bercama A, et al. Domestic violence and children: Prevalence and risk in five major U.S. cities. J Am Acad Child Adolesc Psychiatry 1997;36:116 –22. 36. Zeanah CH, Danis B, Hirschberg L, et al. Disorganized attachment associated with partner violence: A research note. Inf Mental Health J 1999;20:77– 86. 37. Streiner DL. A checklist for evaluating the usefulness of rating scales. Can J Psychiatry 1993;38:140 – 8. 38. Singer MI, Anglin TM, Song LY, et al. Adolescents’ exposure to violence and associated symptoms of psychological trauma. JAMA 1995;273:477– 82. 39. Wekerle C, Wolfe DA. The role of child maltreatment and attachment style in adolescent relationship violence. Dev Psychopathol 1999;10:571– 86. 40. Foshee VA, Bauman KE, Arriaga XB, et al. An evaluation of safe dates, and adolescent dating violence prevention program. Am J Public Health 1998;88:45–50. 41. Mullins G. The battered woman and homelessness. J Law Policy 1994;1:237–55.