Exposure to Violence and Nonviolent Life Stressors and Their Relations to Trauma-Related Distress and Problem Behaviors Among Urban

Early Adolescents

Erin L. Thompson, Jasmine N. Coleman, Kelly E. O’Connor, Albert D. Farrell, and Terri N. Sullivan Virginia Commonwealth University

Objective: The impact of exposure to violence must be considered within the context of a larger constellation of nonviolent life stressors faced by youth in underresourced communities. This study examined nonviolent life stressors, two types of violence exposure, and their associations with trauma- related distress and problem behaviors. Method: Participants were a predominantly African American (80%) sample of early adolescents (Mage � 12.9 years) living in communities with high rates of crime. Structural equation models examined the extent to which nonviolent life stressors and violence exposure (witnessing violence and physical victimization) were associated with adolescents’ frequencies of trauma-related distress (reexperiencing traumatic events, avoidance, and hyperarousal) and problem behaviors (physical aggression, delinquent behavior, and substance use). Results: Nonviolent life stressors, witnessing violence, and physical victimization were each significantly associated with all three symptoms of trauma-related distress and with each of the three problem behaviors. In each case, stronger relations with trauma-related distress and problem behaviors were found for nonviolent life stressors than for physical victimization. After controlling for nonviolent life stressors, both types of violence exposure remained significantly associated with problem behaviors but differed in their patterns of association with trauma-related distress. No gender differences were found among these relations. Conclusion: These findings highlight the need to control for nonviolent life stressors when examining the impact of violence exposure on adjustment. Furthermore, mental health providers may be missing important information related to adolescents’ symptomatology if they fail to inquire about trauma-related distress when adolescents deny exposure to violent and life-threatening events.

Keywords: violence, nonviolent life stressors, trauma-related distress, problem behavior, adolescence

Exposure to violence is a significant public health concern that disproportionally affects adolescents living in urban, low-income communities (Ozer & Weinstein, 2004; Stein, Jaycox, Kataoka, Rhodes, & Vestal, 2003). It includes physical victimization, de- fined as experiencing acts of force, such as being slapped, punched, hit, or shot, and witnessing violence, which involves seeing the physical victimization of someone else. A nationally

representative survey of youth living in the United States indicated that 27% of adolescents aged 10 to 13 and almost half (42%) of adolescents aged 14 to 17 had witnessed community violence in the past year (Finkelhor, Ormrod, & Turner, 2009). These rates are concerning, given the association between violence exposure and various forms of maladjustment, such as trauma-related distress, aggression, delinquency, and substance use (Fowler, Tompsett, Braciszewski, Jacques-Tiura, & Baltes, 2009; Pinchevsky, Fagan, & Wright, 2014). Adolescents in low-income, urban communities are at an increased risk not only for exposure to violence but also for a host of nonviolent stressful experiences that have been linked to maladjustment (Natsuaki et al., 2007; Ozer & Weinstein, 2004). However, few studies have examined the unique impact of expo- sure to violence on adverse outcomes after accounting for nonvi- olent life stressors (for exceptions, see Allison et al., 1999; Brooks- Gunn, Johnson, & Leventhal, 2010; Evans, 2004; Farrell et al., 2007). The purpose of this study was to examine violence exposure and nonviolent life stressors and their associations with adoles- cents’ trauma-related distress and problem behaviors.

Nonviolent Life Stressors

Ecological theory asserts that healthy development occurs most frequently when children’s environments are both consistent and predictable (Bronfenbrenner & Evans, 2000). In contrast, chaotic

This article was published Online First November 7, 2019. X Erin L. Thompson, Jasmine N. Coleman, Kelly E. O’Connor, Albert

D. Farrell, and Terri N. Sullivan, Department of Psychology, Virginia Commonwealth University.

This study was funded by the National Institute of Child Health and Human Development Grant 1R01HD089994, the National Center for In- jury Prevention and Control, Centers for Disease Control and Prevention, CDC Cooperative Agreement 5U01CE001956, and the National Institute of Justice, Grant 2014-CK-BX-0009. The findings and conclusions in this report are those of the authors, and do not necessarily represent the official position of the National Institute of Child Health and Human Development, the Centers for Disease Control and Prevention, or the National Institute of Justice.

Correspondence concerning this article should be addressed to Albert D. Farrell, Department of Psychology, Virginia Commonwealth University, P.O. Box 842018, Richmond, VA 23284-2018. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Psychology of Violence © 2019 American Psychological Association 2020, Vol. 10, No. 5, 509–519 ISSN: 2152-0828 http://dx.doi.org/10.1037/vio0000264

509

environments, characterized by high levels of crowding, noise, and residential instability (Brooks-Gunn et al., 2010), are inversely related to positive well-being (Wachs & Evans, 2010). According to the risk and resilience model of developmental psychopathology (Compas & Andreotti, 2013), nonviolent, and often chronic, life experiences can produce significant physical, cognitive, and envi- ronmental changes that increase the risk for engaging in maladap- tive behaviors (Compas & Andreotti, 2013). These types of envi- ronmental characteristics may be particularly salient among racial and ethnic minority youth living in urban settings, as they face nonviolent risk factors, such as racism, social stratification, and inequitable distribution of wealth (Evans, 2004). Indeed, research has established links between nonviolent life stressors and inter- nalizing and externalizing behaviors of ethnic and racial minority youth (Liu, Bolland, Dick, Mustanski, & Kertes, 2016; Liu, Mus- tanski, Dick, Bolland, & Kertes, 2017; Natsuaki et al., 2007). In addition, a previous study revealed that concentrated neighborhood disadvantage accounted for over a third of the difference in expo- sure to violence between African American and White youth (Zimmerman & Messner, 2013).

Despite their potential impact, few previous studies evaluating the impact of violence exposure on adjustment have taken into account the influence of other concurrent, nonviolent stressors experienced by adolescents. This is a serious limitation, given evidence suggesting that emotional and behavioral difficulties are more highly associated with nonviolent life stressors than with exposure to violence (Liu et al., 2016; Ozer & Weinstein, 2004). Ozer and Weinstein (2004), for example, found that trauma-related distress was more highly correlated with nonviolent life stressors (e.g., “no place to play in the neighborhood”) than with violence exposure (r � .52 vs. .29) among an ethnically diverse sample of seventh graders. They also found that both constructs uniquely predicted increases in trauma-related distress after controlling for one another. Similarly, Liu and colleagues (2016) found that among African American 13- to 19-year-old adolescents, nonvio- lent life stressors and violence exposure were each uniquely asso- ciated with aggressive and rule-breaking behavior in a model that also controlled for racial discrimination (�s � .23 and .16 for nonviolent life stressors and exposure to violence, respectively). Aggressive and rule-breaking behavior was also more highly cor- related with nonviolent life stressors than with exposure to vio- lence (r � .39 vs. .30, respectively). These findings provide empirical support for investigating associations between violence exposure and adjustment within the context of other nonviolent life stressors experienced by youth, particularly among adolescents of color.

Physical Victimization Versus Witnessing Violence

There is growing evidence that physical victimization and wit- nessing violence are related but distinct constructs (Vermeiren, Schwab-Stone, Deboutte, Leckman, & Ruchkin, 2003). In a meta- analysis of 110 studies, Fowler and colleagues (2009) found stron- ger associations between physical victimization, as compared with witnessing violence, and a range of externalizing problems. In contrast, Cyr and colleagues (2017) found that physical victimiza- tion (i.e., assault) was not a significant predictor of posttraumatic stress disorder (PTSD) symptoms after controlling for witnessing violence. Previous studies have also shown that whereas physical

victimization tends to co-occur with witnessing violence, not all youth who witness violence are directly victimized (Ayer et al., 2019; Ford, Grasso, Hawke, & Chapman, 2013). Taken together, these findings underscore the importance of differentiating be- tween witnessing violence and physical victimization to clarify their unique and combined associations with adolescent develop- ment.

Gender Differences

There is also a need to determine how male and female adoles- cents differ in their exposure to violent and nonviolent life stres- sors and how such stressors may influence adolescent adjustment differently. Compared with girls, boys tend to be more frequently exposed to violence (Fowler et al., 2009) and are at greater risk for engaging in problem behaviors (Card, Stucky, Sawalani, & Little, 2008). Girls, in contrast, have been shown to be at a greater risk for developing trauma-related distress (Alisic et al., 2014). However, little research has examined gender differences in nonviolent stressful life events or their differential association with internal- izing and externalizing behaviors. One exception was Liu and colleagues (2016), who found no moderating effects for gender on relations between nonviolent life stressors and externalizing prob- lems (i.e., aggressive and rule-breaking behavior) or internalizing symptoms (i.e., anxiety and depression). Additional work is war- ranted to clarify the moderating role of gender in studies evaluat- ing the unique associations between violence exposure, nonviolent life stressors, and multiple indicators of adjustment.

Current Study

The current study examined nonviolent life stressors and two types of violence exposure (i.e., witnessing violence and physical victimization) and their relations to trauma-related distress (i.e., reexperiencing traumatic events, avoidance, and hyperarousal) and problem behaviors (i.e., physical aggression, delinquent behavior, and substance use) among a predominantly African American sample of early adolescents living in urban, underresourced com- munities. We focused on concurrent relations to determine the extent to which recent experiences including nonviolent life stres- sors, witnessing violence, and physical victimization were associ- ated with trauma-related distress and problem behaviors during early adolescence. We hypothesized as follows:

Hypothesis 1: Nonviolent life stressors and both types of violence exposure would each be correlated with the three symptoms of trauma-related distress and the three problem behaviors;

Hypothesis 2: Compared with exposure to violence, nonvio- lent life stressors would be more highly correlated with each of the outcomes based on previous research;

Hypothesis 3: Nonviolent life stressors would account for a unique proportion of variance in outcomes even after control- ling for both types of violence;

Hypothesis 4: Exposure to violence would account for a unique proportion of variance in the outcomes after control- ling for nonviolent life stressors; and

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

510 THOMPSON, COLEMAN, O’CONNOR, FARRELL, AND SULLIVAN

Hypothesis 5: No specific gender differences were hypothe- sized regarding the relations between violent and nonviolent events and the adjustment variables.

Method

Participants

We conducted a secondary analysis of data from a project that collected 8 years of data between 2010 and 2018 from 2,653 students in three public schools in neighborhoods with high levels of violence (Farrell, Sullivan, Sutherland, Corona, & Masho, 2018). The purpose of that project was to evaluate the Olweus Bullying Prevention Program (Olweus & Limber, 2010). Between 74% and 100% of students at the participating schools were eligible for the federal free or reduced lunch. During Year 1, the project recruited a random sample of English-speaking sixth-, seventh-, and eighth-grade students from the rosters at each school (N � 669). In each subsequent school year, project staff recruited a new sample of 295 to 340 new participants from each school that included a new cohort of sixth-grade students and a sample of seventh- and eighth-grade students to replace those who left the schools or withdrew from the project. Active student assent and parent consent were obtained from about 80% of those eligible.

The final sample had a mean age of 12.9 years (SD � 1.10); 51% were female. The sample was about evenly distributed across the sixth, seventh, and eighth grades (ns � 876 to 891). In all, 17% identified their ethnicity as Hispanic or Latino/a, 11% did not endorse any racial categories, of whom 91% described themselves as Hispanic or Latino/a. Of the rest, the majority (80%) endorsed African American or Black as either the sole category (72%) or as one of several categories (8%). The remainder of participants described themselves as White (5%), Asian (1%), American Indian or Alaska Native (1%), or Native Hawaiian or Other Pacific Islander (1%). Approximately 41% lived with a single mother, 26% with both biological parents, 23% with a parent and step- parent, 7% with a relative without a parent, and 3% with their father without a mother or stepmother. In all, 70% participated while their school was implementing the intervention.

Procedure

The evaluation study used a multiple baseline experimental design wherein the order in which intervention activities were initiated in each school was randomized by having an administra- tor from each school draw a face-down card from a standard deck of playing cards. Intervention activities began in Year 2 at the school whose administrator drew the highest valued card, in Year 3 at the school whose administrator drew the next highest valued card, and in Year 6 at the school whose administrator drew the lowest valued card. The focus of the intervention was on improv- ing school climate through (a) school-level components, including the formation of a bullying prevention coordinating committee to assist in staff training and developing of school rules related to student behavior and (b) classroom-level, weekly classes taught by teachers, including antibullying rules, the bullying circle, leader- ship, and stress management.

Research staff described the study to students and gave them consent forms to take home to their parents. Parental consent and

student assent letters described the study as a project to learn more about school, family, and community-based programs to create safer and healthier schools and communities. Families were also told that lessons would be taught in some sixth, seventh, and eighth grade classrooms and that students would fill out a 45-min survey twice a year. Participants were given $5 gift cards if they returned the consent form, even if parents did not give consent for partic- ipation. Surveys were completed on computer-assisted interviews.

Research assistants administered surveys to small groups of students in the school during the school year and in participants’ homes or public spaces during the summer. Participants received a $10 gift card at each wave when they completed any part of the survey. For more information, see the article by Farrell, Sullivan, et al. (2018). The project collected data four times per year (i.e., every 3 months), using a planned missing data design, wherein participants were randomly assigned to complete two out of four waves during each year they participated. The planned missing data design was used to reduce costs, carryover effects, participant burden, fatigue, and attrition (Graham, Taylor, & Cumsille, 2001). Because the current study focused on relations between concurrent experiences and behavior, we created a cross-sectional data set that included one randomly selected wave for each of the 2,653 par- ticipating students. The university’s institutional review board approved all procedures.

Measures

Nonviolent life stressors. We used the Urban Adolescents Negative Life Experiences Scale to measure the frequency of experiencing nonviolent life stressors. Items were drawn from three sources: The Interpersonal Problem Solving Inventory for Urban Adolescents (Farrell, Ampy, & Meyer, 1998), the Urban Adolescents Life Experiences Scale (Allison et al., 1999), and a qualitative study in which a predominantly African American sample of adolescents from low-income communities identified stressful problem situations (Farrell et al., 2007). Priority was given to selecting items that overlapped across sources. Items that reflected witnessing violence or experiencing victimization were excluded to avoid overlap with the exposure to violence measures. The final set of 20 items included family stressors (e.g., “Family members were getting on your nerves” and “Someone in your family got in serious trouble”), transitions (e.g., “Your parent lost a job” and “Someone in your family that you were close to doesn’t live with you anymore”), resource limitations (“You didn’t get enough to eat” and “You didn’t have transportation to get some- where you wanted to go”), and neighborhood stressors (“You had trouble sleeping at night because it was noisy in your neighbor- hood or your room was too hot or too cold”). Participants rated how frequently each stressor occurred in the past 3 months on a 5-point scale (1 � never, 2 � once or twice, 3 � once or twice a month, 4 � once or twice a week, 5 � almost every day). We created a composite indicator to represent nonviolent life stressors by averaging ratings across items. This was based on Bollen and Bauldry (2011), who argued that a composite indicator may be more appropriate than a latent variable for items that do not meet the assumption of conceptual unity required by latent variables. They noted that it is more appropriate to consider items such as exposure to stressful life events as causes of a construct (exposure to stressful events), the specific pattern of which may vary across

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an Ps

yc ho

lo gi

ca l

A ss

oc ia

tio n

or on

e of

its al

lie d

pu bl

is he

rs .

T hi

s ar

tic le

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

511LIFE STRESSORS AND ADJUSTMENT

individuals rather than as interchangeable indicators that reflect an underlying latent variable. Cronbach’s � for the composite was .81 in the current study.

Violence exposure. We used the Survey of Children’s Expo- sure to Community Violence (Richters & Saltzman, 1990) to assess the frequency of exposure to violence. Although the original version assessed both the frequency and context of the incidents (e.g., relationship to perpetrator or where incident occurred), we only assessed frequency. The resulting measure included 10 items that assess victimization (e.g., “Been chased by gangs or older kids?”) and 10 that assess witnessing violence (e.g., “Seen some- one else being attacked or stabbed with a knife?”). Respondents indicated how often they had been victimized or witnessed vio- lence in the past 3 months on a 6-point scale (1 � never, 2 � 1–2 times, 3 � 3–5 times, 4 � 6–9 times, 5 � 10–19 times, 6 � 20 or more times). The original measure has been used in many studies including the National Institute of Mental Health Community Violence Project (Martinez & Richters, 1993). Based on the same rationale as for nonviolent life stressors, we created composite variables for physical victimization and witnessing violence by averaging ratings across items. Alphas based on the average fre- quency across items were .71 and .86, respectively.

Trauma-related distress. We used the Checklist of Chil- dren’s Distress Symptoms (Richters & Martinez, 1990) to assess trauma-related distress. This 28-item measure was developed to examine the impact of exposure to violence on children’s emo- tional and psychological well-being in a community violence project (i.e., Martinez & Richters, 1993). Items correspond to the Diagnostic and Statistical Manual of Mental Disorders, Third Edition (American Psychiatric Association, 1987) diagnostic cri- teria for PTSD and the PTSD symptom clusters of reexperiencing (“How often do you feel like something bad or frightening from the past is happening all over again?”), avoidance (e.g., “How often do you avoid or try not to go to places or do things that remind you something bad that happened in the past?”), and hyperarousal (e.g., “How often do you watch things around you real closely in order to protect yourself from something bad happening?”). Respondents rated each item on a 5-point scale (1 � never, 2 � seldom, 3 � once in a while, 4 � a lot of the time, 5 � most of the time). Previous research has found higher levels of violence exposure to be associated with higher scores on the Checklist of Children’s Distress Symptoms (Howard, Feigelman, Li, Cross, & Rachuba, 2002).

We conducted a confirmatory factor analysis to evaluate the three-factor solution within our sample. Consistent with previous research (Overstreet & Braun, 2000), responses were recoded to be more clinically meaningful, such that ratings of “never,” “seldom,” and “once in a while” reflected the absence or low level of a symptom (coded 0) and ratings of “a lot” or “most of the time” reflected the presence of an above threshold symptom (coded 1). We evaluated models using the �2 difference test, the root mean square error of approximation (RMSEA), comparative fit index (CFI), and Tucker–Lewis index (TLI). Although the Reexperienc- ing and Avoidance factors were highly correlated (r � .92), the three-factor model with factors representing reexperiencing, avoid- ance, and hyperarousal fit the data adequately, �2(347) � 2855.57, RMSEA � .05, CFI � .92, TLI � .92, and improved upon the fit of the one factor model based on the RMSEA, CFI, and TLI for both boys and girls (��2 � 112.15 and 222.18, ps � .001,

�RMSEA � .00 and �.01, �CFI � .01 and .02, �TLI � .02 and .02, respectively).

Problem behaviors. We used the Problem Behavior Fre- quency Scale–Adolescent Report (PBFS-AR; Farrell, Thompson, Mehari, Sullivan, & Goncy, 2018) to assess the frequency of problem behaviors (e.g., aggression, delinquent behavior, and sub- stance use). Participants reported how frequently they engaged in specific behaviors in the past 30 days using an operationally defined 6-point frequency scale (1 � never, 2 � 1–2 times, 3 � 3–5 times, 4 � 6–9 times, 5 � 10–19 times, 6 � 20 or more times). The PBFS-AR assesses three forms of aggression (in- person physical, in-person relational, and cyber), two forms of victimization (in-person and cyber), substance use, and delinquent behavior. This scoring is based on the study by Farrell, Thompson, et al. (2018), who found support for seven factors based on ordered categorical confirmatory factor analyses of data from a large, predominantly African American sample of middle school stu- dents. This seven-factor model fit the data well and demonstrated strong measurement invariance across groups that differed on gender and grade. Previous studies have found support for the validity of the PBFS-AR based on its pattern of correlations with teacher ratings of adolescents’ behavior and self-report measures of relevant constructs (Farrell, Sullivan, Goncy, & Le, 2016) and with school office discipline referrals (Farrell, Thompson, et al., 2018).

The present study created latent variables based on items from the PBFS-AR physical aggression (five items; e.g., “Hit or slapped someone”), delinquent behavior (six items; e.g., “Taken something from a store without paying for it [shoplifted]”), and substance use (nine items; e.g., “Use marijuana [pot, hash, reefer, K2]”) scales. Our analyses treated the items as ordered categorical variables using weighted least squares mean and variance adjusted estima- tors. Although PBFS-AR items are rated on a 6-point scale, very few participants (i.e., 1.2% or less) endorsed higher frequency categories. Because such low frequencies create problems for the weighted least squares mean and variance adjusted estimator, we recoded all items into four categories by combining the three highest categories. The three-factor model fit the data well, �2(116) � 435.73, RMSEA � .03, CFI � .98, TLI � .98, and improved upon the fit of the one factor model based on the RMSEA, CFI, and TLI for both boys and girls, ��2 � 174.33 and 220.26, ps � .001, �RMSEA � �.03 and �.04, �CFI � .07 and .07, �TLI � .08 and .08, respectively. We therefore used the three-factor solution.

Analysis Plan

We conducted all analyses using Mplus Version 8.0 and used full information maximum likelihood estimation to address miss- ing data. We examined six models to determine both the total and the unique relations between the three “exposure variables” (non- violent life stressors, physical victimization, and witnessing vio- lence), and the six “adjustment va