Complex Trauma and Criminogenic Needs in a Youth Justice Sample: A Gender-Informed Latent Profile Analysis

Abstract

How complex trauma features and criminogenic needs co-vary within youth justice populations requires examination. This study applies latent profile analysis to a sample of 311 justice-involved Canadian youth (211 male, 100 female) to identify if unique profiles of youth would emerge delineated by different combinations of comorbid needs pulled from complex trauma and personality/social learning models. Two similar profiles emerged for males and females alike: a complex trauma with criminogenic needs profile (70% of females, 58.8% of males) and a low overall needs profile (30% of females, 41.2% of males). Surprisingly, the Youth-Level Service/Case Management Inventory predicted recidivism well among the complex trauma/criminogenic need female cases (AUC = .71), but poorly among the complex trauma/criminogenic need male cases (AUC = .59). Trauma-informed approaches that target criminogenic needs in both genders is a clear implication of the findings.

Keywords

adolescence trauma criminogenic needs delinquency gender differences recidivism latent profile analysis

Although youth-perpetrated crime continues to drop in the United States, the proportion of arrests attributed to female youth continues to grow, rising from 20% in 1985 to 30% in 2015 (Puzzanchera & Ehrmann, 2018). In Canada, youth-perpetrated crime is declining in both genders. However, the decrease is greater among males than females (Kong & AuCoin, 2008; Savage, 2019). Notably, it is well established that the apparent increase and/or relatively slower decrease in the female crime rate relative to males is largely a reflection of legal and policy changes rather than changes in actual behavior (Brown et al., 2019). At the same time, gender-focused criminological research has grown exponentially.

Historically, explanatory models of delinquency and crime as well as correctional assessment and intervention approaches have been grounded in decidedly gender-neutral paradigms such as social learning theory (Akers, 1973), personality-based models (Gottfredson & Hirschi, 1990), and developmental life course models (Loeber & Farrington, 2000). Feminist-driven scholarship has changed this landscape. A new generation of gender-responsive scholars is increasingly studying justice-involved girls and/or women (sometimes alongside male comparison groups) using a range of qualitative and quantitative approaches. Unlike mainstream psychological perspectives, prevailing gender-responsive theories emphasize interpersonal trauma, addictions, maladaptive relationships, and impoverished social capital (Bloom et al., 2003; Salisbury et al., 2016). Concomitantly, the study of psychological trauma has evolved considerably since the American Psychiatric Association (APA) introduced posttraumatic stress disorder (PTSD) into the Diagnostic and Statistical Manual of Mental Disorders: Third Edition (DSM-III, 1980). Increasingly, leading trauma experts now recognize that strict adherence to the definitional confines of PTSD precludes more complex forms of trauma resulting from chronic exposure to interpersonal adversity (van der Kolk, 2005).

Gender-responsive and gender-neutral scholars are now regularly exploring heterogeneity within justice-involved samples using person-centered as opposed to variable-centered approaches (Brennan et al., 2012; Lanctôt, 2018; Schwalbe et al., 2008). These approaches attempt to determine if latent or unobserved heterogeneity exists (e.g., different subtypes/profiles/classes of justice-involved people) within seemingly homogeneous groups. This small but significant body of research underscores the diversity within justice-involved samples of females as well as males. However, more work is needed to elucidate the nature and number of latent profiles, and to clarify if different profiles exist who are best characterized by complex trauma (CT) symptomatology or social learning/personality features. This is particularly important to investigate among youth as gender-responsive scholars have been slow to differentiate between adult and adolescent female samples. Using latent profile analysis (LPA; Lazarsfeld & Henry, 1968), this article explores if latent subtypes of justice-involved youth can be delineated using variables derived from CT and social learning models.

Complex Trauma (CT) Theory

The original features of PTSD encapsulated intrusive memories, avoidant/numbing symptoms, and hyperarousal symptoms all resulting from a single event “outside the range of usual human experience” (p. 236, DSM-III; APA, 1980). This definition has been criticized for being unable to capture the complex array of trauma catalysts and ensuing sequalae routinely observed in clinical practice (Briere, 2013). Trauma experts (Ford & Courtois, 2009; Herman, 1992; van der Kolk, 2005) have explained how chronic, interpersonal trauma such as prolonged childhood abuse and neglect or chronic violence and/or forced displacement experienced during war, captivity, or genocide can disrupt: (a) the normal development of secure attachments within caregiving relationships, and (b) core self-regulation competencies related to learning/attention, memory, and emotions. In turn, a myriad of devastating sequalae ensue including but not limited to substance misuse, anxiety, depression, aggression, and impaired daily functioning.

The APA does not recognize CT in the Diagnostic and Statistical Manual of Mental Disorders: 5th Edition (DSM-5; APA, 2013). However, the World Health Organization (WHO) now differentiates between PTSD and complex PTSD in the latest version of the International Classification of Diseases (ICD-11; WHO, 2018). The ICD-11 identifies the following six symptom clusters indicative of complex PTSD: (a) re-experiencing the traumatic event(s), (b) actively avoiding potentially triggering situations, (c) heightened sense of threat/hyperarousal, (d) affect dysregulation, (e) negative self-concept, and (f) difficulties in relationships (Brewin et al., 2017).

Ford et al. (2012) have theorized how complex PTSD can lead to delinquency and aggression vis-à-vis a direct and indirect pathway (the terms CT [in line with Ford and other trauma experts] and complex PTSD [in line with the ICD-11] are used interchangeably throughout the article). In the indirect pathway, complex PTSD is hypothesized to disrupt the normal development of both secure attachment to primary caregivers and self-regulation (attention/learning, memory, and emotional). In turn, attachment and self-regulation deficits escalate the probability of youth experiencing risk factors associated with delinquency and aggression. These risk factors include: (a) oppositional and conduct disorder, (b) risk-taking and impulse control issues, (c) substance misuse, (d) impaired information processing, (e) maladaptive thinking, and (f) delinquent peer association. In the direct pathway, diminished arousal reactions coupled with maladaptive hyperarousal are hypothesized to result in reactive episodes of anger/rage, and assaultive behaviors that are then criminalized. Arguably, the emergent CT paradigm provides an elegant framework to potentially unify gender-responsive pathways perspectives.

Gender-Responsive Pathways Theory of Female-Perpetrated Crime

A parallel body of research collectively known as feminist pathways research (Chesney-Lind, 1997; Daly, 1992) or more simply, pathways theory (Belknap, 2014) emerged in the 1990s to explain female crime. To date, about 25 pathways studies (sample sizes: 16 to 300) have identified one prominent (but not the only) pathway to the justice system for girls and women. This particular pathway illustrates how aversive family environments (e.g., abuse, neglect) propel girls from their homes and schools into the streets. In turn, this leads to further victimization in the form of prostitution and poor coping strategies such as drug use, and criminalized survival strategies such as robbery, fraud, and drug trafficking (see Belknap, 2014 for a detailed review). Daly (1992), one of the most commonly cited pathways researchers identified the following five classes of women based on thematic analyses of presentence reports: (a) street women characterized by abuse in the home, which leads to running away and living on the streets, prostitution, and substance abuse; (b) harmed and harming women characterized by experiences of abuse and psychological harm in childhood with the inability to cope; (c) drug-connected women characterized using and selling drugs with family members and intimate partners; (d) battered women characterized by involvement in an abusive relationship with a romantic partner; and (e) other/economically motivated women whereby crime is motivated by the need for a comfortable and secure lifestyle. Notably two of Daly’s classes (i.e., street women, harmed, and harming women) dovetail seamlessly with the CT model that has emerged independently within clinical psychology and psychiatry.

A Gender-Neutral Theory of Crime: General Personality and Cognitive Social Learning

The general personality and cognitive social learning (GPCSL) theory of crime (Bonta & Andrews, 2017) is a multidisciplinary perspective with psychology at the forefront. Grounded heavily in social learning and self-control theories, GPCSL theorists posit that an individual will engage in crime when the rewards for doing so outweigh the costs. Factors that tip the balance in favor of criminal rewards include situational factors such as opportunity, intoxication, and intense emotions such as anger, as well as person-centered factors such as self-regulation deficits, cognitions favorable to crime, social support for crime/criminal associates, an antisocial personality pattern, and having a history of antisocial behaviors. Interpersonal-level (family factors) and community-level variables (school, neighborhood) also figure prominently in the GPCSL theory of crime, but to a lesser extent. GPCSL theorists posit that structural-level factors such as age, socio-economic status, gender and race influence how rewards and costs are distributed within a given group, but they do not aid in explaining individual differences in criminal conduct.

The GPCSL perspective does share some commonalities with CT pathways models. For example, each paradigm underscores the importance of regulation deficits, negative emotions, substance misuse, and familial deficits. However, unlike pathways and CT models, the GPCSL does not emphasize the importance of chronic childhood abuse/instability or attachment dysfunction as key catalysts of criminal conduct. Nor does the GPCSL identify mental health or self-concept variables as criminogenic treatment priorities. In addition, unlike pathways and CT theorists, GPCSL theorists prioritize antisocial peers, antisocial cognitions, and an antisocial personality pattern as treatment targets.

Thus, depending on one’s perspective, correctional treatment targets and approaches will vary. Pathway and complex PTSD theorists advocate for trauma-informed and relational based approaches that prioritize self-concept, mental health, healthy relationships, regulation deficits, and problem solving, whereas GPCSL theorists advocate for the use of cognitive behavioral approaches that prioritize criminal thinking, criminal peers, and substance misuse as well as regulation deficits and problem-solving skills. The extent to which different clusters of justice-involved youth—male and female alike—may reflect a CT subtype, a personality/social learning type (i.e., GPCSL) or some hybrid requires investigation. Thus, this study examines heterogeneity within a youth justice sample using LPA, a person-centered approach (Lazarsfeld & Henry, 1968).

The Person-Centered Approach in Justice Samples

Over a decade ago, Reisig et al. (2006) illustrated that the predictive validity of the Level of Supervision Inventory—Revised (LSI-R; Andrews & Bonta, 1995)—a gender-neutral risk assessment tool—was reduced in a subsample of American women probationers. In particular, the LSI-R did not predict recidivism well for women who the researchers had qualitatively classified as following a pathway to crime characterized by abuse, poverty, and mental health challenges. In contrast, the LSI-R performed as expected for women classified as economically motivated (a pathway described as masculine). This finding has yet to be replicated. Furthermore, the extent to which the Reisig et al. results will generalize to a justice-involved youth sample that includes males as well as females or will replicate using a quantitative classification technique like LPA requires exploration.

Person-centered approaches like LPA are advantageous for two main reasons. First, they can identify unobserved homogeneous classes of individuals who, based on their shared characteristics, may require differential treatment approaches. Second, they can help unify two seemingly disparate theories of crime such as the gender-responsive CT pathway, and the gender-neutral personality/social learning model (i.e., GPCSL) by examining both perspectives in a complimentary holistic fashion.

Person-centered studies using LPA or LCA (latent class analysis) have become increasingly prominent in the justice field, particularly for girls. To date, this small, but burgeoning literature has collectively identified between three and five classes (or profiles) of seemingly homogeneous groups of justice-involved adolescents, irrespective of gender. One class is typically always a low-risk/low-need group characterized by low-offending patterns and/or low needs, and another class is typically a serious and chronic offending group characterized by a broad spectrum of needs (adolescent female studies: Lanctôt, 2018; Odgers et al., 2007; mixed sex adolescent studies: Campbell et al., 2019; Whitney et al., 2010). The remaining classes tend to be more heterogenous ranging from subtypes of individuals characterized by a victimized pathway with co-occurring addictions and mental health needs (adolescent female study: Cusworth-Walker et al., 2016) or a peer and/or addictions group with none, or only low levels of mental health and victimization histories (adolescent female study: Cusworth-Walker et al., 2016; mixed sex adolescent study: Schwalbe et al., 2008). All of the extant studies have been cross-sectional in nature with one exception. Henneberger et al. (2014) followed up 109 American adolescent females who participated in an earlier LCA study conducted by Odgers et al. (2007). Odgers et al.’s seminal LCA analysis identified three groups of females: (a) violent and delinquent (e.g., truancy, theft), (b) delinquent only, and (c) low-level violent/delinquent group. Three to six years later, Henneberger et al.’s follow-up study revealed that the violent/delinquent subgroup experienced the greatest level of impairment during young adulthood including high levels of violent recidivism and internalizing psychopathology.

While it is apparent that there is a small, yet chronic class of adolescent offending females, it is unknown whether or not this chronic subtype can be further differentiated. In particular, it is unknown whether or not all chronic offending females (or males) are best conceptualized as CT cases, or if a subtype exists whom are better characterized by antisocial personality features alongside social learning–based risk factors such as criminal cognitions and criminal associates. Previous LPA/LCA analyses involving justice-involved girls (and boys) have also yet to include measures encapsulated by the CT paradigm and social learning and personality-based models simultaneously. Furthermore, few LPA studies (see Odgers et al., 2007 for an exception) have included measures of affect regulation and attachment—key CT constructs. Finally, whether latent profiles will impact the accuracy of traditional risk tools has not been examined since Reisig et al. (2006).

Study Objectives and Hypotheses

This study explores unobserved heterogeneity among justice-involved youth using variables informed by both CT models and personality and social learning models. If unobserved heterogeneity does exist, the study will further explore: (a) if the resultant latent profiles differ as a matter of degree (e.g., low, moderate or high risk) and/or in kind (e.g., different combination of risk and need profiles emerge) and (b) the validity of the resultant latent profiles in terms of the extent to which they vary as a function of theoretically relevant factors such as race, internalizing and externalizing problems, criminal history, and criminal recidivism. Based on previous studies, it is hypothesized that two to four latent profiles will emerge. It is also hypothesized that at least one latent profile will be comprised of a lower need group while one additional latent profile will be characterized by CT symptoms; it is also expected that females will be over-represented in the CT profile.

Method

Participants

Three hundred and eleven consenting justice-involved youth (n_females = 100, 32.2%; n_males = 211, 67.8%) participated in the study. Participants were recruited from six different sites in Ontario, Canada, that service justice-involved youth (i.e., one probation office, two secure custodial facilities, two open custodial facilities, and one mental health center). The sample ranged in age from 12 to 21 (M = 16.8, SD = 1.3); notably 87.5% of the sample was between the ages of 15 and 18. Age did not vary as a function of gender (females: M = 16.7, SD = 1.2 vs. males: M = 16.9, SD = 1.3, t [309] = 1.4, p = .15, d = .18). Participants self-identified as: White (45.3%, n = 141), Black (30.2%, n = 94), mixed racial heritage (5.5%, n = 17), Indigenous (4.8%, n = 15), Middle/Far East (4.8%, n = 15), Asian (3.2% n = 10), and Hispanic (1.9% n = 6). Four percent (n = 13) opted not to self-identify. Based on a collapsed race variable (i.e., White, Black, other), there were significant gender differences (χ² = 17.9, p = .001, Phi = .25) in ethnic distributions with proportionately more White females (65.0%, n = 64) than White males (38.7%, n = 77), and more Black males (37.2%, n = 74) than Black females (20.0%, n = 20). At the time of the assessment, 42.8% (n = 133) of the sample had been remanded to custody pending trial outcome or sentencing determination, while 57.6% (n = 179) had been adjudicated.

At the time of the original assessment, participants had been charged with and/or convicted with the following offenses: homicide-related (4.2%), serious violence (e.g., kidnapping/forcible confinement, robbery, assault; 53.5%), sex-related (9.3%), less serious person-related (e.g., criminal harassment, uttering threats; 11.2%), weapons-related (15.1%), property/theft/fraud-related (34.0%), drug trafficking (5.1%), drug possession (4.8%), and administration of justice/obstruction of justice (55.8%). Significant gender differences emerged for sexual offenses (n_females = 1 [3.4%] vs. n_males = 28 [13.3%]; χ² = 12.2, p = .001, Phi = −.20), and weapons-related offenses (n_females = 4 [4.0%] vs. n_males = 43 [20.5%]; χ² = 14.5, p = .001, Phi = −.22). Overall, the sample had committed serious offenses and was not representative of the larger youth justice population in Canada. The typical Canadian justice-involved youth has committed a minor, nonviolent crime (e.g., theft) and has either been diverted from the justice system or has been sanctioned in the community (Allen & Superle, 2016).

Measures

The following seven constructs were used to capture complex PTSD as defined by the ICD-11 (WHO, 2018) and Ford et al. (2012): (a) Adverse Childhood Experiences (ACEs), (b) PTSD, (c) Affect Regulation Deficits, (d) Learning Regulation Deficits, (e) Avoid Attachment Difficulties, (f) Anxious Attachment Difficulties, and (g) Poor Self-concept. The following seven central needs as identified within the GPCSL model were also measured: (a) Criminal Associates, (b) Criminal Attitudes, (c) Psychopathy, (d) Education-Related Problems, (e) Family-Related Problems, (f) Substance Misuse, and (g) Poor Use of Leisure Time. Auxiliary variables included age, race, general offense history, number of violent index charges and/or convictions, and internalizing and externalizing problems. One distal outcome was included: official recidivism.

Demographics and Criminal History Measures

Demographic (e.g., age, sex, and race) and criminal history variables were coded using a combination of data. On-site file reviews were coded first. Next, electronic data collected directly from the Ministry of Children and Youth Services Ontario were retrieved to address any gaps.

Achenbach Youth Self-Report

The Youth Self-Report (YSR) has been extensively validated (Achenbach, 1991). It consists of 112 self-report items tapping emotional and behavioral problems within the past 6 months. Items are rated on a 3-point Likert-type scale (0 = Not true, 1 = Somewhat true, 2 = Often true). The following YSR subscales were used in subsequent analyses: PTSD proxy (males: α = .82, females: α = .83), attention problems (males: α = .79, females: α = .80), internalizing problems (males: α = .87, females: α = .89), and externalizing problems (males: α = .91, females: α = .90). The lesser known PTSD proxy subscale is comprised of 14 YSR items; sample items include: I have nightmares, I am too fearful or anxious, and, I cannot get my mind off certain thoughts.

Adolescent Relationship Scales Questionnaire

The Adolescent Relationship Scales Questionnaire (ARSQ) consists of 17 self-report items measuring four dimensions of attachment: secure, fearful, preoccupied, and dismissing, each rated on a 7-point Likert-type scale (1 = Not at all like me, 4 = Somewhat like me, 7 = Very much like me) (Scharfe & Bartholomew, 1995). Observed alphas were low for each subscale (.08 > α < .69). Low observed alphas coupled with the paucity of published validation on the ARSQ necessitated a principal component analysis (PCA) be conducted using all 17 items; varimax rotation with parallel analysis was used. A resultant 2-component solution (accounting for 38.9% of the variance) emerged. Component #1: anxious attachment was characterized by six items such as: I worry that I will be hurt if I become too close to others (males: α = .72, females: α = .77). Component #2: avoidant attachment was characterized by four items such as: I find it hard to count on other people (males: α = .85, females: α = .76).

Rosenberg Self-Esteem Scale

The Rosenberg Self-Esteem Scale (RSES) is a validated self-esteem measure comprised of 10 self-report items focusing on overall feelings of self-worth and self-acceptance (Rosenberg, 1965). Items are rated on a 4-point Likert-type scale (0 = Strongly disagree to 3 = Strongly agree; plausible range: 0 to 30). Higher scores are indicative of greater self-esteem (males: α = .83, females: α = .87).

ACEs

Ten dichotomously scored ACEs (e.g., physical abuse, sexual abuse, parental divorce; plausible range: 0 to 10) were coded by proxy from the Youth Assessment and Screening Instrument (YASI; Felitti et al., 1998; Orbis Partners, 2000). This method has demonstrated validity (Baglivio et al., 2015). ACEs were reliably coded in the study [males: α = .74, females: α = .74; inter-rater reliability estimate based on 21 cases: one-way random single Intraclass Correlation Coefficient (ICC) = .88].

ARC

The Affect Regulation Checklist (ARC) is a 12-item self-report scale with three subscales assessing affect control, affect suppression, and adaptive reflection (Moretti, 2003). Each item is measured on a 3-point Likert-type scale (0 = not like me to 2 = a lot like me; plausible range: 0 to 8). The study used the 4-item affect control subscale (males: α = .78; females: α = .88); sample items include: I have a hard time controlling my feelings, and It is very hard for me to calm down when I get upset.

Measures of Criminal Attitudes and Associates

The 12-item (dichotomously scored) attitudes toward violence subscale (ATV; plausible range: 0 to 12; sample items include: It is understandable to hit someone who insults you, someone who makes you very angry deserves to be hit) from the Measures of Criminal Attitudes and Associates (MCAA) was used to assessed attitudes toward violence (Mills et al., 2002). Previous research supports the ATV’s validity with both justice-involved girls and boys (O’Hagan et al., 2018). Higher scores reflect attitudes supportive of violence (males: α = .83, females: α = .83).

Youth Level of Service—Case Management Inventory, 2.0 (YLS-CMI, 2.0)

The YLS-CMI is a validated 42-item, interview-based tool (scored by the interviewer) to assess risk to reoffend and criminogenic needs among youth (Hoge & Andrews, 2011). The study used the YLS-CMI total score (males: α = .89, females: α = .87, single inter-rater ICC = .81) and the following subdomains: prior/current offenses (males: α = .78, females: α = .72; single inter-rater ICC = .84), education/employment (males: α = .71, females: α = .58; single inter-rater ICC = .33), peer relations (males: α = .69, females: α = .64; single inter-rater ICC = .55), substance abuse (males: α = .75, females: α = .76; inter-rater reliability estimate based on 21 cases: one-way random single ICC = .85), family/parenting (males: α = .57, females: α = .46; single inter-rater ICC = .71), and poor use of leisure time (males: α = .50, females: α = .49; single inter-rater ICC = .84).

Hare Psychopathy Checklist: Youth Version (PCL: YV)

The well-validated 20-item PCL: YV—scored by the rater from a semiclinical interview and file review—was used to assess psychopathic traits; PCL: YV total scores were used in the analyses (Forth et al., 2003). Items are measured on a three-point scale (0 = No; 1 = Maybe; 2 = Yes). Higher scores are associated with more prototypical levels of psychopathy (males: α = .89, females: α = .82, inter-rater reliability estimate based on 21 cases: one-way random single ICC =. 74).

Recidivism

Youth and adult reconviction information was coded from official recidivism records obtained from the province of Ontario (Ministry of Community Safety and Correctional Services and the Ministry of Children and Youth Services), and the federal policing agency (Royal Canadian Mounted Police [RCMP]) to ensure coverage of youth and adult criminal offenses in Ontario, as well as across the country. Recidivism was coded as any new criminal conviction (excluding technical violations) (coded: 0 = No, 1 = Yes) occurring over a 3-year fixed follow-up period. 56.9% the sample recidivated with males significantly more likely to recidivate than females (62.0% vs. 46.5%, χ² = 6.57, p = .01, Phi = −.15)

Procedure

After all requisite ethics and legal clearances were obtained, consenting participants were interviewed either by a trained researcher or a clinician (at the CAMH site only). Custody participants received $30.00 worth of canteen money and community participants received $30.00 in gift certificates. The assessment process took approximately 8 to 10 hours per participant; this included face-to-face interviews, on-site file reviews, self-report questionnaire administration, and instrument scoring where applicable (e.g., PCL-YV, YLS/CMI).

Statistical Approach

Using maximum likelihood procedures, LPA generates latent groups (profiles) alongside mean estimates for each observed variable within each latent profile (Lazarsfeld & Henry, 1968; Muthén & Muthén, 1998–2017). LPA assumes that population-based correlations among variables are a by-product of variations in variable means among unobserved heterogeneity within the population. The resultant profiles are hypothesized to have substantive meaning in the sense that there are theoretical reasons to expect that individuals will behave differently depending upon profile membership in terms of antecedents (e.g., etiological pathways) and distal outcomes. Males were over-represented by a ratio of 2 to 1 in the sample; thus, two separate LPAs were conducted for each gender to ensure that any male-specific results did not eclipse any potential female-specific results; power implications are explored and discussed.

Results

Models consisting of one through five profiles were estimated in Mplus, version 8.1 (Muthén & Muthén, 1998–2017) using the robust maximum likelihood estimator. Models were also estimated separately for each gender. The resultant number of profiles was determined based on multiple statistical criteria, a priori-hypotheses, parsimony, and meaningfulness. Statistical criteria included the following fit indices: (a) Akaike Information Criterion (AIC), (b) Bayesian Information Criterion (BIC), and (c) sample size adjusted Bayesian Information Criterion (SSA-BIC). Noteworthy, lower fit indices are indicative of a better fitting model. Entropy was also used as a measure of classification accuracy with values approaching 1 indicative of stronger classification (values >.80 are indicative of highly discriminating latent profiles; Muthén & Muthén, 1998–2017). Finally, likelihood ratio (LR) statistical tests were used. These tests compare the fit of the K₀-profile model to the K₋₁ profile model (where K₋₁ < K₀). Smaller p values indicate that the K₀-profile model is a significantly better fit to the observed data than the K₋₁ profile model. The following LR tests were used: (a) adjusted Lo–Mendell–Rubin (LMR) test, and (b) Vuong–Lo–Mendell–Rub likelihood ratio test (VLMRT). There is no firm agreement among statistical experts regarding which statistical index is superior. Although simulation studies do suggest that the SSA-BIC, LMR, and the VLMR are among the strongest (Nylund et al., 2007) and that the AIC and entropy values are consistently unreliable (Tein et al., 2013).

As Table 1 illustrates, the VLMR and the LMR tests support a two-profile solution for both males and females. Notably the corresponding entropy values were also strong (.89 for females, .85 for males). The entropy values and AIC, BIC, and SSA-BIC fit indices also supported a four- and/or five-profile solution in both genders. However, the 2-profile solution was deemed more reliable, given that AIC indices and entropy values have demonstrated poor reliability irrespective of sample size, degree of profile separation, or number of indicators (Tein et al., 2013). See Figure 1 for a graphical representation of the profiles and Table 1 for mean values associated with each profile. Corresponding Cohen’s d values representing the degree of profile separation between indicators are also presented in Table 2. The correlations between the 14 LPA indicators ranged from −.17 to .74 in the males, and from −.32 to .74 in the females.

Table 1:

Latent Profile Analyses Model Fit Results by Gender

Profile #	AIC	BIC	SSA-BIC	Entropy	VLMR	LMR
Female (n = 100)
1	6,355.67	6,428.62	6,340.19	na	na	na
2	6,227.99	6,340.01	6,204.21	.89	−3,149.84**	155.43**
3	6,159.80	6,310.90	6,127.72	.88	−3,070.99 ns	96.79 ns
4	6,127.77	6,317.95	6,087.89	.90	−3,021.90 ns	61.14 ns
5	6,103.87	6,333.13	6,055.20	.92	−2,990.87 ns	53.13 ns
Male (n = 211)
1	13,017.12	13,110.97	13,022.25	na	Na	na
2	12,652.62	12,796.75	12,660.50	.85	−6,480.56***	389.64***
3	12,544.63	12,749.04	12,565.26	.79	−6,283.31 ns	126.42 ns
4	12,476.37	12,721.05	12,489.74	.86	−6,219.31 ns	108.26 ns
5	12,455.95	12,750.91	12,472.07	.89	−6,165.18 ns	49.80 ns

Note. AIC = Akaike information criterion; BIC = Bayesian information criterion; SSA-BIC = sample size adjusted BIC; VLMR = Vuong–Lo–Mendell–Rub likelihood ratio test; LMR = Lo–Mendell–Rubin test.

p < .01. ***p < .001.

Table 2:

Total Sample and Latent Profile Means (M), Standard Deviations (SDs) and Cohen’s ds by Gender

Variable	Total	Males (N = 211)			Females (N = 100)
	Total	CPTSD and Criminogenic	Low overall needs	d	CPTSD and Criminogenic	Low overall needs	d
	(N = 311)M (SD)	(n = 124)M (SD)	(n = 87)M (SD)	d	(n = 70)M (SD)	(n = 30)M (SD)	d
PTSD	10.77 (5.94)	11.92 (4.63)	6.34 (4.09)	1.28	15.27 (4.62)	6.77 (2.92)	1.22
ACEs	4.05 (2.72)	4.28 (2.28)	1.53 (1.58)	1.40	6.18 (2.29)	4.00 (2.76)	0.86
Anxious attachment	21.55 (8.51)	20.98 (7.40)	17.03 (6.52)	0.57	30.31 (6.24)	18.31(5.75)	1.99
Avoidant attachment	19.75 (5.26)	19.75 (5.34)	18.63 (5.63)	0.21	20.52 (4.86)	20.00 (4.50)	0.11
Self-esteem	20.35 (5.14)	19.88 (4.44)	22.78 (4.50)	0.65	16.70 (4.16)	24.00 (4.68)	1.65
Dysregulation-emotions	3.65 (2.51)	3.60 (2.25)	2.23 (1.98)	0.65	5.20 (2.19)	2.71 (2.56)	1.05
Dysregulation-attention	7.22 (3.77)	8.30 (3.39)	4.69 (3.35)	1.07	10.13 (2.96)	3.96 (2.41)	2.29
Education issues	2.69 (1.69)	3.90 (1.84)	2.37 (2.13)	0.77	3.24 (1.63)	1.97 (1.65)	0.78
Substance misuse	2.27 (1.61)	3.36 (1.38)	1.02 (1.05)	1.91	2.30 (1.81)	2.13 (1.48)	0.10
Family/parenting issues	2.58 (1.44)	3.45 (1.31)	1.45 (1.17)	1.62	3.52 (1.42)	2.63 (1.69)	0.57
Violent attitudes	6.12 (3.44)	7.16 (3.08)	4.87 (3.16)	0.74	5.83 (3.41)	5.00 (3.22)	0.25
Criminal associates	2.61 (1.27)	3.20 (0.98)	1.77 (1.20)	1.30	2.65 (1.27)	2.43 (1.19)	0.18
Psychopathy	16.82 (7.07)	19.43 (6.93)	10.83 (6.62)	1.27	16.64 (6.86)	13.81 (5.91)	0.44
Poor use of leisure	1.75 (0.93)	2.11 (0.69)	1.39 (0.91)	0.89	1.67 (0.92)	1.40 (0.89)	0.29

Note. SD = standard deviation; CPTSD = complex posttraumatic stress disorder; d = Cohen’s d; ACE = adverse childhood experiences.

Figure 1:

Resultant Latent Profiles

The two male profiles are best described as a (a) complex PTSD with criminogenic needs subgroup, henceforth known as complex needs, and a (b) low overall needs subgroup, henceforth known as low needs. The Complex Needs subgroup is characterized by male youth scoring high on complex PTSD symptoms: anxious attachments, emotional and attentional regulation deficits, ACEs, PTSD, low self-esteem, and high on all seven criminogenic needs: education, substance misuse, family/parenting problems, violent attitudes, criminal associates, poor use of leisure time, and psychopathy. In contrast, the low needs subgroup evidenced comparably lower levels on all complex PTSD and criminogenic indicators. Importantly, with the exception of avoidant attachment (d = 0.21) the magnitude of the mean differences between all of the profile indicators was considerably high (0.57 > d < 1.91). Notably 10 of the 14 indicators evidenced profile separations characterized as large (0.66 > d < 1.00) or very large (d > 1.00).

The two resultant female profiles: (a) complex PTSD with criminogenic needs (henceforth known as complex needs) and (b) low overall needs (henceforth know as low needs) mirrored the male profiles with two key distinctions. First, with the exception of avoidant attachment (d = 0.11), the degree of separation between the female profiles was stronger between the complex PTSD indicators (0.86 > d < 2.29; average d = 1.51) in comparison to the corresponding male profiles (average d = 0.91). Also noteworthy, members of the female complex needs profile scored particularly high on anxious attachment and particularly low on self-esteem relative to both the male complex needs and the female low needs subgroups. Second, the degree of separation between criminogenic indicators was substantially larger between the male profiles (0.74 > d < 1.91; average d = 1.21) than the female profiles (0.10 > d < 0.78; average d = 0.37). Unlike the males, there was very little difference between the female profiles in terms of substance misuse, criminal associates, and poor use of leisure time (0.10 > d < 0.29). Furthermore, while the differences in education, family/parenting, and psychopathy between the female profiles were large (.44 > d <.78), they were substantially smaller in comparison to the males (.77 > d < 1.62). Overall, although the discrimination between both profiles averaged across all 14 indicators for the females was large (d = 0.84), it remained comparably smaller relative to the male profiles (d = 1.02). See Table 3 for a gender by profile breakdown.

Table 3:

Resultant Latent Profile Distributions by Gender

Latent profile	Males, % (n/211)	Females, % (n/100)
Complex PTSD with criminogenic needs	58.8 (124)	70.0 (70)
Overall low needs	41.2 (87)	30.0 (30)

Note. PTSD = posttraumatic stress disorder.

Next, the meaningfulness and practical utility of the profiles were assessed using a series of t-tests, chi-square analyses, and receiver operator characteristic analysis in IBM SPSS, version 25. These analyses examined the extent to which the latent profiles were related to six auxiliary variables: age, race, internationalizing and externalizing problems, criminal offense history (YLS/CMI offense history domain score), and number of violent index convictions. The analyses also examined the relationship between latent profile membership and one distal outcome, recidivism.

As Table 4 illustrates, only three of the six auxiliary variables were significantly related to profile membership. Interestingly, the same pattern emerged in both genders. Relative to the low needs subgroup, both male and female participants classified as complex needs evidenced more extensive offense histories as measured by the YLS/CMI’s criminal history domain. However, the effect was more pronounced for males (d = 0.91) than females (d = 0.42). Similarly, both male and female complex needs profiles evidenced more internalizing and externalizing problems as measured by the YSR (Achenbach, 1991) in comparison to their low need profile counterparts. The magnitude of the externalizing effect however was stronger for males (d = 1.65) than females (d = 1.13). Conversely the magnitude of the internalizing effect was stronger for females (d = 1.33) than males (d = 0.95).

Table 4:

Latent Profiles by Auxiliary Variables Separated by Gender

Auxiliary variables	Females (N = 100)				Males (N = 211)
	Complex needs (n = 70)	Low needs (n = 30)	t	d	Complex needs (n = 124)	Low needs (n = 87)	t	d
	M (SD)	M (SD)	t	d	M (SD)	M (SD)	t	d
Age	16.57 (1.22)	16.95 (1.19)	1.49 ns	0.31	16.90 (1.01)	16.93 (1.58)	0.19ns	0.02
Offense History	2.03 (1.50)	1.41 (1.47)	1.92*	0.42	2.59 (1.64)	1.17 (1.47)	−6.39***	0.91
#Violent offenses	0.97 (0.93)	0.87 (1.11)	−0.49 ns	0.10	1.25 (1.07)	1.02 (0.96)	−0.20ns	0.23
Internalizing^a	27.97 (10.89)	9.43 (6.48)	−8.40***	1.33	18.17 (9.58)	10.01 (7.45)	−8.16***	0.95
Externalizing^b	30.33 (10.37)	19.00 (9.61)	−4.97***	1.13	29.00 (9.19)	13.99 (8.98)	−15.00***	1.65
	Females (N = 99)^c				Males (N = 199)^d
	Complex needs, % (n/70)	Low needs, % (n/29)	χ²	V	Complex needs, % (n/118)	Low needs, % (n/81)	χ²	V
Race
White	68.6 (48)	55.2 (16)	5.34 ns	.23	64.9 (50)	35.1 (27)
Black	14.3 (10)	34.5 (10)			56.8 (42)	43.2 (32)	1.74 ns	.09
Other	17.1 (12)	10.3 (3)			54.2 (26)	45.8 (22)

Note. SD = standard deviation.

Youth Self-Report (YSR) internalizing problem score. ^bYSR externalizing problem score. ^cRace was missing for one female youth. ^dRace was missing for 12 male youths.

p< .05. ***p <.001.

Although race was not a significant predictor of profile membership in either gender, it is important to underscore that 68.6% of White females were classified as complex needs, versus only 14.3% and 17.1% of Black and Other females, respectively. In contrast, the proportion of males classified as complex needs was generally consistent across race (White = 64.9%, Black = 56.8%, Other = 54.2%). Finally, the results clearly illustrated no association between profile membership and age, nor the number of violent index charges/convictions. These results were consistent across gender.

Next, four separate AUC analyses were conducted to examine the extent to which the predictive validity of the YLS/CMI total score would vary as a function of profile membership. As Table 5 illustrates, profile membership influenced the predictive accuracy of the YLS/CMI total score, and the results varied as function of gender. For females, the YLS/CMI total score predicted recidivism more accurately among the complex needs profile cases (AUC = .71) than the low needs profile cases (AUC = .65). Albeit, as expected given the small sample sizes, the 95% confidence intervals did overlap substantially. The reverse pattern was found among males. For males, the YLS/CMI total score predicted recidivism more accurately among the low need profile cases (AUC = .64) than the complex need profile cases (AUC = .59). However, once again it is important to underscore the overlapping 95% confidence intervals.

Table 5:

Two Latent Profile Solution Predicting Recidivism: Receiver Operator Characteristic and Chi-Square Results by Gender

Profile	Recidivated
Profile	N	AUC (SE)	95% CI	% (n/N)	χ²	V ^a
Females
Complex needs	70	.71 (.06)	[0.59, 0.83]	45.70 (32/70)	0.93 ns	0.009
Low needs	30	.65 (.11)	[0.45, 0.86]	46.70 (14/30)
Males
Complex needs	124	.59 (.06)	[0.48, 0.71]	73.30 (88/120)	15.91***	0.280
Low needs	87	.64 (.06)	[0.52, 0.76]	45.90 (39/85)

Note. AUC = area under the curve; SE = standard error; CI = confidence intervals.

Cramer’s V.

***

p < .001.

Finally, a chi-square analysis examining the relationship between recidivism and profile membership was significant for males but not females (Table 5). Noteworthy, the complex needs male subtype evidenced the highest recidivism rate (73.3%). In contrast, the recidivism rates were lower, as well as similar among the remaining profile groups (range: 45.7%–46.7%). In addition, for females, YLS/CMI total mean scores did not vary significantly (t = −1.77, df = 94, p = .08) as a function of profile membership: complex needs (M = 20.65, SD = 7.57), low needs (M = 17.70, SD = 7.60), albeit the magnitude of the difference was classified as moderate (Cohen’s d = 0.39). However, YLS/CMI total mean scores did vary significantly for males (t = −17.01, df = 203, p < .0001) as a function of profile membership: complex needs (M = 25.78, SD = 4.26), low needs (M = 13.23, SD = 6.31). The size of difference was classified as very large (Cohen’s d = 2.33).

Discussion

This study explored heterogeneity in a Canadian sample of justice-involved girls and boys using the increasingly popular LPA technique. LPA was conducted separately for males and females. In both genders two similar latent profiles emerged: (a) complex needs (complex PTSD with criminogenic needs) and (b) overall low needs (no CT /low criminogenic needs). Interestingly, both females and males were more likely to be classified as complex needs versus low needs. However, the proportion of females classified as complex needs was slightly higher (70.0% vs. 58.8% of the males). Surprisingly, the proportion of males in our sample falling in the complex needs profile was notably higher than those reported by Charak et al. (2019). Charak et al. reported an overall prevalence rate of 20% for poly-victimization cases with emotional regulation deficits in a justice-involved sample of male and female youth. However, similar to our results, Charak et al. also found that 64.8% of the girls were classified as poly-victimization/emotional regulation deficit cases. As hypothesized, we also found an overall low needs group as previously reported in similar studies (Lanctôt, 2018; Odgers et al., 2007). Notably, 30% of the females and 41% of the males were classified as low needs. The small magnitude of this gender difference was somewhat unexpected, given past theorizing and evidence that justice-involved females have more needs than males, albeit not necessarily criminogenic (Salisbury et al., 2016).

It is possible that our study may have been underpowered to detect more complex profiles and hence more nuanced gender differences. Tein et al.’s (2013) LPA power simulation study suggests that when the size of separation between profile indicators averages in the range of d = 0.80, a sample size of 500 is most likely required to accurately identify the correct number of profiles. Although our study evidenced moderate to large average separation between profiles for both males (average Cohen’s d = 1.02) and females (average Cohen’s d = 0.84), our sample size fell considerably short of 500 for males and females alike.

Notwithstanding power limitations, there were some notable gender differences within the complex needs profiles. Specifically, relative to their male counterparts, the female complex needs profile evidenced considerably higher levels of anxious attachment, PTSD, ACEs, attentional and emotional dysregulation, and lower levels of self-esteem. Conversely, with the exception of family problems, the male complex need profile scored higher than their female counterparts on all criminogenic need areas (i.e., education, substance misuse, violent attitudes, criminal associates, psychopathy, and poor use of leisure time). In addition, the auxiliary analyses revealed that CT cases were also associated with higher levels of internalizing and externalizing problems as measured by the Youth Self-Report (Achenbach, 1991) in both genders. However, the magnitude of the externalizing effect was stronger for males. Collectively, these results are consistent with the position that females, in general, are lower risk than their male counterparts in terms of risk-elevating criminogenic factors, but yet, are higher need on factors that are not (or only weakly) predictive of recidivism (Salisbury et al., 2016).

Contrary to Reisig et al. (2006), our study illustrated that the YLS/CMI predicted recidivism well among complex need female cases (AUC = .71), whereas the predictive accuracy of the YLS/CMI was lower among the low need female cases (AUC = .65). The opposite pattern emerged for the males; the YLS/CMI predicted recidivism better among the low need males (AUC = .64) and worse among the complex need male cases (AUC = .59). Whether these findings were the result of random sampling bias, developmental differences (Reisig et al., used an adult sample), restricted variance issues, reliance on a statistical classification approach as opposed to Reisig et al.’s qualitative classification approach, or simply an accurate reflection of the data requires further investigation using larger, more diverse samples.

Recidivism likelihood did vary as a function of profile membership—but only for males. Specifically, the complex needs male group was significantly more likely to recidivate (73.3%) than their low needs counterparts (45.9%). In contrast, the recidivism rates between the complex needs female profile and the low needs female profile were virtually identical, 45.7% and 46.7%, respectively. These results are expected given that males in the complex needs group scored particularly high on criminogenic needs relative to their low need counterparts.

Implement Universal Trauma-Informed Services in Youth Justice Settings

The results support the need for trauma-informed services in youth justice settings for both genders as almost 60% of the male youth and 70% of the female youth in this sample were classified as complex PTSD cases with co-occurring criminogenic needs. The presence of CT symptoms coupled with criminogenic needs elevated the risk to recidivate and minimized the predictive accuracy of the YLS/CMI only in males; unexpectedly these findings were not observed in females. Nonetheless, the exceedingly high prevalence rates of CT symptoms in males and females alike, has implications for the responsivity principle (Bonta & Andrews, 2017). Specifically, system-wide services that realize what trauma is, and how it impacts behavior, recognize the signs of trauma, and resist practices that are potentially re-traumatizing are needed (Substance Abuse and Mental Health Services Administration, 2014). Specific examples may include: (a) system-wide screening for trauma coupled with correctional plans that match trauma severity to an appropriate response level, (b) staff training for all justice personnel (e.g., teachers, parole officers, correctional officers/wardens), not just mental health staff, (c) correctional programs focused on criminogenic needs that are trauma-informed (i.e., the main target may be substance misuse, but the program recognizes the impact of trauma and minimizes re-traumatizing participants), and (d) mental health treatment where the primary target is the trauma itself.

Holistic, Trauma-Informed Services Are Consistent With the RNR Framework

Advocating for trauma-informed services is not in conflict with correctional treatment practices embedded within the RNR framework. This approach embodies the specific responsivity principle as originally intended by its architects (Bonta & Andrews, 2017) and is in no way incompatible with the general responsivity principle. General responsivity emphasizes social learning approaches such as cognitive behaviorism but also underscores the need for respectful and empathic interactions which are key tenets of trauma-informed services. Moreover, cognitive behavior therapies are increasingly being fused with emotions-focused therapies to address emotional regulation deficits (e.g., trauma-focused cognitive behavioral therapy [TF-CBT]; Cohen et al., 2010, 2017) suggesting that it is no longer an “either”, “or” paradigm). Finally, the bridge between trauma and responsivity principles can be further fortified by conceptualizing traditional criminogenic treatment targets (e.g., emotional dysregulation, substance misuse) as proximally observed sequalae to distally occurring ACEs.

The growing practice of fusing seemingly divergent treatment modalities is highly recommended. Our study, as well as others (Odgers et al., 2007), have found that externalizing and internalizing symptoms co-occur for a significant subset of justice-involved youth, particularly females. These results also speak to the need to implement holistic treatment regimens that simultaneously target both internalizing and externalizing symptoms.

Limitations and Future Directions

Despite the study’s strengths, several cautions are in order. We argue that our findings support the need for holistic, trauma-informed services that simultaneously target features of CT and criminogenic needs. However, the implications for etiological pathways are less clear. Our results are based on the retrospective recall of youth, and as such, temporal order cannot be established based on our research design. Furthermore, our sample was small and comprised of a relatively unique sample of youth. A disproportionate number of these youth had serious index offenses and were at higher risk to reoffend in comparison to typical youth in the justice system. In addition, our measurement battery was imperfect. For example, the PTSD proxy subscale was relatively crude, and our attachment and emotional regulation measures were relatively untested and require further validation. Specifically, it was assumed that both avoidant and anxious attachment styles would figure prominently among CT cases. But this was not the case, for males or females. Only anxious attachment styles defined CT cases, whereas avoidant attachments styles did not. As well, our recidivism definition was based solely on official as opposed to self-report information. Finally, our analyses were most likely underpowered in both genders due to sample size, and thus were most likely unable to detect more complex profiles.

It is important to note that the distinction between latent profiles is one of relativity whereby a given profile is said to score higher on variable X relative to another profile. Future LPA analyses should explore whether or not the observed relative differences in mean scores also translate into meaningful practical implications. Importantly, future research should explore how statistically derived typologies compare to clinically built typologies such as those in the DSM. More research regarding the intersection of race, gender, and typology building is also required. For example, the finding that female Black youth or other non-White female youth were less likely to be classified within the CT /criminogenic need group was unexpected and requires further exploration.

In closing, the results support the need to consider both criminogenic and CT symptomology within samples of justice-involved youth, males and females alike. Importantly, this research underscores the dangers of over-stating gender differences. When we erroneously assume that trauma is exclusively the purview of justice-involved females, we risk over-looking the very real possibility that trauma also permeates the lives of justice-involved males. Thus, like trauma-informed approaches that have been developed specifically for girls/women, the extent to which trauma-informed approaches that have been developed specifically for boys/men (e.g., Helping Men Recover; Covington et al., 2011) require serious consideration and rigorous evaluation.

Footnotes

Authors’ Note:

Leigh Greiner is now at British Columbia Corrections, Ministry of Public Safety and Solicitor General. We have no conflicts of interests associated with this research. This research was funded by the Social Sciences Humanities Research Council, Grant 76034, Principal Investigator: Shelley L. Brown & co-investigator: Tracey A. Skilling.

ORCID iD

Shelley L. Brown

Shelley L. Brown has been an associate professor in the Department of Psychology, Carleton University, Ottawa, Canada since 2010. Shelley studies gender differences in complex trauma, violence, strengths, and risk assessment practices among justice-involved populations.

Kayla A. Wanamaker received her PhD in 2020 from the Psychology Department at Carleton University in Ottawa, Canada. Her doctoral research assessed patterns of change in dynamic risk and strength factors and the relationship to reoffending outcomes for justice-involved men and women. She also examined typologies of men and women on community supervision that can be used to inform service delivery and rehabilitation efforts.

Leigh Greiner received her PhD in 2016 from the Psychology Department at Carleton University, Ottawa, Canada. Leigh is currently the Director of Research and Strategic Planning, British Columbia Corrections, Ministry of Public Safety and Solicitor General.

Terri Scott received her PhD in 2017 from the Psychology Department at Carleton University in Ottawa, Canada. Her doctoral research examined risks, strengths, and gender differences among justice-involved youth.

Tracey A. Skilling is a psychologist and clinician scientist at the Center for Addiction and Mental Health and an assistant professor in the Department of Psychiatry at the University of Toronto with a graduate appointment in the Department of Applied Psychology and Human Development. Her research interests include antisocial behavior, mental health issues, substance use in adolescents, and risk assessment issues in both male and female young offenders. Her current projects include studies evaluating risk assessment practices and understanding the desistance process among justice-involved male and female offenders.

References

Achenbach

T. M.

(1991). Manual for the youth self-report and 1991 profile. University of Vermont.

Akers

R. L.

(1973). Deviant behavior: A social learning approach. Wadsworth.

Allen

Superle

(2016). Youth crime in Canada, 2014 (Juristat Catalogue No. 85-002-X). Canadian Centre for Justice Statistics, Statistics Canada. https://www150.statcan.gc.ca/n1/en/pub/85-002-x/2016001/article/14309-eng.pdf?st=maKumDdM

American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.).

American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). American Psychiatric Publishing.

Andrews

D. A.

Bonta

(1995). The Level of Supervision Inventory—Revised (LSI-R). Multi-Health Systems.

Baglivio

M. T.

Wolff

K. T.

Piquero

A. R.

Epps

(2015). The relationship between adverse childhood experiences (ACE) and juvenile offending trajectories in a juvenile offender sample. Journal of Criminal Justice, 43, 229–241. https://doi.org/10.1016/j.jcrimjus.2015.04.012

Belknap

(2014). The invisible woman: Gender, crime and justice (4th ed.). Cengage Learning.

Bloom

Owen

Covington

S. S.

(2003). Gender-responsive strategies for women offenders: Research, practice, and guiding principles for women offenders. U.S. Department of Justice, National Institute of Corrections. https://s3-amazonaws-com-443.web.bisu.edu.cn/static.nicic.gov/Library/018017.pdf

10.

Bonta

Andrews

D. A.

(2017). The psychology of criminal conduct (6th ed.). Routledge.

11.

Brennan

Breitenbach

Dieterich

Salisbury

E. J.

van Voorhis

(2012). Women’s pathways to serious and habitual crime: A Person-centered analysis incorporating gender responsive factors. Criminal Justice and Behavior, 39, 1481–1508. https://doi.org/10.1177/0093854812456777

12.

Brewin

C. R.

Cloitre

Hyland

Shevlin

Maercker

Bryant

R. A.

Humayun

Jones

L. L.

Kagee

Rousseau

Somasundaram

Suzuki

Wessely

Ommeren

Reed

G. M.

(2017). A review of current evidence regarding the ICD-11 proposals for diagnosing PTSD and complex PTSD. Clinical Psychology Review, 58, 1–15. https://doi.org/10.1016/j.cpr.2017.09.001

13.

Briere

(2013). Treatment of complex trauma: A sequenced, relationship-based approach. Guilford Press.

14.

Brown

S. L.

Blanchette

K. D.

Dean

R. J.

(2019). Are females really more violent and more criminal than ever? In Morgan

R. D.

(Ed.), The Sage encyclopedia of criminal behavior. SAGE.

15.

Campbell

C. A.

Miller

Papp

Barnes

A. R.

Onifade

Anderson

V. R.

(2019). Assessing intervention needs of juvenile probationers: An application of latent profile analysis to a risk–need–responsivity assessment model. Criminal Justice and Behavior, 46, 82–100. https://doi.org/10.1177/0093854818796869

16.

Charak

Ford

J. D.

Modrowski

C. A.

Kerig

P. K.

(2019). Polyvictimization, emotion dysregulation, symptoms of posttraumatic stress disorder, and behavioral health problems among justice-involved youth: A latent class analysis. Journal of Abnormal Child Psychology, 47(2), 287–298. https://doi.org/10.1007/s10802-018-0431-9

17.

Chesney-Lind

(1997). The female offender: Girls, women and crime. SAGE.

18.

Cohen

J. A.

Berliner

Mannarino

(2010). Trauma focused CBT for children with co-occurring trauma and behavior problems. Child Abuse & Neglect, 34(4), 215–224. https://doi.org/10.1016/j.chiabu.2009.12.003

19.

Cohen

J. A.

Mannarino

A. P.

Deblinger

(2017). Treating trauma and traumatic grief in children and adolescents (2nd ed.). Guilford Press. https://doi.org/10.1080/07317107.2017.1375719

20.

Covington

S. S.

Griffin

Dauer

(2011). Helping men recover: A program for treating addiction. Wiley.

21.

Cusworth-Walker

Bishop

A. S.

Nurius

P. S.

Logan-Greene

(2016). The heterogeneity of treatment Needs for justice-involved girls: A typology using latent class analysis. Criminal Justice and Behavior, 43(3), 323–342. https://doi.org/10.1177/0093854815615162

22.

Daly

(1992). Women’s pathways to felony court: Feminist theories of lawbreaking and problems of representation. Southern California Review of Law and Women’s Studies, 2(1), 11–52.

23.

Felitti

V. J.

Anda

R. F.

Nordenberg

Williamson

D. F.

Spitz

A. M.

Edwards

Koss

M. P.

Marks

J. S.

(1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The Adverse Childhood Experiences (ACE) Study. American Journal of Preventive Medicine, 14(4), 245–258. https://doi.org/10.1016/S0749-3797(98)00017-8

24.

Ford

J. D.

Chapman

Connor

D. J.

Cruise

K. R.

(2012). Complex trauma and aggression in secure juvenile justice settings. Criminal Justice and Behavior, 39, 694–724. https://doi.org/10.1177/0093854812436957

25.

Ford

J. D.

Courtois

C. A

. (2009). Defining and understanding complex trauma and complex traumatic stress disorders. In Courtois

C. A.

Ford

J. D.

(Eds.), Treating complex traumatic stress disorders: An evidence-based guide (pp. 13–30). Guilford Press.

26.

Forth

A. E.

Kosson

Hare

R. D.

(2003). The Hare Psychopathy Checklist: Youth version. Multi-Health Systems.

27.

Gottfredson

M. R.

Hirschi

(1990). A general theory of crime. Stanford University Press.

28.

Henneberger

A. K.

Oudekerk

B. A.

Reppucci

N. D.

Odgers

C. L.

(2014). Differential subtypes of offending among adolescent girls predict health and criminality in adulthood. Criminal Justice and Behavior, 41, 181–195. https://doi.org/10.1177/0093854813500957

29.

Herman

J. L.

(1992). Trauma and recovery: The aftermath of violence—From domestic to political terror. Basic Books.

30.

Hoge

R. D.

Andrews

D. A.

(2011). Youth level of service/case management inventory: YLS/CMI 2.0 interview guide (2nd ed.). Multi-Health Systems.

31.

Kong

AuCoin

(2008). Female offenders in Canada. Canadian Centre for Justice Statistics, Statistics Canada. https://www150.statcan.gc.ca/n1/pub/85-002-x/85-002-x2008001-eng.pdf

32.

Lanctôt

(2018). Gender-responsive programs and services for girls in residential centers: Meeting different profiles of rehabilitation needs. Criminal Justice and Behavior, 45, 101–120. https://doi.org/10.1177/0093854817733495

33.

Lazarsfeld

P. F.

Henry

N. W.

(1968). Latent structure analysis. Houghton Mifflin.

34.

Loeber

Farrington

D. P.

(2000). Young children who commit crime: Epidemiology, developmental origins, risk factors, early interventions and policy implications. Development and Psychopathology, 12, 737–762. https://doi.org/10.1017/S0954579400004107

35.

Mills

J. F.

Kroner

D. G.

Forth

A. E.

(2002). Measures of criminal attitudes and associations (MCAA): Development, factor structure, reliability, and validity. Assessment, 9(3), 240–253. https://doi.org/10.1177/1073191102009003003

36.

Moretti

M. M.

(2003). The affect regulation checklist [Unpublished research measure]. Simon Fraser University. http://www.adolescenthealth.ca/?page_id=71

37.

Muthén

L. K.

Muthén

B. O.

(1998–2017). Mplus user’s guide (8th ed.).

38.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535–569. https://doi.org/10.1080/10705510701575396

39.

Odgers

C. L.

Moretti

M. M.

Burnette

M. L.

Chauhan

Waite

Reppucci

N. D.

(2007). A latent variable modeling approach to identifying subtypes of serious and violent female juvenile offenders. Aggressive Behavior, 33, 1–14. https://doi.org/10.1002/ab.20190

40.

O’Hagan

H. R.

Brown

S. L.

Jones

N. J.

Skilling

T. A.

(2018). The reliability and validity of the Measure of Criminal Attitudes and Associates and the Pride in Delinquency Scale in a mixed sex sample of justice-involved youth. Criminal Justice and Behavior, 46(5), 751–769. https://doi.org/10.1177/0093854818810459

41.

Orbis Partners. (2000). Youth Assessment Screening Inventory (YASI). https://www.orbispartners.com/juveniles-at-risk-assessment

42.

Puzzanchera

Ehrmann

(2018). Spotlight on girls in the juvenile justice system. National Center for Juvenile Justice, Office of Juvenile Justice and Delinquency Prevention. http://www.ncjj.org/pdf/Data%20snapshots%202016/DataSnapshot_Girls2015.pdf

43.

Reisig

M. D.

Holtfreter

Morash

(2006). Assessing recidivism risk across female pathways to crime. Justice Quarterly, 23, 384–405. https://doi.org/10.1080/07418820600869152

44.

Rosenberg

(1965). Society and the adolescent self-image. Princeton University Press.

45.

Salisbury

Boppre

Kelly

(2016). Gender-responsive risk and need assessment: Implications for the treatment of justice-involved women. In Taxman

(Ed.), Handbook on risk and need assessment: Theory and practice (Vol. 1, pp. 220–243). Taylor & Francis.

46.

Savage

(2019). Female offenders in Canada, 2017. Canadian Centre for Justice Statistics, Statistics Canada. https://www150.statcan.gc.ca/n1/pub/85-002-x/2019001/article/00001-eng.pdf

47.

Scharfe

Bartholomew

(1995). Accommodation and attachment representations in young couples. Journal of Social and Personal Relationships, 12(3), 389–401. https://doi.org/10.1177/0265407595123004

48.

Schwalbe

C. S.

Macy

R. J.

Day

S. H.

Fraser

M. W.

(2008). Classifying offenders: An application of latent class analysis to needs assessment in juvenile justice. Youth Violence and Juvenile Justice, 6(3), 279–294. https://doi.org/10.1177/1541204007313383

49.

Substance Abuse and Mental Health Services Administration. (2014). National Mental Health Services Survey (N-MHSS): 2014. Data on mental health treatment facilities. Department of Health and Human Services, Substance Abuse and Mental Health Services Administration. https://www.samhsa.gov/data/sites/default/files/2014_National_Mental_Health_Services_Survey.pdf

50.

Tein

J. Y.

Coxe

Cham

(2013). Statistical power to detect the correct number of classes in latent profile analysis. Structural Equation Modeling: A Multidisciplinary Journal, 20, 640–657. https://doi.org/10.1080/10705511.2013.824781

51.

van der Kolk

B. A

. (2005). Developmental trauma disorder: Toward a rational diagnosis for children with complex trauma histories. Psychiatric Annals, 35(50), 401–408.

52.

Whitney

S. D.

Renner

L. M.

Herrenkohl

(2010). Gender differences in risk and promotive classifications associated with adolescent delinquency. Journal of Genetic Psychology, 171(2), 116–138. https://doi.org/10.1080/00221320903548092

53.

World Health Organization. (2018). ICD-11 for mortality and morbidity statistics. https://icd.who.int/browse11/l-m/en