Literature Reivew

Poly-Victimization across Time in Juvenile Justice-Involved Youth Receiving Behavioral Health Treatment Fredrick Butcher , Krystel Tossone, Maureen Kishna, Jeff M. Kretschmar, and Daniel J. Flannery

Jack, Joseph and Morton Mandel School of Applied Social Sciences, Case Western Reserve University, Cleveland, Ohio, USA

ABSTRACT Diversion and community-based treatment for juveniles involved in the justice system is commonly used across jurisdictions as an alter- native to incarceration. While evidence-based approaches to assess- ment have allowed courts and treatment providers to more accurately identify areas of treatment needs, studies have rarely examined how needs change during treatment. One specific area of treatment needs common to juvenile justice-involved youth is expo- sure to violence (ETV). This study extended a previous cross-sectional study that examined the importance of contextual location on con- ceptualizing ETV by examining how ETV changes during treatment for youth involved in the justice system. Using Latent Transition Analyses, the data revealed three stable groups: (a) low ETV, (b) Home and School ETV, and (c) Poly-Location ETV. While there was little interclass movement, family history of mental health problems predicted movement from the low ETV to another class. Results confirm the importance of examining the contextual location of the exposure, the relative stability of exposure during treatment, and the importance of providing family-based behavioral health treatment for JJI youth. These findings suggest that juvenile justice-involved youth receiving community-based treatment are likely to continue to experience ETV and that treatment approaches should address this area of need.

KEYWORDS Adolescent victimization; juvenile justice; treatment; diversion programs

The juvenile justice system has undergone some major changes over the past several decades in an attempt to reduce the incarcerated population. These efforts have focused on promoting diversion and community-based treatment, as well as identifying assessment practices that match youth with services that best address the risks and needs of the individual (Andrews & Bonta, 2010; Harvell et al., 2016). There has been much interest around implementing the Risk Needs Responsivity (RNR) model in the juvenile justice system, which is designed to identify criminogenic risk in youth and match the intensity of treatment to specific needs to reduce risk for recidivism (Bonta & Andrews, 2007 Lipsey, Howell, Kelly, Chapman, & Carver, 2010; Schwalbe, 2007). While much of the literature has focused on how criminogenic risk changes during treatment and how to tailor treatment to these changes, little attention is paid to changes in factors that affect responsivity to treatment (Holloway, Cruise, Morin, Kaufman, & Steele, 2018; Taxman,

CONTACT Fredrick Butcher [email protected] Begun Center for Violence Prevention Research & Education, Jack, Joseph and Morton Mandel School of Applied Social Sciences, Case Western Reserve University, 11402 Bellflower Road, Cleveland, OH 44106-7167, USA.

VICTIMS & OFFENDERS 2020, VOL. 15, NO. 1, 22–42 https://doi.org/10.1080/15564886.2019.1680467

© 2019 Taylor & Francis Group, LLC

2014). One specific area of need that affects responsivity to treatment for many juvenile justice-involved youth is exposure to violence (ETV). ETV is an issue that affects a large proportion of youth in the juvenile justice system and may place youth at risk for future justice system involvement (Wolff & Baglivio, 2016). Increasingly, juvenile justice systems have moved toward trauma-informed treatment to reduce incarceration and prevent recidivism (Griffin, Germain, & Wilkerson, 2012; Ko et al., 2008). However, to adequately implement treatment in juvenile justice-involved youth, we must first understand how youth experience violence exposure and how best to assess and identify issues around ETV.

The current study builds on a previous study (see Butcher, Holmes, Kretschmar, & Flannery, 2016), which used latent class analysis (LCA) to examine typologies of juvenile justice-involved youth exposed to violence. Data reported in the 2016 study indicated that the contextual location in which the violence exposure occurred is an important point of assessment for juvenile justice-involved youth. Building off of this finding, the current study uses latent transition analysis (LTA) to examine whether ETV based on the typologies defined in the previous study changes during community-based treatment for youth diverted from the juvenile justice system. Latent class modeling approaches are often used to find identifiable subgroups with similar characteristics within populations that are seemingly heterogeneous (McCutcheon, 1987). Subgroups refer to smaller clusters of individuals with similarities on a given variable such as ETV. Identifying these clusters can help tailor treatment approaches to target groups of youth who can most benefit from a given type of treatment. While the focus of the first study was to better understand the conceptualization of ETV, specifically around the importance of understanding where exposure to violence occurred, the current study looks to extend this knowledge by examining how ETV changes during treatment for youth diverted from the juvenile justice system. In the previous study, we found subgroups of youth who generally had low probabilities of violence exposure, those who had higher probabilities of violence exposure in homes and school, and those who had high probabilities of ETV in the home, school, and neighborhood (Butcher et al., 2016). Given these findings, the previous study had clear implications around including questions that are location specific when screen- ing and assessing juvenile justice-involved youth for exposure to violence. Based on these conceptualizations from the previous study, the current study will examine whether youth move from subgroup to subgroup (e.g. from elevated probabilities of ETV in all three locations (home, school, and neighborhood) to low probabilities of violence exposure) during treatment and the variables that may predict this movement.

Literature review

Across the United States, youth report ETV across multiple social contexts including the home, school, and neighborhood (Butcher et al., 2016; Finkelhor, Turner, Shattuck, & Hamby, 2013; Stein, Jaycox, Kataoka, Rhodes, & Vestal, 2003; Turner, Shattuck, Finkelhor, & Hamby, 2016). Youth exposed to violence are at an increased risk of internalizing and externalizing problems (Ford, Elhai, Connor, & Frueh, 2010; Singer, Anglin, Song, & Lunghofer, 1995; Zarling et al., 2013). Particularly at risk are poly- victims who are exposed to multiple types of violent victimization and experience trauma symptomatology at a higher rate (Finkelhor, Ormrod, & Turner, 2007).

VICTIMS & OFFENDERS 23

Conceptually, research on poly-victimization distinguishes itself from earlier conceptua- lizations of ETV by focusing on the cumulative effects of multiple types of violence exposure rather than the frequency of single types of exposure (Finkelhor et al., 2007). Further adding to the complexity of conceptualizing violence exposure, poly-victims experience violence both directly as victims and indirectly as witnesses and in locations such as the home, schools, and neighborhoods (Finkelhor et al., 2013). While males and females are at a similar level of risk for experiencing violence in neighborhoods and schools (Singer et al., 1995), females report higher levels of victimization in the home (Mitchell & Finkelhor, 2001). Factors such as family history of mental health issues also raise the risks of experiencing ETV in the home (Finkelhor, Ormrod, Turner, & Holt, 2009; Tossone et al., 2015). Poly-victimization is associated with trauma symptomatol- ogy which often negatively affects the ways in which youth can build and maintain positive social relationships, ultimately changing a youth’s responsivity to treatment and increasing their risk for recidivism (Butcher, Galanek, Kretschmar, & Flannery, 2015; Holloway et al., 2018).

Widom’s (1989) work on the Cycle of Violence three decades ago has informed research around the effects of children’s maltreatment and ETV. Children who are exposed to violence are at a greater risk for perpetrating violence themselves (Maxfield & Widom, 1996; Widom, 1989) mediated by factors that can help to protect youth from the effects of ETV (Wright, Turanovic, O’Neal, Morse, & Booth, 2019). One theoretical approach to explaining the process in which ETV in youth increases the risk for violence perpetration is the social information-processing theory (Dodge, Bates, & Pettit, 1990; Widom & Wilson, 2015). Youth who experience violence directly and indirectly are more likely to misinterpret social cues and to respond aggressively (Dodge & Crick, 1990). Social interactions are also nested in the context in which they occur and interpretation and reaction to social cues takes into account the settings in which these interactions take place (Lösel, Bliesener, & Bender, 2007).

While there is generally an extensive literature around the impact of ETV, there is still considerable debate around conceptualizing ETV particularly as new and innovative meth- odologies are developed. Recently, researchers have utilized mixture modeling, a person- oriented analytical technique to classify youth based on their ETV and poly-victimization (Aebi, Giger, Plattner, Metzke, & Steinhausen, 2014; Bender, Ferguson, Thompson, & Langenderfer, 2014; Ford et al., 2010; Ford, Grasso, Hawke, & Chapman, 2013; Kretschmar, Tossone, Butcher, & Flannery, 2017; Obsuth, Mueller-Johnson, Murray, Ribeaud, & Eisner, 2017; Reid & Sullivan, 2009; Tossone et al., 2015). Mixture modeling techniques including LCA are fairly familiar in violence research and can help researchers to identify subgroups within a study population (Nurius & Macy, 2008; Swartout & Swartout, 2012). Identifying subgroups can help researchers and practitioners to make sense of data by classifying youth into different groups based on the probability of experiencing a given phenomenon. For example, one study found six subgroups of adolescents exposed to violence, with four of these groups with a high likelihood of different types of polyvictimiza- tion (e.g. polyvictimization in the community; Ford et al., 2010). Youth who were identified as experiencing multiple types of abuse and assault are at a higher risk for delinquent behaviors (Ford et al., 2010). Studies have consistently found between 3 (e.g. Aebi et al., 2014; Charak et al., 2016) and 8 (Adams et al., 2016) distinguishable subgroups depending on how ETV was measured with poly-victimization consistently found to be associated with

24 F. BUTCHER ET AL.

behavioral health problems (Adams et al., 2016; Burns, Lagdon, Boyda, & Armour, 2016; Turner et al., 2016).

While previous studies incorporating LCA on ETV have consistently and accurately identified subgroups of youth exposed to violence, few studies have examined the impor- tance of contextual location in the study of poly-victimization. Often, ETV occurs in multiple settings, which may be cumulative and particularly damaging as the threat of victimization can become pervasive in these youth’s lives (Mrug & Windle, 2010). Several studies have found subgroups of youth exposed to violence based on the contextual location in which the incidents occur (Butcher et al., 2016; Turner et al., 2016). Findings suggest that youth who experience ETV in multiple locations including the home, school, and neighborhood, termed poly-location victims, have higher levels of externalizing problems (Butcher et al., 2016). Conceptualizing ETV by contextual location may be particularly important as there may be a differential impact of ETV on childhood development based on the ecological context in which the violence occurs.

Bronfenbrenner’s (1974, 1994) work on the ecological models of human development argues that the effect of interpersonal interactions depends on factors nested in the environment. Specific to child maltreatment and violence exposure, Cicchetti and Lynch (1993) proposed the ecological-transactional model which examines the complex hierar- chy of risk and protective factors that affects a youth’s susceptibility to violence in the home and community and its effects on child development. They argue that ETV at one level (e.g. community) does not necessarily affect the risk of violence in the home. However, the interaction of the risk and effects of violence in the community with the risk and effects of violence exposure in the home is important in developmental outcomes (Cicchetti & Lynch, 1993; Spano, Rivera, Vazsonyi, & Bolland, 2008).

Accurate assessment for ETV at intake and during treatment is of particular impor- tance in juvenile justice-involved youth receiving community-based treatment. Studies have found that in comparison with community samples, juvenile justice samples con- sistently report higher prevalence rates of ETV (Ford, Hartman, Hawke, & Chapman, 2008; Wasserman & McReynolds, 2011; Wilson et al., 2013) with one study reporting that as many as 90% experience violent victimization either directly or indirectly (Abram et al., 2004). Females involved in the juvenile justice system are particularly at risk for poly- victimization (Ford et al., 2013). As part of decarceration efforts in the juvenile justice system, a number of states across the US have invested in strategies to divert youth from the justice system by providing evidence-based treatment while keeping youth away from costly and ineffective out of home placements (Harvell et al., 2016). Treating juvenile justice-involved youth in the community can potentially mean that youth who are exposed to violence in multiple settings continue to remain at risk for violence exposure during treatment. While treatment approaches should address issues around ETV (Cohen, Berliner, & Mannarino, 2000), treating juvenile justice-involved youth in nonresidential settings presents a unique challenge. Given that these youth may be at continued risk of experiencing violence during treatment, it is important that the research examines the ways in which ETV changes for youth during treatment and the factors that predict this change. Understanding ETV during treatment has strong implications for assessment, case planning, and treatment for juvenile justice-involved youth.

While studies that have identified subgroups of youth exposed to violence have important implications for understanding poly-victimization at a single time point, few

VICTIMS & OFFENDERS 25

studies have examined how these subgroups change over time. This is particularly salient for juvenile justice-involved youth who are undergoing treatment as ETV is associated with factors that greatly affect an individual’s responsivity to treatment (Ford, Hawke, Alessi, Ledgerwood, & Petry, 2007). Latent Transition Analyses (LTA) which is an exten- sion of LCA for longitudinal data have generally found that youth in community samples experience consistent levels of ETV over time. In a study of 543 African–American middle school students, Lambert, Nylund-Gibson, Copeland-Linder, and Ialongo (2010) found fairly little movement in youth exposed to violence. Students who reported either high or low levels of violence in 6th grade reported similar levels of ETV throughout middle school. Similarly, Choi and Temple (2016) found that among youth exposed to teen dating violence, there was little movement between classes over time. While these studies have examined subgroups of youth exposed to violence in community samples, little is known about how these models work in juvenile justice-involved youth undergoing treatment in the community. Understanding changes in ETV for juvenile justice-involved youth is particularly important as decarceration efforts have focused on treating youth in the community and ongoing ETV may affect approaches to treatment and recidivism. Therefore, research into the changes in ETV during treatment and the variables that predict these changes are essential to assessment and treatment planning.

Current study

In sum, juvenile justice-involved youth experience ETV and poly-victimization at a high rate. As the juvenile justice system has moved toward providing community-based treatment, it is important that we understand how youth experience ETV during treatment and how to identify those at risk for ETV during treatment. While recent studies have utilized person- oriented approaches such as latent class analysis to identify typologies of experiencing ETV, few studies have examined how these typologies change over time and during the course of treatment. The current study builds upon a previous study that conceptualized exposure to violence and polyvictimization as dependent on social context by examining the data over time to see whether subgroups of juvenile justice-involved youth based on ETV change during treatment. Further, we examine the variables that predict whether youth will transi- tion from one subgroup to another. Understanding how to identify youth who are at risk for continued violence exposure during treatment is important to treatment planning and care. The current study examines the following research questions:

1. How do subgroups of youth expose to violence change over time? 2. What is the probability that an individual will stay in the same subgroup or move to different subgroups across time?

3. Do family mental health history, gender, and age affect the probability of transition- ing to different subgroups?

Method

Study population and design

The study population consists of 1,275 juvenile justice-involved youth who successfully completed participation in the Behavioral Health Juvenile Justice Initiative (BHJJ),

26 F. BUTCHER ET AL.

a community-based diversion program in multiple sites across a Midwestern state. Counties contracted with treatment providers to provide BHJJ youth with community- based behavioral health treatment that best serves their population. To be eligible for BHJJ participation, youth were required to have a history of juvenile justice involvement, be between 10 and 18 years old, and present with at least one Diagnostic and Statistical Manual of Mental Disorders (4th ed.; DSM-IV; American Psychiatric Association, 2000) diagnosis. Success in treatment often depended upon individual goals and treatment plans, and while there is no general definition, youth were required to attend treatment sessions and meetings with caseworkers, progress in therapy and comply with the terms of the treatment plan, in addition to other county and program-specific criteria. Progress in therapy depends upon individual needs and can include, for example, improving relation- ships with family members. County-specific criteria may include the terms of probation for counties that served youth on probation while program-specific criteria include programmatic goals such as increasing prosocial activities. Each juvenile court, with the help of their service providers, was asked to identify whether a youth had completed treatment successfully. Youth who did not successfully complete the program often failed to return to treatment providers to complete termination paperwork and therefore were removed from the total sample (n = 704). Table 1 presents the characteristics of the sample, with the majority of participants being male, white, and about 15 years old. Data for the present study were collected at the county level by a trained caseworker during intake and termination interviews with the participant. The average treatment time for participants was 7.32 months (Table 1). After appropriate assent and consent by partici-

Table 1. Population characteristicsa.

Variable in Model Time 1 Time 2

Latent Class Indicator

Victim of a Threat- Home 301 (25.5%) 170 (18.1%) Witness a Threat- Home 211 (18%) 119 (9.3%)

Victim Slap, Hit or Punch- Home 438 (37.2%) 221 (23.6%) Witness Slap, Hit or Punch- Home 360 (30.6%) 197 (21.2%)

Victim of Beating- Home 109 (9.3%) 50 (5.4%) Witness Beating- Home 135 (11.5%) 67 (7.2%) Victim of a Threat- School 398 (33.8%) 208 (22.2%)

Witness a Threat- School 582 (45.6%) 316 (24.8%) Victim Slap, Hit or Punch- School 316 (26.9%) 166 (17.7%)

Witness Slap, Hit or Punch- School 685 (58.4%) 368 (39.4%) Victim of Beating- School 85 (7.2%) 28 (3%)

Witness Beating- School 642 (54.6%) 354 (37.9%) Victim of a Threat- Neighborhood 257 (21.8%) 165 (17.6%)

Witness a Threat- Neighborhood 391 (33.4%) 236 (25.3%) Victim Slap, Hit, Punch- Neighborhood 225 (19.1%) 137 (14.7%) Witness Slap, Hit, Punch- Neighborhood 482 (40.9%) 272 (29.1%)

Victim of Beating- Neighborhood 91 (7.7%) 46 (4.9%) Witness Beating- Neighborhood 438 (37.4%) 270 (28.9%)

Covariate

Family Mental Health History 782 (66.5%)

Male 723 (56.7%) Treatment Time 7.32 Months (4.5 Months) Age at Beginning of Treatment 15.1 Years (1.5 Years)

White 723 (56.8%) a Neighborhood Disorganization, Treatment Time and Age at Beginning of Treatment employ mean (standard deviation) – All others use frequency (rounded valid percent)

VICTIMS & OFFENDERS 27

pants and their caregivers, the counties sent de-identified data to researchers. The parti- cipating institution’s IRB approved all study protocols.

Measures

Exposure to violence The Recent Exposure to Violence Scale (REVS) contains 26 items that use self-report to measure ETV in the past 12 months as a witness or victim in three locations (school, home, and neighborhood; Singer et al., 1999). Previous research indicates acceptable internal consistency for this scale (Butcher, Kretschmar, Lin, Flannery, & Singer, 2014; Singer et al., 1995). Because we were interested in only the location items (18 location- specific items), we only included those in the analysis, employing the same items as a previous latent class analysis on the REVS in the BHJJ population (Butcher et al., 2016). These items examine several forms of ETV, including experiencing or witnessing threats, being slapped, hit or punched, and beatings for each location. Responses ranged between 0 (Never) and 3 (Almost Every Day). Due to cell scarcity for the 1, 2 and 3 responses, we combined these responses to make each item a dichotomous variable – 0 being an absence of that item or 1 being a presence of that item. While the original scale included two items around witnessing and being a victim of sexual abuse, the frequency of youth who responded yes to these questions was low at intake (victim 6.0%, n = 134; witness 6.2%, n = 138) and at termination (victim 3.4%, n = 37; witness 3.3%, n = 36). Preliminary analyses were conducted with the sexual abuse items in the proposed models but convergence was not achieved indicating extreme values in those variables and were removed from the models presented here. Internal consistency testing employing Cronbach’s Alpha indicated that a dichotomized version (Time 1 α = .84, Time 2 α = .87) performs similarly to the original scale (Time 1 α = .86, Time 2 α = .88).

As described in a previous study (Butcher et al., 2016), ETV at intake was conceptua- lized as a latent construct composed of three latent classes or sub-populations in the study population. These three sub-populations were dependent upon the location of the expo- sure – home, school and/or neighborhood. The Low ETV class or sub-population con- sisted of participants who have a low probability of indicating yes to any of the ETV items. The Home and School ETV class consisted of participants who have a high probability of indicating yes to any of the home items and school items, particularly being a victim to or witnessing threats or slapping, hitting or punching. The Poly-Location ETV class con- sisted of participants who have a high probability of indicating yes to the neighborhood and school items, and a somewhat higher probability of indicating yes to the home items.

Family mental health history Family mental health history was a combination of two items measured as part of the Caregiver Information Questionnaire administered during intake. The first item indicated the presence of depression in the participant’s biological family and the second items indicated the presence of any other mental health illness in the participant’s biological family. The Caregiver Information Questionnaire is administered to the participant’s caregiver. The family mental health history item was a dichotomous variable, 0 meaning no history of mental health issues and 1 being family mental health history (yes to clinical depression or yes to other mental health illness).

28 F. BUTCHER ET AL.

Other study covariates Gender was coded as Female (1) or Male (0) and was a self-report item. Due to cell scarcity in some of the race categories, race was coded as a dichotomous variable, White (1) and Nonwhite (0). Nonwhite represented African-American, Hispanic, Asian, or Other. Age at the beginning of treatment was a continuous variable representing age in years at the start of treatment. Treatment time was a continuous variable representing time in months from the beginning date of treatment to the termination date of treatment.

Analysis

This analysis seeks to build upon a previous study that examined the subclasses of juvenile justice-involved youth exposed to violence at one time point to examine changes in subclasses during treatment. For the purposes of this paper and the analyses, subclasses refer to smaller clusters of individuals who share characteristics on a given variable, in this case, exposure to violence (McCutcheon, 1987). In the previous paper, we identified three underlying subclasses with a portion of the same population studied here (Butcher et al., 2016). The following sections will describe the results of a Latent Transition Analysis (LTA), a longitudinal mixture modeling technique that examines how subclasses change over time (Collins & Lanza, 2010; Graham, Collins, Wugalter, Chung, & Hansen, 1991).

LTA is the longitudinal extension of LCA, a cross-sectional, individual-based modeling approach that identifies a latent variable composed of observed categorical items within that latent variable that is represented by heterogeneous sub-populations (McCutcheon, 1987). In contrast with other types of latent variable modeling techniques that focus on variables that can be scaled to create a construct (i.e. factor analysis), the focus of the research is on identifying groups of individuals who experience a phenomenon similarly. Furthermore, the current analysis is intended to examine how these groups change over the course of community-based treatment. As Lambert et al. (2010) note, LTA modeling is useful for measuring the model change in ETV because it allows for the examination of violence exposure over time in addition to explaining the subclasses of youth experien- cing ETV.

In the current study, the analyses sought to accomplish four objectives. First, we examined how well the model established in the previous paper fits the sample in this study at intake. The model established in the previous paper consisted of three classes of youth; low probability of ETV, higher probabilities of Home and School ETV, and higher probabilities of ETV in the Home, School, and Neighborhood. Second, the analysis sought to establish the model at termination by examining how well the intake model fits the data at termination. Third, we examined how subgroups of youth change from class to class from intake to termination. For example, youth may stay in the same ETV class from intake to termination while others move from a low probability of ETV to a high probability of ETV during treatment. Finally, the fourth objective of the analysis is to examine the factors that predict whether a youth moves or stays in a subgroup.

To accomplish these four objectives, we fit an LTA model using the three-step approach recommended by Nylund (2010) and Collins and Lanza (2010). We assessed how well the model fit by first including only the REVS variables, and then we added in the important covariates that were identified in the previous study. As with LCA, LTA relies on multiple indicators for assessing model fit, including relative information criteria, the Bayesian

VICTIMS & OFFENDERS 29

Information Criterion (BIC), Akaike’s Information Criterion (AIC), and the adjusted Bayesian Information Criterion (aBIC), entropy, to compare models against each other while favoring the most parsimonious and interpretable class solution (Geiser, 2012). These fit criteria, in addition to the Bootstrap LR Difference Test, were also employed to compare between classes when fitting the time 2 LCA model before fitting the LTA model. All LTA models employed random start values to ensure that results are global rather than only local solutions. These random starts allow researchers to see whether estimating the model multiple times results in the same or a different solution.

We fit several LTA models. First, we fit one that did not include covariates in order to obtain the baseline transitional probabilities and classification of individuals into possible classifications. As with an LCA, we examined fit statistics to determine which model fits the data best. Another important aspect of fitting an LTA model is the examination of measurement invariance from Intake to Termination. LTA measurement invariance is defined as the models, including the number and nature of classes, being expressed similarly over a period of time (Collins & Lanza, 2010). While measurement invariance is desirable, it is not necessary as we can estimate using partial measurement invariance. To estimate measurement invariance, we compare the selected LTA model using all freely estimated parameters with the same model except one where all of the parameters are set to equal each other over time. Then, we can calculate the amount that these two agree by conducting an LRT test (Perra, 2012). If the p-value is < .05, then we cannot assume measurement invariance and will fit a model that is partially non-invariant by estimating which parameters should be fixed based on theory.

Following proper LTA model fit, the second LTA model included all covariates related to Time 1 membership: gender, family mental health history, age at beginning of treat- ment, race, and treatment time in months. We hypothesized that family mental health history and treatment time in months would be related to the transition probabilities from Time 1 to Time 2; therefore, the third model included statistically significant covariates affecting Time 1 membership (gender and family mental health history) and family mental health history and treatment time in months affecting Time 1 to Time 2 membership. The tables presented reflect the multinomial logistic regression analyses from the second model and third model. Only statistically significant results are presented from the third model. Analyses were conducted in MPlus version 7.2 (Muthén & Muthén, 1998/2014) and the SAS PROC LCA and LTA Macro Version 1.3.2 (Lanza, Dziak, Huang, Wagner, & Collins, 2015).

Results

Table 1 displays the population characteristics according to the 18 Recent Exposure to Violence Scale (REVS) class indicators employed in the LCAs and LTA, and the covariates employed in the LTA model. A total of 1,094 participants had all the covariates of interest in this study. Generally, the prevalence of each REVS variable decreased from Time 1 to Time 2. Nearly two-thirds (66.5%) of successfully completed youth had a family history of mental health issues as indicated by their caregivers. Most of the successfully completed population is male (56.7%) and white (56.8%). The average age at the beginning of treatment was 15.1 years. The average treatment time was 7.3 months. We examined differences in all ETV variables in the study for successful and unsuccessful completers at

30 F. BUTCHER ET AL.

intake and found no statistically significant differences. These population characteristics were also similar to the cross-sectional model presented previously (Butcher et al., 2016).

The original Time 1 LCA presented in a previous article was conducted on all youth enrolled in BHJJ (N = 2,124; see Butcher et al., 2016). Prevalence rates for all violence exposure items are nearly identical (see Butcher et al., 2016) suggesting that the two samples are similar to the variables of interest. Considering the population characteristics including the endorsement of the REVS latent indicators were very similar for the current analyses, we used the same LCA three-class model and found nearly identical fit statistics and conditional probability plot (Figure 1). The three classes are as such: Low ETV (low probability of endorsement on all items), Home and School ETV (high probability of endorsement on home and school items), and Poly-Location ETV (high probability of endorsement on home, school and neighborhood items). For more information on the measurement portion of the Time 1 LCA model, please see Butcher et al. (2016).

While the fit criteria for the Time 2 LCA indicated a superior fit (lower AIC, BIC, aBIC; Higher entropy) for a four-class model as opposed to a three class model, the interpreta- tion of the fourth class was unclear and endorsement of the indicators were in the .4 to .5 range. This class may indicate a group of individuals that are guessing their answers, decreasing the interpretability of an optimal class solution (Geiser, 2012). Due to the poor interpretation of the fourth class, we opted to remain with the three-class solution determined in the Time 1 LCA. Figure 1 presents the Time 1 LCA (top figure) and Time 2 LCA (bottom figure) conditional probability plots for the study population. For the Low ETV class, endorsement of REVS indicators was lower at Time 2 than at Time 1. For Time 2, there appeared to be a clearer divide between the Poly-Location ETV class and the Home and School ETV class, particularly for the Home indicators.

We suspected that due to the shift in the home items and the neighborhood items for the Home and School ETV class that there might be partial measurement invariance from Time 1 to Time 2. To test for this, we compared the LTA with full measurement invariance and the LTA with freely estimated parameters and found that there was a statistically significant difference between the two models (p = .0003). Then, we compared the LTA with full measurement invariance and the LTA with freely estimated parameters for the home and neighborhood items (partial measurement invariance) and found that there was no statistically significant difference between the two models (χ2 = 32.012, df = 36, p = .30). Therefore, we ran the model with constrained parameters on the School items, and freely estimated parameters on the home and neighborhood items.

According to the results of the LTA (no covariates), the prevalence of latent classes from Time 1 to Time 2 shifts (Table 2). In Time 1, the prevalence of the latent classes is about equal, with a larger prevalence in the Poly-Location ETV class (38.1%) than in the Low ETV (33.8%) or Home and School ETV class (28.1%). This changes by Time 2, where the majority of the prevalence is in the Low ETV class (53.9%), and the other two classes each experience a prevalence decrease of about 10%. This shift in prevalence is supported by the transition probabilities demonstrated in Figure 2 (Table 3). The highest probabilities are in the cells that represent no movement from one class to a different class (i.e. Low ETV to Low ETV). Following that, the highest probabilities are from the Home and School ETV and Poly-Location ETV to Low ETV (.39). Therefore, in this population, someone has a higher likelihood of remaining in the same class from Time 1 to Time 2 than of moving to another class, particularly if that

VICTIMS & OFFENDERS 31

person is in the Low ETV class (.85). If someone is in the Home and School ETV or Poly-Location ETV class, that person has about a 40% chance of moving to the Low ETV class.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

m e

tI g

ni s

r o

d n

E f

o y

tili b

a b

o r

P

Time 1, N = 1,184

Poly Location, 37.1% Home and School, 25.6% Low ETV, 37.3%

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

m e

tI g

ni s

r o

d n

E f

o y

tili b

a b

o r

P

Time 2, N = 942

Poly Location, 16.9% Home and School, 37.6% Low ETV, 45.4%

Figure 1. Conditional item probability plots for 3-Class LCA models for Time 1 and Time 2.

32 F. BUTCHER ET AL.

Based on this model, there are nine statuses that can represent the spectrum of transition- ing (or staying) from one class to another between Time 1 and Time 2 (Table 4). Staying in the Low ETV class represents 28.6% of the study population (n = 365). While staying in the other two classes represents 32.6% of the study population, moving fromHome and School ETV or

Figure 2. Transition probabilities Time 1 to Time 2.

Table 2. Prevalence of latent classes according to time. Low ETV Home and School ETV Poly- Location ETV

Time 1 371 (33.8%) 308 (28.1%) 418 (38.1%) Time 2 592 (53.9%) 197 (18%) 308 (28.1%)

VICTIMS & OFFENDERS 33

Poly-Location ETV to Low ETV represents about 26.1% of the study population. Additionally, few move from a lower ETV class to a higher ETV class (9.3%, n = 119).

After examining the baseline LTA model, we included covariates into the model using multinomial logistic regression analysis. Table 5 presents the results of the regression analysis of the impact of covariates on Time 1 class membership. Family mental health history compared to no family mental health history (p = .0039) and gender (female compared to male; p = .0188) significantly impacted Time 1 class membership. Of particular interest is family mental health history, where those who have family mental health history have higher odds of being in the Home and School ETV class (OR = 1.93) or Poly-Location ETV class (OR = 1.45) than being in the Low ETV class. Females had lower odds of being in the Poly-Location ETV class versus the Low ETV class (OR = 0.72); however, they had slightly higher odds of being in the Home and School ETV class versus the Low ETV class (OR = 1.15) compared to males.

While we examined the impact of both statistically significant variables on the transi- tion from Time 1 to Time 2, we focused on the Odds Ratios for only family mental health history due to the interest of the study and because the data revealed generally null ORs (about 1.00) for each class comparison. Table 6 presents the Odds Ratios for family mental health history (compared to no family mental health history). Similar to the Time 1 LCA

Table 5. Multinomial logistic regression modeling Time 1 covariate impact. Change in

2*LogLikelihood Type III p-Value

Home and School ETV Odds Ratioa

Poly- Location ETV Odds Ratioa

Family Mental Health Historyb 11.09 .0039 1.93 1.45 Genderc 7.94 .0188 1.15 0.72 Treatment Time 0.17 .9174 1.00 1.00 Age at Beginning of Treatment 2.96 .2272 0.90 0.90 Raced 0.25 .8813 1.05 0.95

aClass 1 (Low ETV) is the Reference Group bNo Family Mental Health History is the Reference Group cMale is the Reference Group dNonwhite is the Reference Group

Table 3. Transition probabilities of latent class membership. Time 2

Time 1 Low ETV Home and School ETV Poly-Location ETV

Low ETV 0.85 0.03 0.11 Home and School ETV 0.39 0.46 0.15 Poly-Location ETV 0.39 0.09 0.51

Table 4. Classification of individuals based on the estimated model (N = 1,275). Time 1 Class Time 2 Class Number Rounded Percent

Low ETV Low ETV 365 28.6% Low ETV Home & School ETV 17 1.3% Low ETV Poly ETV 45 3.5% Home & School ETV Low ETV 146 11.5% Home & School ETV Home & School ETV 166 13% Home & School ETV Poly ETV 57 4.5% Poly ETV Low ETV 186 14.6% Poly ETV Home & School ETV 42 3.3% Poly ETV Poly ETV 250 19.6%

34 F. BUTCHER ET AL.

covariate analysis, we find that those who go from the Low ETV class to a higher ETV class (Home and School ETV or Poly-Location ETV) have a higher odds of family mental health history (OR = 1.26 and 1.22, respectively) than those who stay in the Low ETV class. When examining the impact of family mental health history on transition prob- ability and status membership (Table 7), we find that the transition probabilities are similar regardless of family mental health history. However, those with a family mental health history have a slightly higher probability of transitioning from Low ETV to Home and School ETV, Low ETV to Poly-Location ETV, Home and School ETV to Low ETV, Poly-Location ETV to Home and School ETV than those without a family mental health history.

Discussion

The current study extended a previous cross-sectional study by examining how subgroups of juvenile justice-involved youth exposed to violence change over time. As with the previous study (see Butcher et al., 2016), we found that location was an important distinguishing factor for youth experiencing ETV. Youth exposed to violence can be classified into three subgroups according to the contextual location in which the violence occurred: Low ETV, Home and School ETV, and Poly-Location ETV. Recent literature has shown that in addition to under- standing the cumulative nature of violence exposure, conceptualizing ETV in youth should focus on the location in which the incidents occurred and the effects of victimization across

Table 7. Prevalence of latent class membership and transition probabilities according to family mental health history.

Time Prevalence Transition Probability

Time 1 Class Time 2 Class Family Mental Health Hx

No Family Mental Health Hx

Family Mental Health Hx

No Family Mental Health Hx

Low ETV Low ETV 187 (25.6%) 128 (34.8%) .84 .88 Low ETV Home & School

ETV 13 (1.7%) 4 (1.1%) .06 .03

Low ETV Poly ETV 25 (3.4%) 14 (3.8%) .11 .09 Home & School ETV

Low ETV 90 (12.3%) 27 (7.4%) .40 .33

Home & School ETV

Home & School ETV

106 (14.5%) 37 (10.1%) .47 .46

Home & School ETV

Poly ETV 30 (4.1%) 17 (4.6%) .13 .21

Poly ETV Low ETV 96 (13.1%) 63 (17.2%) .34 .45 Poly ETV Home & School

ETV 30 (4.2%) 8 (2.2%) .11 .06

Poly ETV Poly ETV 153 (20.9%) 69 (18.8%) .55 .49

Table 6. Odds ratios predicting the effect of family mental health history on transitions from Time 1 to Time 2.a

Time 2

Time 1 Low ETV Home and School ETV Poly-Location ETV

Low ETV Reference 1.26 1.22 Home and School ETV 0.96 Reference 0.86 Poly-Location ETV 0.87 1.08 Reference

aFamily Mental Health History compared to No Family Mental Health History

VICTIMS & OFFENDERS 35

multiple social contexts (Butcher et al., 2016; Turner et al., 2016). Data presented here also demonstrated the stability of this conceptualization of ETV across time. The current study was also restricted to using only youth who successfully completed programming, and therefore, the stability in the conceptual model further confirms the importance of contextual location as a factor in conceptualizing ETV in juvenile justice-involved youth.

Consistent with findings reported by Lambert et al. (2010) and Choi and Temple (2016), data presented here suggest that subgroups of youth who experience ETV change with fairly low probability during treatment. Particularly, the Low ETV class was the least likely to change from Time 1 to Time 2. It is encouraging that subgroups of youth with a low probability of ETV at intake into treatment are likely to remain at low risk for ETV throughout treatment. Similar to the finding for youth in the Low ETV class, there was a substantial probability that youth in Poly-Location ETV class at intake would continue to experience ETV in all three contextual locations throughout treatment. These youth present a challenge for treatment providers as ETV is likely to be ongoing.

While interclass movement was fairly low for all three subgroups of youth exposed to violence, family history of mental health issues predicted movement from the Low ETV class into either the Home and School ETV class or the Poly-Location ETV class. Previous studies have found that family history of mental health issues is a predictor of ETV class membership (Butcher et al., 2016; Tossone et al., 2015). The current study builds upon these previous studies by demonstrating the importance of family mental health problems in the dynamic nature of a youth’s experience with ETV.

Strengths and limitations

Several strengths of the current study are of note. Data from the current study represent a relatively large sample of youth involved with the juvenile justice system. Behavioral health problems in juvenile justice-involved youth are still a relatively understudied issue that can have a significant impact across child-serving systems. While there are a number of cross-sectional studies on ETV in this population, data at intake and termination are fairly unique. Youth in the juvenile justice system are more likely to be from areas of high structural disadvantage and are at an increased risk for ETV (Butcher et al., 2015). Therefore, the current study design allowed for an examination of youth who report high prevalence rates of ETV across social contexts and during treatment. High prevalence rates for each of these items allowed for a clear separation of latent classes. While the sample is restricted to youth in a Midwestern state, these youth represent 11 geographi- cally diverse counties across the state.

While the current study has several strengths, findings reported here have several limitations. Youth are involved with the juvenile justice system so findings may not be generalizable to community samples. Further, ETV is a sensitive topic that participants may feel uncomfortable discussing. Self-report surveys on sensitive topics have the implicit limitation of underreporting or malingering (Butcher et al., 2014; Tourangeau & Yan, 2007). Due to data scarcity in cells, ETV items were dichotomized to reflect whether the youth had or had not experienced each type of incident in the past year. The items included in the REVS have been dichotomized in previous studies (e.g. Butcher et al., 2016). Data for this study were derived entirely of youth who successfully completed treatment. As youth often did not return to the juvenile court or treatment agency

36 F. BUTCHER ET AL.

following their unsuccessful termination, we did not have the opportunity to collect data at the second time point. However, the LCA model at intake reported in this study fit the data similarly when compared to the Butcher et al. (2016) study which included all youth participating in the BHJJ program including those who would go on to complete the program unsuccessfully. Further, it is important to note that exposure to violence is experienced by the majority of juvenile justice-involved youth. Regardless of their pro- gram completion status, the prevalence of ETV at intake was similar. The current study also does not include questions around sexual violence. While we did collect these data, there were not enough youth who answered yes for the models to converge. Finally, while the current study is an examination of the effects of change in ETV during treatment, the average length of time between intake and termination was 7 months. Longer time periods between measurement may uncover a more meaningful change in latent profiles. However, several studies using latent transition analysis examining ETV class membership across time show that ETV class membership across time is fairly stable (Choi & Temple, 2016; Lambert et al., 2010).

Conclusions

Exposure to violence is an important challenge for juvenile justice-involved youth and treatment providers as more juvenile justice systems expand diversion programming. Research on juvenile justice-involved youth have consistently recommended screening for ETV and trauma as part of the intake and screening procedures (Branson, Baetz, Horwitz, & Hoagwood, 2017; Ford, Cruise, Grasso, & Holloway, 2018; Fox, Perez, Cass, Baglivio, & Epps, 2015; Ko et al., 2008). Screening procedures can help to identify treatment needs for youth and can help systems to reduce recidivism. The current study reinforces the importance of capturing the contextual location in which the violence exposure occurs by examining ETV at two time points. This finding supports previous research that examined location-based ETV using person-oriented approaches (Butcher et al., 2016; Turner et al., 2016).

While the purpose of the current study was to examine the conceptualization of ETV during treatment and to understand how subgroups of youth experience ETV, data presented here also have significant implications for family-based treatment in juvenile justice settings. The findings here suggest that family history of mental health was associated with ETV in multiple settings throughout treatment. This may support treatment models that involve both the youth and their caregivers to build and sustain supportive relationships. Evidence supported treatment models appropriate for juvenile justice-involved youth that have been successfully implemented in the community while maintaining fidelity include Multisystemic Therapy (MST), Functional Family Therapy (FFT), and Treatment Foster Care-Oregon (TFC-O, formerly Multidimensional Treatment Foster Care, MTFC; Henggeler & Sheidow, 2012). While these treatments do not target PTSD specifically, they do address associated and common behavioral and functional issues (Mahoney, Ford, Ko, & Siegfried, 2004). A major benefit of these types of treatments is that they coach caregivers to intervene in the multiple domains of their child’s life (school, peers, and community) to varying degrees. This makes these programs extremely well-positioned to address the youth’s complex need for safety, stability, and structure, in order for caregivers to best guide the child in navigating the

VICTIMS & OFFENDERS 37

often fraught terrain of adolescence into adulthood. These treatment approaches often address issues beyond the reasons for referral, and target other co-occurring issues in order to build up protective factors that are needed to help the youth to achieve and sustain positive changes, including targeting the youth and caregivers’ substance abuse and mental health needs (Henggeler & Sheidow, 2012).

Validated and evidence-based treatments that have shown that they can reduce symp- toms of mental health issues for caregivers are particularly worth consideration for use with juvenile justice-involved youth with violence exposure and a history of family mental health issues. Employing a treatment approach that addresses caregiver mental health needs can be powerful in activating protective parenting approaches that aid in a youth’s recovery and resiliency. Data here indicated that these youth are particularly vulnerable to repeated victimization and trauma and point to the importance of providing juvenile justice-involved youth with family-based treatment approaches.

While the current study has significant implications for the assessment of ETV during community-based treatment and the factors associated with ETV, future studies should examine the impact of ETV on treatment engagement and outcomes. Juvenile justice- involved youth are likely to experience ETV across multiple social and this can have an impact on treatment outcomes. Further, the data showed a fairly high likelihood that youth who experience ETV in multiple settings will continue to experience ETV through- out community treatment. Future studies should examine how well-specific treatment modalities can help youth to build resilience and how resilience affects both treatment outcomes and future justice involvement.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This research was supported in part by grants from the Ohio Department of Youth Services and the Ohio Department of Mental Health & Addiction Services [4AS3190].

ORCID

Fredrick Butcher http://orcid.org/0000-0002-2009-2681

References

Abram, K. M., Teplin, L. A., Charles, D. R., Longworth, S. L., McClelland, G. M., & Dulcan, M. K. (2004). Posttraumatic stress disorder and trauma in youth in juvenile detention. Archives of General Psychiatry, 61, 403–410. doi:10.1001/archpsyc.61.4.403

Adams, Z. W., Moreland, A., Cohen, J. R., Lee, R. C., Hanson, R. F., Danielson, C. K., & Briggs, E. C. (2016). Polyvictimization: Latent profiles and mental health outcomes in a clinical sample of adolescents. Psychology of Violence, 6, 145–155. doi:10.1037/a0039713

Aebi, M., Giger, J., Plattner, B., Metzke, C. W., & Steinhausen, H. C. (2014). Problem coping skills, psychosocial adversities and mental health problems in children and adolescents as predictors of

38 F. BUTCHER ET AL.

criminal outcomes in young adulthood. European Child Adolescent Psychiatry, 23, 283–293. doi:10.1007/s00787-013-0458-y

American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders: DSM-IV-TR. Washington, DC: American Psychiatric Association.

Andrews, D. A., & Bonta, J. (2010). Rehabilitating criminal justice policy and practice. Psychology, Public Policy, and Law, 16, 39. doi:10.1037/a0018362

Bender, K., Ferguson, K., Thompson, S., & Langenderfer, L. (2014). Mental health correlates of victimization classes among homeless youth. Child Abuse & Neglect, 38, 1628–1635. doi:10.1016/ j.chiabu.2014.03.001

Bonta, J., & Andrews, D. A. (2007). Risk-need-responsivity model for offender assessment and rehabilitation (User Report 2007-06). Ottawa, Ontario: Public Safety Canada. doi:10.1094/ PDIS-91-4-0467B

Branson, C. E., Baetz, C. L., Horwitz, S. M., & Hoagwood, K. E. (2017). Trauma-informed juvenile justice systems: A systematic review of definitions and core components. Psychological Trauma: Theory, Research, Practice, and Policy, 9, 635–646. doi:10.1037/tra0000255

Bronfenbrenner, U. (1974). Developmental research, public policy, and the ecology of childhood. Child Development, 45, 1–5. doi:10.2307/1127743

Bronfenbrenner, U. (1994). Ecological models of human development. In M. Gauvian & M. Cole. (Eds.), Readings on the development of children (4th ed., pp. 37–43). New York, NY: Worth Publishers.

Burns, C. R., Lagdon, S., Boyda, D., & Armour, C. (2016). Interpersonal polyvictimization and mental health in males. Journal of Anxiety Disorders, 40, 75–82. doi:10.1016/j. janxdis.2016.04.002

Butcher, F., Galanek, J. D., Kretschmar, J. M., & Flannery, D. J. (2015). The impact of neighborhood disorganization on neighborhood exposure to violence, trauma symptoms, and social relation- ships among at-risk youth. Social Science & Medicine, 146, 300–306. doi:10.1016/j. socscimed.2015.10.013

Butcher, F., Holmes, M. R., Kretschmar, J. M., & Flannery, D. J. (2016). Polyvictimization across social contexts: Home, school, and neighborhood violence exposure. Criminal Justice and Behavior, 43, 1726–1740. doi:10.1177/0093854816662679

Butcher, F., Kretschmar, J. M., Lin, Y., Flannery, D. J., & Singer, M. I. (2014). Analysis of the validity scales in the trauma symptom checklist for children. Research on Social Work Practice, 24, 695–704. doi:10.1177/1049731513516803

Charak, R., Byllesby, B. M., Roley, M. E., Claycomb, M. A., Durham, T. A., & Ross, J., ... & Elhai, J. D. (2016). Latent classes of childhood poly-victimization and associations with suicidal behavior among adult trauma victims: Moderating role of anger. Child Abuse & Neglect, 62, 19–28. doi:10.1016/j.chiabu.2016.10.010

Choi, H. J., & Temple, J. R. (2016). Do gender and exposure to interparental violence moderate the stability of teen dating violence? Latent transition analysis. Prevention Science, 17, 367–376. doi:10.1007/s11121-015-0621-4

Cicchetti, D., & Lynch, M. (1993). Toward an ecological/transactional model of community violence and child maltreatment: Consequences for children’s development. Psychiatry, 56, 96–117. doi:10.1080/00332747.1993.11024624

Cohen, J. A., Berliner, L., & Mannarino, A. P. (2000). Treating traumatized children: A research review and synthesis. Trauma, Violence, & Abuse, 1, 29–46. doi:10.1177/1524838000001001003

Collins, L. M., & Lanza, S. T. (2010). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. New York, NY: Wiley.

Dodge, K. A., Bates, J. E., & Pettit, G. S. (1990). Mechanisms in the cycle of violence. Science, 250, 1678–1683. doi:10.1126/science.2270481

Dodge, K. A., & Crick, N. R. (1990). Social information-processing bases of aggressive behavior in children. Personality and Social Psychology Bulletin, 16, 8–22. doi:10.1177/ 0146167290161002

Finkelhor, D., Ormrod, R., Turner, H., & Holt, M. (2009). Pathways to poly-victimization. Child Maltreatment, 14, 316–329. doi:10.1177/1077559509347012

VICTIMS & OFFENDERS 39

Finkelhor, D., Ormrod, R. K., & Turner, H. A. (2007). Poly-victimization: A neglected component in child victimization. Child Abuse & Neglect, 31, 7–26. doi:10.1016/j.chiabu.2006.06.008

Finkelhor, D., Turner, H. A., Shattuck, A., & Hamby, S. L. (2013). Violence, crime, and abuse exposure in a national sample of children and youth: An update. JAMA Pediatrics, 167, 614–621. doi:10.1001/jamapediatrics.2013.42

Ford, J. D., Cruise, K. R., Grasso, D. J., & Holloway, E. (2018). A study of the impact of screening for poly-victimization in juvenile justice: The rocky road to a successful investigation in the real world. Journal of Interpersonal Violence, 33, 810–831. doi:10.1177/0886260517744844

Ford, J. D., Elhai, J. D., Connor, D. F., & Frueh, B. C. (2010). Poly-victimization and risk of posttraumatic, depressive, and substance use disorders and involvement in delinquency in a national sample of adolescents. Journal of Adolescent Health, 46, 545–552. doi:10.1016/j. jadohealth.2009.11.212

Ford, J. D., Grasso, D. J., Hawke, J., & Chapman, J. F. (2013). Poly-victimization among juvenile justice-involved youths. Child Abuse & Neglect, 37, 788–800. doi:10.1016/j.chiabu.2013.01.005

Ford, J. D., Hartman, J. K., Hawke, J., & Chapman, J. F. (2008). Traumatic victimization, posttrau- matic stress disorder, suicidal ideation, and substance abuse risk among juvenile justice-involved youth. Journal of Child & Adolescent Trauma, 1, 75–92. doi:10.1080/19361520801934456

Ford, J. D., Hawke, J., Alessi, S., Ledgerwood, D., & Petry, N. (2007). Psychological trauma and PTSD symptoms as predictors of substance dependence treatment outcomes. Behaviour Research and Therapy, 45, 2417–2431. doi:10.1016/j.brat.2007.04.001

Fox, B. H., Perez, N., Cass, E., Baglivio, M. T., & Epps, N. (2015). Trauma changes everything: Examining the relationship between adverse childhood experiences and serious, violent and chronic juvenile offenders. Child Abuse & Neglect, 46, 163–173. doi:10.1016/j.chiabu.2015.01.011

Geiser, C. (2012). Data analysis with Mplus. New York, NY: Guilford Press. Graham, J. W., Collins, L. M., Wugalter, S. E., Chung, N. K., & Hansen, W. B. (1991). Modeling

transitions in latent stage-sequential processes: A substance use prevention example. Journal of Consulting and Clinical Psychology, 59, 48–57. doi:10.1037/0022-006X.59.1.48

Griffin, G., Germain, E. J., & Wilkerson, R. G. (2012). Using a trauma-informed approach in juvenile justice institutions. Journal of Child and Adolescent Trauma, 5, 271–283. doi:10.1080/ 193615212012.697100

Harvell, S., Welsh-Loveman, J., Love, H., Durnan, J., Eisenstat, J., Golian, L., & La Vigne, N. (2016). Reforming sentencing and corrections policy: The experience of justice reinvestment initiative states. Washington, DC: Urban Institute.

Henggeler, S. W., & Sheidow, A. J. (2012). Empirically supported family-based treatments for conduct disorder and delinquency in adolescents. Journal of Marital and Family Therapy, 38, 30–58. doi:10.1111/j.1752-0606.2011.00244.x

Holloway, E. D., Cruise, K. R., Morin, S. L., Kaufman, H., & Steele, R. D. (2018). Juvenile probation officers’ evaluation of traumatic event exposures and traumatic stress symptoms as responsivity factors in risk assessment and case planning. Law and Human Behavior, 42, 369–384. doi:10.1037/lhb0000283

Ko, S. J., Ford, J. D., Kassam-Adams, N., Berkowitz, S. J., Wilson, C., Wong, M., & Lyne, C. M. (2008). Creating trauma-informed systems: Child welfare, education, first responders, health care, juvenile justice. Professional Psychology: Research and Practice, 39, 396–404. doi:10.1037/0735- 7028.39.4.396

Kretschmar, J. M., Tossone, K., Butcher, F., & Flannery, D. J. (2017). Patterns of poly-victimization in a sample of at-risk youth. Journal of Child and Adolescent Trauma, 10, 363–375. doi:10.1007/ s40653-016-0109-9

Lambert, S. F., Nylund-Gibson, K., Copeland-Linder, N., & Ialongo, N. S. (2010). Patterns of community violence exposure during adolescence. American Journal of Community Psychology, 46, 289–302. doi:10.1007/s10464-010-9344-7

Lanza, S. T., Dziak, J. J., Huang, L., Wagner, A., & Collins, L. M. (2015). PROC LCA & PROC LTA (Version 1.3.2) [Software]. University Park, PA: The Methodology Center, Penn State. Retrieved from http://methodology.psu.edu

40 F. BUTCHER ET AL.

Lipsey, M. W., Howell, J. C., Kelly, M. R., Chapman, G., & Carver, D. (2010). Improving the effectiveness of juvenile justice programs. Washington, DC: Center for Juvenile Justice Reform at Georgetown University.

Lösel, F., Bliesener, T., & Bender, D. (2007). Social information processing, experiences of aggres- sion in social contexts, and aggressive behavior in adolescents. Criminal Justice and Behavior, 34, 330–347. doi:10.1177/0093854806295833

Mahoney, K., Ford, J. D., Ko, S. J., & Siegfried, C. B. (2004). Trauma-focused interventions for youth in the juvenile justice system. Retrieved from https://www.nctsn.org/resources/trauma-focused- interventions-youth-juvenile-justice-system

Maxfield, M. G., & Widom, C. S. (1996). The cycle of violence: Revisited 6 years later. Archives of Pediatrics & Adolescent Medicine, 150, 390–395. doi:10.1001/archpedi.1996.02170290056009

McCutcheon, A. L. (1987). Latent class analysis. Newbury Park, CA: SAGE Publications, Inc. Mitchell, K. J., & Finkelhor, D. (2001). Risk of crime victimization among youth exposed to

domestic violence. Journal of Interpersonal Violence, 16, 944–964. doi:10.1177/ 088626001016009006

Mrug, S., & Windle, M. (2010). Prospective effects of violence exposure across multiple contexts on early adolescents’ internalizing and externalizing problems. Journal of Child Psychology and Psychiatry, 51, 953–961. doi:10.1111/j.1469-7610.2010.02222.x

Muthén, L. K., & Muthén, B. O. (1998/2014). Mplus user’s guide (7th ed.). Los Angeles, CA: Muthén & Muthén.

Nurius, P. S., & Macy, R. J. (2008). Heterogeneity among violence-exposed women applying person-oriented research methods. Journal of Interpersonal Violence, 23, 389–415. doi:10.1177/ 0886260507312297

Nylund, K. (2010). Latent transition analysis: Modeling extensions and an application to peer victimization (Doctoral Dissertation, University of California Los Angeles). Retrieved from http://www.statmodel.com/download/Nylund%20dissertation%20Updated1.pdf

Obsuth, I., Mueller-Johnson, K., Murray, A. L., Ribeaud, D., & Eisner, M. (2017). Violent poly- victimization: The longitudinal patterns of physical and emotional victimization throughout adolescence (11–17 years). Journal of Research on Adolescence, Advanced online publication. doi:10.1111/jora.12365

Perra, O. (2012). Latent transition analysis [PowerPoint Slides]. Retrieved from http://www.restore. ac.uk/latentvariablemodels/workshopfiles/Latent%20Transition%20Analysis_Oliver%20Perra/ Course%20Presentation%20and%20Exercises/LTA_course_UU.pdf

Reid, J. A., & Sullivan, C. J. (2009). A latent class typology of juvenile victims and exploration of risk factors and outcomes of victimization. Criminal Justice and Behavior, 36, 1001–1024. doi:10.1177/ 0093854809340621

Schwalbe, C. S. (2007). Risk assessment for juvenile justice: A meta-analysis. Law and Human Behavior, 31, 449. doi:10.1007/s10979-006-9071-7

Singer, M. I., Anglin, T. A., Song, L. Y., & Lunghofer, L. (1995). Adolescents` exposure to violence and associated symptoms of psychological trauma. JAMA, 273, 477–482. doi:10.1001/ jama.1995.03520300051036

Singer, M. I., Miller, D. B., Guo, S., Flannery, D. J., Frierson, T., & Slovak, K. (1999). Contributors to violent behavior among elementary and middle school children. Pediatrics, 104, 878–884. doi: 10.1542/peds.104.4.878

Spano, R., Rivera, C., Vazsonyi, A. T., & Bolland, J. M. (2008). The impact of exposure to violence on a trajectory of (declining) parental monitoring: A partial test of the ecological-transactional model of community violence. Criminal Justice & Behavior, 35, 1411–1428. doi:10.1177/ 0093854808321735

Stein, B. D., Jaycox, L. H., Kataoka, S., Rhodes, H. J., & Vestal, K. D. (2003). Prevalence of child and adolescent exposure to community violence. Clinical Child and Family Psychology Review, 6, 247–264. doi: 1096-4037/03/1200-0247/0

Swartout, K. M., & Swartout, A. G. (2012). Shifting perspectives: Applying person-centered analyses to violence research. Psychology of Violence, 2, 309–312. doi:10.1037/a0029910

VICTIMS & OFFENDERS 41

Taxman, F. S. (2014). Second generation of RNR: The importance of systemic responsivity in expanding core principles of responsivity. Federal Probation, 78, 32–40.

Tossone, K., Jefferis, E. S., Grey, S. F., Bilge-Johnson, S., Bhatta, M. P., & Seifert, P. (2015). Poly- traumatization and harmful behaviors in a sample of emergency department psychiatric intake response center youth. Child Abuse & Neglect, 40, 142–151. doi:10.1016/j.chiabu.2014.11.015

Tourangeau, R., & Yan, T. (2007). Sensitive questions in surveys. Psychological Bulletin, 133, 859–883. doi:10.1037/0033-2909.133.5.859

Turner, H. A., Shattuck, A., Finkelhor, D., & Hamby, S. (2016). Polyvictimization and youth violence exposure across contexts. Journal of Adolescent Health, 58, 208–214. doi:10.1016/j. jadohealth.2015.09.021

Wasserman, G. A., & McReynolds, L. S. (2011). Contributors to traumatic exposure and posttrau- matic stress disorder in juvenile justice youths. Journal of Traumatic Stress, 24, 422–429. doi:10.1002/jts.20664

Widom, C. S. (1989). The cycle of violence. Science, 244, 160–166. doi:10.1126/science.2704995 Widom, C. S., & Wilson, H. W. (2015). Intergenerational transmission of violence. In J. Lindert &

I. Levav (Eds.), Violence and mental health (pp. 27–46). Dordrecht, Netherlands: Springer. Wilson, H. W., Berent, E., Donenberg, G. R., Emerson, E. M., Rodriguez, E. M., & Sandesara, A.

(2013). Trauma history and PTSD symptoms in juvenile offenders on probation. Victims & Offenders, 8, 465–477. doi:10.1080/15564886.2013.835296

Wolff, K. T., & Baglivio, M. (2016). Adverse childhood experiences, negative emotionality, and pathways to juvenile recidivism. Crime and Delinquency, 1–27. doi:10.1177/0011128715627469

Wright, K. A., Turanovic, J. J., O'Neal, E. N., Morse, S. J., & Booth, E. T. (2019). The cycle of violence revisited: childhood victimization, resilience, and future violence. Journal of Interpersonal Violence, 34(6), 1261–1286. doi:10.1177/0886260516651090

Zarling, A. L., Taber-Thomas, S., Murray, A., Knuston, J. F., Lawrence, E., Valles, N. L., & Bank, L. (2013). Internalizing and externalizing symptoms in young children exposed to intimate partner violence: Examining intervening processes. Journal of Family Psychology, 27, 945–955. doi:10.1037/a0034804

42 F. BUTCHER ET AL.

Copyright of Victims & Offenders is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.