Abstract
Child maltreatment research is increasingly recognizing the need to capture patterns of co-occurrence between different types of abuse/neglect and to consider their associations with psychosocial functioning. Few studies have examined these issues in justice-involved youth despite the fact that rates of maltreatment and trauma-related psychopathology are disproportionately high among this population. This study examined profiles of self-reported child physical abuse, sexual abuse, emotional abuse, and neglect among incarcerated juveniles in Victoria, Australia, using latent class analysis. We also investigated associations between maltreatment profiles and mental health and behavioral problems. Data pertaining to juveniles’ experiences of maltreatment and mental health and behavioral functioning were collected from interviews, questionnaires, files, and administrative datasets. A three-class solution provided the best fit for the data and was conceptually meaningful: a “low/rare maltreatment” class (41%); “high physical and emotional abuse” class (23%); and a “poly-victimization” class (36%). Youth in the “poly-victimization” class experienced especially serious mental health and behavioral disturbances, including higher rates of mental illness, greater severity of internalizing and externalizing symptoms, impulsivity, substance abuse, self-harm and suicidal behavior, irritability, and early-onset violence. Results suggest there may be benefit in considering screening and assessment procedures in youth justice settings to identify poly-victimized youth in need of more intensive monitoring and treatment to address their complex clinical and behavioral profiles.
Child maltreatment is a complex social, human rights, and public health problem that has significant and enduring ramifications for victims, their families, and broader society. In the USA, more than 3.5 million children were referred to child protection services for maltreatment concerns in 2018 alone; 678,000 were found to be victims of maltreatment (U.S. Department of Health & Human Services, 2020). Concerning rates of child protection notifications have also been reported in Australia, the location of the current study (Australian Institute of Health and Welfare [AIHW], 2020a). Moreover, child protection records substantially underestimate the extent of child maltreatment in the community, as not all maltreatment cases come to the attention of authorities. Recent reviews of the international prevalence of self-reported maltreated suggest it affects around 127 per 1,000 for sexual abuse, 226 per 1,000 for physical abuse, 363 per 1,000 for emotional abuse, 163 per 1,000 for physical neglect, and 184 per 1,000 for emotional neglect (Stoltenborgh et al., 2015). Child maltreatment has been associated with a range of negative outcomes, such as mental health problems (Green et al., 2020), substance abuse (Widom et al., 2006), risk-taking behaviors (Kerig, 2019), revictimization (Widom et al., 2008), and criminal offending (Malvaso et al., 2017; Papalia et al., 2018).
The association between childhood maltreatment and offending behavior is complex and not deterministic. However elevated rates of child maltreatment have been found among justice-involved young people (Baidawi & Sheehan, 2019; Malvaso et al., 2018). Contemporary data from the study jurisdiction, for instance, demonstrate that one quarter of children before the criminal courts were the subject of substantiated maltreatment, and over one-fifth (22%) had been subject to a child protection court order (Baidawi & Sheehan, 2019; Sentencing Advisory Council, 2019). The offending profiles of maltreated children and young people also demonstrate a heightened risk of early-onset offending (Baglivio et al., 2015; Baidawi & Sheehan, 2019), recidivism (Baglivio et al., 2016), and violence throughout the life course (Fitton et al., 2018; Widom, 2017). These profiles illustrate that maltreated children form not only a disproportionate number of justice-involved youth but a subset of this population also face greater risk of ongoing serious offending, including progression to the adult justice system. Indeed, youth in custodial settings not only evidence a high prevalence of childhood maltreatment (Malvaso et al., 2018), but many have child protection system involvement (Sentencing Advisory Council, 2019) and have been placed in the care of child welfare authorities (Justice Health & Forensic Mental Health Network, 2017). Such children who traverse child protection and youth justice services are alternately termed “crossover,” “dual status,” or “dually-involved,” in recognition of their involvement with these distinct statutory systems.
Although there are high rates of maltreatment among justice-involved youth, prospective cohort studies have shown that only a small proportion of children who are maltreated (according to child protection records) have later youth criminal convictions (between 1% and 10%; Malvaso et al., 2017; Vidal et al., 2017). As such, an emerging body of research has sought to identify factors associated with an increase in maltreated children’s likelihood of subsequent criminal justice involvement and to undertake descriptive analyses of the maltreatment, mental health, and behavioral profiles of justice-involved youth who have endured child maltreatment.
Trauma-related psychopathology has been identified as one potential pathway to criminal justice system involvement among maltreated youth (Kerig, 2019). A large body of work, including prospective longitudinal research shows that maltreatment is associated with increased risk of psychiatric diagnoses in adolescence and adulthood including post-traumatic stress disorder (Widom, 1999), mood, anxiety, and personality disorders (Johnson et al., 1999; Thornberry et al., 2001; Widom et al., 2007), substance-use disorders (Widom et al., 2006), and psychotic symptoms (Arseneault et al., 2011). The increased risk of psychopathology following child maltreatment extends to a range of emotional and behavioral outcomes including emotion recognition and regulatory difficulties (Dvir et al., 2014), insecure/disorganized attachment (Cyr et al., 2010), impulsivity and risk-taking (Kerig, 2019), self-harm and suicidal behavior (Fergusson et al., 2008), as well as aggression and violence (Fitton et al., 2018; Lee & Hoaken, 2007). Developmental theories propose a range of potential mechanisms by which trauma-related psychopathology increases risk of criminal justice system contact in youth. Some of these include impaired risk recognition, maladaptive affect regulation, attempts to regain personal integrity and a sense of control, and recklessness or thrill-seeking behavior related to impaired fear circuitry (for a theoretical review, see Kerig, 2019). Providing some support for this perspective, recent research has found higher rates of mental health and behavioral disturbances among justice-involved youth who report having experienced maltreatment, relative to justice-involved youth without maltreatment histories (Meurk et al., 2019). Longitudinal and case-control research also identifies emotional and behavioral problems as predictors of offending (Barrett et al., 2014; Jonson-Reid, 2002), including earlier-onset of offending (Cho et al., 2019), among maltreated youth. While trauma-related psychopathology may relate to criminal justice system contact, whether a young person offends, particularly repeatedly, is influenced by a range of personal, interpersonal, and social factors, and particularly the presence of internal criminogenic factors (e.g., antisocial attitudes, antisocial personality traits) that are reinforced over time (Andrews & Bonta, 2017). This likely explains why the vast majority of maltreated children, including those who develop mental health and emotional challenges, do not offend.
Studies attempting to ascertain the association between specific maltreatment types and offending risk report varied findings. Aside from definitional and measurement variations, victims of child abuse and neglect are typically exposed to multiple forms of maltreatment, generating additional challenges to discerning the impact of maltreatment type. Additionally, there is some overlap in these concepts, observable for instance, by the fact that childhood physical and sexual abuse are often inherently emotionally abusive experiences (Baidawi, 2019). Nonetheless, while any maltreatment appears to relate to an increased risk of criminal justice system involvement, some evidence suggests that experiences of physical abuse and neglect are most strongly related to a range of offense types (Maxfield & Widom, 1996; Vidal et al., 2017; Widom & Massey, 2015). Further, the risk of offending among maltreated youth appears to be increased by experiences of cumulative harm, for instance maltreatment recurrence and persistence into adolescence (Cho et al., 2019; Malvaso et al., 2017).
Related to the notion of cumulative harm, child maltreatment researchers are increasingly turning their attention to the adverse outcomes experienced by multi-type or poly-victims—that is, those who have been exposed to not just one but multiple types of child maltreatment and traumatic victimization (Finkelhor et al., 2007). Person-centered techniques, such as latent class analyses, provide a reliable alternative to traditional variable-centered analyses that enable identification of distinct subgroups of youth and adults based on their patterns of exposure to different types of child abuse and neglect (Debowska et al., 2017). Such approaches avoid the “artificial compartmentalization” of examining siloed maltreatment exposures that fails to reflect real-world experiences (Widom, 2017, p. 195). Poly-victimization is overrepresented in youth justice populations and is associated with serious mental health and behavioral problems (Debowska et al., 2017; Ford et al., 2018, 2010). In addition to potentially requiring clinical interventions, these problems may affect or impede an individual’s response to criminogenic treatments. This may partly explain why poly-victimized youth are at increased risk of reoffending and deepening involvement in the justice system (Fox et al., 2015).
Very few studies have applied person-centered approaches to examine patterns of maltreatment co-occurrence among justice-involved populations. Those that do are difficult to compare due to differences in study samples (e.g., community vs. custodial; single vs. two-gender studies; youth vs. adult systems) and measures of traumatic exposure (e.g., self-report vs. administrative data; maltreatment only vs. a range of potential traumas and adversities). Nevertheless, these studies consistently identify several latent classes of maltreatment/trauma among justice-involved youth (Aebi et al., 2015; Charak et al., 2019; Ford et al., 2013) and adults (Debowska & Boduszek, 2017; Zhang & Zheng, 2018), commonly between three and five classes, distinct in the level and type of victimization experienced. Invariably, a class characterized by poly-victimization is identified, showing increased emotional and behavioral disturbances. Studies with incarcerated youth have generally encompassed a wider range of traumas and adversities, beyond child maltreatment (e.g., witnessing an accident, seeing someone assaulted, disaster, being in a war zone; Charak et al., 2019; Ford et al., 2013). Although young people can encounter violence and victimization in a range of settings, victimization in the home may have the strongest impact on a child’s development (Margolin & Gordis, 2000; Turner et al., 2006). To our knowledge, only one study has specifically examined profiles of child maltreatment in a custodial youth population (Aebi et al., 2015). This study, conducted in Vienna, identified a small group (8%) of incarcerated male adolescents characterized by high levels of multi-type maltreatment (i.e., physical abuse, sexual abuse, emotional abuse). This group evidenced higher rates of psychiatric illness and subsequent reincarceration than incarcerated male adolescents with no/mild maltreatment.
Aims of the Current Study
An emerging body of research points to the importance of probing the associations between child maltreatment, psychopathology, and justice system involvement. While emotional and behavioral difficulties appear to be associated with offending behavior among maltreated children, little research has examined these relationships, particularly in detention settings, where children and youth experience a higher prevalence of these concerns. Such knowledge has attendant impacts on the identification of needs and responsivity to treatment and rehabilitation measures. To our knowledge, this is the first study to investigate profiles of five forms of child maltreatment—physical abuse, sexual abuse, emotional abuse, physical neglect, and emotional neglect—in detained youth in Australia using latent class analysis. We sought to examine associations between the maltreatment latent profiles and mental health and behavioral variables across four broad domains: psychopathology (e.g., mental illness, internalizing/externalizing symptoms), impulsivity/inhibitory control, maladaptive coping (e.g., self-harm/suicidal behavior, substance abuse), and aggression (e.g., early-onset violence). In addressing these aims, we add to a small evidence base and extend the Aebi et al. (2015) study in two ways. First, we include two additional types of maltreatment—physical and emotional neglect—with research suggesting that neglect is an important predictor of negative developmental outcomes (see Widom, 2017). Second, we include incarcerated female youth. Although females make up a relatively small proportion of our sample (18%), this is highly representative of the youth justice population in Australia (AIHW, 2020b). Based on existing research, we hypothesized that several distinct subgroups of youths with varied maltreatment experiences would emerge. We expected one or more of these groups to be characterized by high endorsement rates of multiple types of maltreatment, or poly-victimization. Second, consistent with notions of cumulative harm, we hypothesized that youth in the poly-victimization subgroup(s) would evidence greater mental health and behavioral problems, including higher rates of psychiatric disorder and psychopathology, impulsivity, self-harm and suicidality, substance abuse, and aggression.
Method
Sample Setting
We use existing data (see Shepherd et al., 2018) of 215 youth from two justice centers in Victoria Australia: Parkville Youth Justice Precinct (PYJP) and Malmsbury Youth Justice Centre (MYJC). Victoria is the second most populous Australian state, with a population of more than 6.5 million, concentrated mainly in the capital city. PYJP houses remanded/sentenced males and females aged 10–17 years and sentenced females aged 18–20 years. MYJC houses sentenced males aged 18–20 years. Victoria’s unique “dual track” system under the Sentencing Act 1991 enables adult courts to sentence 18–20-year-olds to serve custodial sentences in youth detention rather than adult prison; hence, their inclusion in this sample.
Measures
Child Maltreatment
The Childhood Trauma Questionnaire (CTQ; Bernstein et al., 2003) is a self-report inventory used to assess experiences of child maltreatment across five domains: physical abuse (PA), emotional abuse (EA), sexual abuse (SA), physical neglect (PN), and emotional neglect (EN). Each domain includes five questions, with respondents asked to provide answers about their “experiences growing up as a child and a teenager”, anchored on a 5-point scale (1 = never true to 5 = very often true). Scores from each domain can then be summed (range: 5–25) to indicate the level/frequency of maltreatment experienced. The CTQ has been widely validated and is frequently used with justice-involved youth (Aebi et al., 2015; Kenny et al., 2006). The manual classifies maltreatment subscale scores into severity quartiles: “none/minimal,” “low to moderate,” “moderate to severe,” and “severe to extreme.” For each domain, we coded scores in the “none/minimal” range as indicating “no maltreatment,” (i.e., PA ≤ 7; EA ≤ 8; SA ≤ 5; PN ≤ 7; EN ≤ 9) and scores in the remaining three quartiles as indicating “maltreatment.” Other studies have used a similar approach to delineate the occurrence of maltreatment using the CTQ (Aebi et al., 2015; Kenny et al., 2006; Zhang & Zheng, 2018).
Mental Health Service Contacts and Diagnoses
We extracted consenting participants’ public mental health records from the statewide Redevelopment of Acute & Psychiatric Information Directions (RAPID) database. This included all contacts with, and admissions to, an outpatient or inpatient mental health service in Victoria. Admissions to inpatient beds in the private sector or contacts with private outpatient services are not captured by RAPID. We collapsed diagnoses into major groups, in line with common psychiatric diagnostic systems, including: psychotic disorders (e.g., schizophrenia, schizoaffective, drug-induced psychosis); mood disorders (e.g., major depressive, bipolar affective, dysthymia); anxiety disorders (e.g., generalized anxiety, obsessive–compulsive); trauma/stress-related disorders (e.g., post-traumatic stress, adjustment disorder, acute stress, attachment disorder); and substance use disorders.
Traumatic Brain Injury (TBI)
Lifetime head injuries and associated symptoms were screened using the HELPS Brain Injury Screening Tool (Picard et al., 1999). There are five general questions about TBI events and any symptoms/experiences associated with those events. These include: H—Have you ever Hit your Head or been Hit on the Head? E—Were you ever seen in the Emergency room, hospital or by a doctor because of an injury to your head? L—Did you ever Lose consciousness or experience a period of being dazed and confused because of an injury to your head? P—Do you experience any of these Problems (list included) in your daily life since you hit your head? and S—Have you had any significant Sicknesses? A participant is screened in for possible TBI when they endorse questions regarding: (a) an event that could have resulted in a brain injury (i.e., questions H, E, or S); (b) medical assessment for a head injury or a loss of consciousness associated with the event (i.e., questions E or L); and (c) at least two enduring problems arising from the traumatic injury (i.e., question P).
Self-Reported Psychopathology
The Youth Self Report (YSR; Achenbach & Rescorla, 2001) questionnaire assesses emotional and behavioral problems in youth aged 11–18 years. It comprises 112 items rated on a 3-point scale (0 = not true, 1 = somewhat or sometimes true, 2 = very true or often true), which generate eight scale scores encompassing a broad range of symptoms present in the past six months: social withdrawn/depressed; somatic complaints; anxious/depressed; social problems; thought problems; attention problems; delinquent behavior; and aggressive behavior. Higher scores indicate higher levels of psychopathology. Adequate reliability estimates have been reported for the subscales (Achenbach & Rescorla, 2001; Ebesutani et al., 2011).
Impulsivity/Inhibitory Control
We used two tools to assess impulsivity and inhibitory control: the Barratt Impulsivity Scale (BIS) and the go/no-go task. The BIS is a 30-item tool encompassing attentional, motor and nonplanning impulsiveness subfactors (Stanford et al., 2009). Answers are rated on a 4-point scale (1 = never/rarely to 4 = almost/always). Total scores of 72 and above suggest high levels of impulsivity. The BIS is well validated, and is frequently employed with justice-involved populations (Stanford et al., 2009). Inhibitory control was assessed using a go/no-go task adapted by Rubia et al. (1998). The task is administered as a computer-based program and requires participants to either initiate (Go) or inhibit (No-Go) a motor response (pressing the “space” bar), depending on whether a picture of an airplane (Go) or bomb (No-Go) appears on the computer screen. The stimuli, 70% of which are airplanes and 30% are bombs, are delivered as one block of 180 trials and appear in a random order for a duration of 200 ms with an interstimulus interval of 1,600 ms. Participants were instructed to respond as quickly as possible to the Go stimulus and to inhibit responding when the No-Go stimulus appeared. The mean response time to the Go stimulus and number of errors of omission and commission were recorded.
Self-Harm or Suicide Attempts
We collated information on self-harm and suicidal behaviors from two sources: the Victorian Offending Needs Indicator for Youth (VONIY) and the Structured Assessment of Violence Risk in Youth (SAVRY) item 5 (Shepherd et al., 2018). The VONIY was a criminogenic needs tool that assisted Victorian youth justice case workers with intervention prioritization. When the study was conducted, the VONIY was completed by youth justice staff following an interview with the young person and review of collateral information. The VONIY item, ‘self-harm concerns’, reflecting any history of self-harm/suicidal behaviors (i.e., suicidal ideation, acts of self-mutilation, suicide attempts), was included in this study. The SAVRY is a structured professional judgment tool for assessing risk for violence in youth aged 12–18 years (Borum et al., 2006). We used item 5, “history of self-harm or suicide attempts,” which is coded as; Low: no history of self-harm or suicide attempts; Moderate: a history of self-harm or suicidal actions that did not require medical care and had no clear suicidal intent; High: a history of medically severe self-harm (requiring medical care or hospitalization) or one or more suicide attempts (Borum et al., 2006). The SAVRY was scored after a semi-structured interview with the participant conducted by Masters-level clinical researchers. For this analysis, and consistent with Shepherd et al. (2018), participants were coded as having a history of self-harm if they received a “Moderate” or “High” rating on the SAVRY and/or the presence of self-harm concerns on the VONIY. Participants were coded as having a history of suicide attempts if they received a “High” rating on the SAVRY only.
Illicit Polysubstance Use
The Kiddie-Schedule for Affective Disorders and Schizophrenia—Present and Lifetime (K-SADS-PL) is a semi-structured diagnostic interview designed to assess current and past episodes of psychopathology in children and adolescents according to DSM-IV criteria (Kaufman et al., 1997). While the instrument can generate 35 child and adolescent psychiatric diagnoses, only the Substance-Related Disorders module screening and diagnostic interview questions were administered. The total number of different illicit substances used in the past six months were coded based on the K-SADS-PL interview. The K-SADS-PL interviewers were two doctoral-level clinical psychologists and two doctoral students under clinical supervision.
Early-Onset Violence
Early-onset violence was assessed using the SAVRY (Borum et al., 2006) Item 3, “Early Initiation of Violence.” This item is coded as; Low: no known violent acts before age 14; Moderate: first known violent act between ages 11 and 13; High: first known violent act prior to age 11 (Borum et al., 2006). For the purposes of this study, participants were coded as having early-onset violence if they received a “High” rating on the SAVRY Item 3.
Aggressiveness
We utilized two tools to assess aggressiveness: the Impulsive Aggression Quick Screen (IA–QS) and the Impulsive/Premeditated Aggression Scale (IPAS). The IA–QS (Stanford et al., 1995) is a brief semi-structured interview that screens for impulsive aggression. It uses DSM-IV-TR criteria for Intermittent Explosive Disorder along with the Irritability subscale from the Buss-Durkee Hostility Inventory. The IA–QS focuses on an individual’s aggressive behavior in the past six months. The IPAS (Stanford et al., 2003) is a 30-item self-report tool measuring subtypes of aggressive behavior. Participants consider their aggressive behavior in the preceding six months and rate their responses on a 5-point scale (1 = strongly disagree, 5 = strongly agree). The tool generates two subscales; one indicating impulsive/spontaneous aggression and the other indicating premeditated/instrumental aggression. The IPAS has been validated in adolescent and adult samples (Mathias et al., 2007; Stanford et al., 2003).
Procedure
The study procedures have been reported elsewhere (Shepherd et al., 2018). Briefly, researchers approached youth in custody between July 2011 and June 2012 and invited them to participate in the study. Those interested and willing to participate provided written informed consent. Participants under the age of 18 can provide informed consent in line with the “mature minor” principal described in Victorian legislation, provided they have sufficient capacity to comprehend the nature of the study and what participation involves. Participants were interviewed separately in private rooms, during which assessments and questionnaires were completed. Each interview lasted approximately 90 minutes. VONIY data were obtained from the then Victorian Department of Health and Human Services. The consent procedures were approved by the Victorian Department of Health and Human Services and Monash University Human Research Ethics Committees.
Analyses
Several variables had a small amount of missing data: early-onset violence (n = 2; 0.9%); suicide attempt (n = 2; 0.9%); TBI (n = 7; 3.3%); and illicit polysubstance use (n = 1; 0.5%). Cases with missing data were excluded from analyses involving that particular variable. To examine patterns of maltreatment co-occurrence, we conducted latent class analysis (LCA), a statistical technique used to determine the number of homogeneous groups (or classes) from categorical data. LCA assumes that relationships between observed categorical variables—in this case, the five maltreatment subtypes—can be explained by a finite number of mutually exclusive latent groups (Debowska & Boduszek, 2017). Multiple models with one through six latent classes were fitted to the data using Mplus Version 8 (Muthén & Muthén, 2015). We used several fit indices to assist with model selection: Akaike’s information criterion (AIC); Bayesian information criterion (BIC); sample-size-adjusted BIC (aBIC); likelihood ratio (LR) χ2 test; Lo–Mendell–Rubin adjusted likelihood ratio test (LMR LRT); and parametric bootstrapped likelihood ratio test (BLRT). Smaller AIC, BIC, and aBIC values suggest the most parsimonious and preferred model. The LR χ2 test measures absolute model fit, where significant values reflect a discrepancy between the model and the data. The LMR LRT and BLRT compare the current model (k classes) with the preceding model (k – 1 classes). A non-significant value indicates that the model with one fewer class offers a more parsimonious fit. Entropy, while not recommended as an indicator for model selection, quantifies the performance of the model in classifying individuals into classes; values closer to one suggest better classification and separation between classes. We also considered whether the final model reflected coherent, distinct, and theoretically meaningful subgroups (Debowska & Boduszek, 2017).
After selecting the optimal model, participants were assigned to their most likely class according to the posterior class membership probabilities. Next, Chi-squared tests of association and ANOVAs were used to examine differences in latent classes on demographic, psychopathology, impulsivity/inhibitory control, maladaptive coping, and aggression variables at the omnibus level, with α = .05 corrected for multiple comparisons using the Benjamini–Hochberg method. Due to unequal class sizes, we used Welch’s test for omnibus ANOVA and Games–Howell post hoc comparisons where the assumption of homogeneity of variance was violated. Post hoc comparisons were only conducted where the omnibus analysis yielded a statistically significant p-value following correction. Benjamini–Hochberg corrections for multiple comparisons were then applied at the post hoc level, except for ANOVA post hoc analyses, which are already adjusted for multiple testing. Where significant differences across classes were observed, we calculated odds ratios (ORs) with confidence intervals (CIs) and Cohen’s d for categorical and continuous correlates, respectively.
Results
Descriptive Statistics
The average age of participants (n = 215; 82% male) was 16.82 (SD = 1.83, range: 12–21) years. Participants self-identified diverse cultural backgrounds, including white and/or Caucasian participants of European descent (categorized as English-speaking backgrounds; ESB, 48.8%, n = 105), a range of non-English-speaking-background minority groups (e.g., Vietnamese, Sudanese, Pacific Islander; categorized as Culturally and Linguistically Diverse; CALD, 31.2%, n = 67), and Aboriginal Australian or Torres Strait Islander heritages (categorized as Indigenous, 20.0%, n = 43). Most participants (86.5%, n = 186) had prior police charges for violent crimes and all participants self-reported a history of violence. The most common index offenses were assault (36.7%), burglary/theft (17.7%), and robbery (15.2%). The average sentence length that participants were serving was 8.78 months (SD = 16.02), and the mean total number of prior community-based or custodial orders was 3.74 (SD = 3.97). Using dichotomized CTQ subscale scores, 77.2% (n = 166) of the sample endorsed at least one form of maltreatment; 51.2% (n = 110) reported experiencing EA, 58.6% (n = 126) reported PA, 16.7% (n = 36) reported SA, 37.7% (n = 81) reported EN, and 42.3% (n = 91) reported PN. Additional descriptive data are presented in Table 1.
Descriptive Statistics for the Total Young Offender Sample (N = 215).
Note. M = mean; SD = standard deviation; YSR = Youth Self-Report; MRT = mean reaction time; IAQS = Impulsive Aggression Quick Screen; IA = impulsive aggression; IPAS = Impulsive/Premediated Aggression Scale.
Patterns of Maltreatment Co-Occurrence
LCA resulted in the selection of a three-class model based on the fit indices (Table 2) and substantive meaning of classes. The AIC and aBIC were smaller for the three-class model relative to the two-class model; however, the BIC favored a two-class solution. With samples under 500 and unequal class sizes, the aBIC performs better than the BIC (Dziak et al., 2014; Nylund et al., 2007). The non-significant LRT and BLRT for the four-class model confirmed that it did not provide a better fit over the three-class solution. The average posterior probabilities (AvPPs) of assignment to the most likely class for the chosen three-class model suggested that, on average, participants were classified with a high degree of certainty, with AvPPs of 0.902, 0.932, and 0.917, for classes 1 to 3, respectively. The three-class model is depicted in Figure 1. Latent class 1 (40.47% of the sample; 42.9% of males and 28.9% of females) consisted of youth reporting relatively low endorsement of all maltreatment types. Individuals in this class most frequently endorsed EN (0.19), followed by PN (0.185), EA (0.156), PA (0.122), and SA (0.00). Subsequently, this class was labeled the “low/rare maltreatment” group. Youth in Class 2 (23.26% of the sample; 21.5% of males and 31.6% of females) showed a high item-response probability of reporting PA (0.912), a moderate probability of reporting EA (0.54), and a relatively low probability of reporting SA (0.188), EN (0.00) and PN (0.00). Thus, this class was considered to reflect a “high physical and emotional abuse” class. Class 3 (36.27% of the sample; 35.6% of males and 39.5% of females) consisted of youth who highly endorsed PA (0.902), PN (0.894), EA (0.87), and EN (0.769). The probability of endorsing SA was moderate in this class (0.332), albeit considerably higher than the endorsement rate in the complete sample (16.7%). Therefore, this class was labeled a “poly-victimization” group.
Comparison of Latent Class Models for the Sample of Young Offenders (N = 215) based on Dichotomous CTQ Subscales.
Note. CTQ = Childhood Trauma Questionnaire; AIC = Akaike’s information criterion; BIC = Bayesian information criterion; aBIC = sample-size-adjusted BIC; LR χ2 = likelihood ratio Chi-square test; LMR LRT = Lo–Mendell–Rubin adjusted likelihood ratio test; BLRT = parametric bootstrapped likelihood ratio test.
Boldface type suggests the best-fitting model for that particular indicator.
Note. Conditional response probabilities reflect the probability of a particular observed response (in this case, positive endorsement) on a particular variable (in this case, the maltreatment subtypes) conditional on latent class membership.
Mental Health and Behavioral Characteristics
Associations between latent classes and demographic, mental health and behavioral variables are shown in Table 3. There were no significant differences between the classes on demographic variables. For psychopathology variables, youth in the poly-victimization class had significantly higher rates of diagnosed psychotic disorders than those in the low/rare maltreatment class. Differences between the classes on rates of major mood disorders, anxiety disorders, trauma-related disorders, and possible TBI approached but failed to reach statistical significance. Average scores on the YSR differed significantly across classes for each syndrome scale, with the majority of these differences reflecting a greater degree of emotional and behavioral symptoms among poly-victimized youth relative to one or both of the other two classes. Youth in the “high physical and emotional abuse” group scored higher than “low/rare maltreatment” youth on social problems, rule-breaking behavior, and aggressive behavior as per the YSR. Across impulsivity and coping domains, poly-victimized youth demonstrated significantly higher impulsivity, a greater number of omission errors on the go/no-go task, higher rates of self-harm concerns (including suicide attempts), greater number of illicit substances used, and higher rates of diagnosed substance use disorders. In relation to aggression variables, the odds of showing an early-onset of violence were more than two times higher among poly-victimized youth relative to low/rare maltreatment youth. Poly-victimized youth also evidenced a higher mean irritability score on the IAQS.
Associations Between Demographic, Mental Health and Behavioral Variables, and the Three Maltreatment Classes (N = 215).
Note. M = mean; SD = standard deviation; OR = odds ratio; CI = confidence interval; d = Cohen’s d; ESB = English-speaking background; CALD = culturally and linguistically diverse; YSR = Youth Self-Report; MRT = mean response time; IAQS = Impulsive Aggression Quick Screen; IA = impulsive aggression; IPAS = Impulsive/Pre-meditated Aggression Scale; NA = not available as an OR cannot be calculated when everyone or no one in one group develops the outcome of interest.
ap-values typed in boldface indicate statistical significance following correction for multiple comparisons. bPairwise comparisons typed in boldface indicate statistical significance following correction for multiple comparisons. cFisher’s exact test.
Discussion
This study examined profiles of child maltreatment in detained youth and their relation to mental health and behavioral problems. Supporting our first hypothesis, we identified three latent classes of child maltreatment, including a “poly-victimization” group, a “high physical and emotional abuse” group, and a “low/rare maltreatment” group. Consistent with hypothesis two, youth in the “poly-victimization” group evidenced greater problems than youth in one or both of the other two classes across psychopathology, impulsivity/inhibitory control, maladaptive coping, and aggression domains. The main points are: (a) more than three quarters (77%) of detained juveniles report exposure to child maltreatment of any kind; (b) a subgroup of poly-victimized justice-involved youth are especially likely to have endured all five forms of maltreatment, and to need intervention to address the severe mental health and behavioral problems and risks they experience (Charak et al., 2019); and (c) another subgroup, although less likely than poly-victimized youth to have suffered pervasive maltreatment, have still experienced significant physical and emotional abuse, and may benefit from increased support to address the difficulties they experience in social relationships, general delinquency, and aggression.
The identified classes parallel findings from prior research. Studies with justice-involved youth often find a group characterized by low/minimal/mild maltreatment and adversity (Aebi et al., 2015; Ford et al., 2013). Our findings also revealed a class of young people exposed to high rates of physical and emotional abuse (23%). Previous research has found similar subgroups of justice-involved youth (Aebi et al., 2015; Charak et al., 2019) and imprisoned adults (Debowska & Boduszek, 2017; Zhang & Zheng, 2018) showing high levels of exposure to violent environments, which are often inherently emotionally abusive environments (Baidawi, 2019).
Similarly, the identification of “poly-victimization” group in this study, characterized by high endorsement of the five maltreatment types, replicates what has been found in other studies with justice-involved samples. However, the membership rate for this class (36%) was higher than previous studies (Aebi et al., 2015; Charak et al., 2019; Ford et al., 2013). Studies identifying smaller groups of poly-victims often measure a broader range of adverse and traumatic experiences (up to 26 types), which could account for these disparities. Aebi et al. (2015) is an exception, where the authors focused on just three types of child maltreatment (physical, sexual, and emotional), and only 8% of the sample endorsed high rates of each abuse type. Cultural (i.e., Austria vs. Australia) and gender (i.e., 0% female vs. 18% female) differences may partly explain the discrepancy in size of the poly-victimization class found for Aebi et al. and this study, with some research suggesting that girls are more likely to be classed as poly-victims than boys (Charak et al., 2019; Ford et al., 2013). Interestingly, we did not find significant differences across classes according to gender and ethnicity, which is novel and requires further exploration in deep-end justice-involved youth. Despite these inconsistencies, the fact that more than one third of detained youth endorsed relatively high rates of all five forms of maltreatment is concerning and highlights the substantial and cumulative trauma exposure among this population (Abram et al., 2004).
Young people in the “high physical and emotional abuse” subgroup reported comparably high rates of physical abuse exposure (about 90%) to the poly-victimization group; but, unlike the latter, they reported lower rates of the remaining four maltreatment types. This accords with research findings demonstrating high levels of violence exposure among justice-involved youth (Abram et al., 2004). “High physical and emotional abuse” youth evidenced more serious problems with rule-breaking, aggressive behavior, and impaired peer relationships as per the YSR than “low/rare maltreatment” youth. This supports prior work showing significant externalizing behavior among young people who experience family violence (Mrug & Windle, 2010). Violent family environments may contribute to the development of antisocial attitudes (Bandura, 1973) and foster a learned emotional detachment as a self-protective mechanism, resulting in a callous appearance (Bennett & Kerig, 2014). Like Charak et al. (2019) suggest, justice-involved youth in the “high physical and emotional abuse” group may therefore be incorrectly identified as high in primary callous-unemotional traits, when in reality their presentations may be better described as “trauma-related acquired callousness” (Bennett & Kerig, 2014, p. 415), which may be responsive to trauma-focused treatments (Charak et al., 2019; Ford et al., 2016).
Consistent with prior research and models of cumulative harm, we anticipated that young people in the “poly-victimization” class would show the most significant disturbances in mental health and behavioral functioning. This was confirmed through higher rates of psychotic illness, greater severity of internalizing (e.g., depression, anxiety, somatic complaints) and externalizing (e.g., rule-breaking, aggressive behavior) symptoms, impulsivity, substance abuse, self-harm and suicide attempts, irritability, and early-onset violence. Although not significant following correction, rates of trauma-related disorders and possible TBI were also high (around one third) among poly-victimized youth, consistent with existing research (Perron & Howard, 2008). Collectively, these findings point to a pattern of severe and pervasive emotional, cognitive, and behavioral dysregulation among justice-involved youth exposed to poly-victimization, consistent with prior work (Aebi et al., 2015; Charak et al., 2019; Ford et al., 2013). Some of the challenges faced by poly-victimized youth are known risk factors for recidivism—for example, aggressive behavior, early-onset violence, substance use, and impulsivity (Andrews & Bonta, 2017). Others may be construed as non-criminogenic needs or factors that, if unaddressed, may impair the young person’s capacity to benefit from criminogenic interventions aimed at reducing recidivism (e.g., mental illness, psychological distress). With the knowledge that base rates of maltreatment and trauma are extremely high among justice-involved youth, identification of youth with poly-victimization histories may provide a more feasible, efficient, and precise screening focus than simply identifying young people based on the presence of childhood trauma (Ford et al., 2018).
The primary difference between the poly-victimized class and the “high physical and emotional abuse” class was the former’s exposure to additional traumas, including sexual abuse and physical and emotional neglect. Relative to “high physical and emotional abuse” youth, poly-victimized youth also evidenced greater internalizing symptoms (i.e., anxiety, depression, social withdrawal, thought problems), trauma-related disorders (non-significant post-correction), impulsivity, self-harm and suicide attempts, and substance use disorders. Therefore, the additional trauma and attachment disruption that likely occurs as a result of the violations inherent to sexual abuse and the lack of adequate, stable, and nurturing caregivers who provide safety and positive modeling may contribute to the more pervasive emotional and behavioral regulatory deficits seen in poly-victimized youth (Charak et al., 2019). This is consistent with existing research highlighting the role of neglect, abandonment, maternal rejection, and sexual abuse in predicting risky, impulsive, and self-destructive behaviors, such as intentional self-injury and substance abuse (see Kerig, 2019).
There are several limitations associated with this research. First, incarcerated youth represent an extremely unique and high-risk sample, making up just 0.02% of the Victorian youth population. Thus, the findings are not generalizable to the wider population of youth who come into contact with police or who are found guilty of a criminal offense. Second, the accuracy of participants’ reports of sensitive topics such as childhood trauma, mental health, suicidal behavior, substance use, and aggression may have been influenced by factors such as reliability of memory, psychopathology, and/or fear of stigma or legal implications (Charak et al., 2019; Kenny et al., 2006; Papalia et al., 2018). Second, the data are cross-sectional and descriptive, which precludes determining any causal, directional, or temporal relationships (Shepherd et al., 2018). The analysis could not determine whether maltreatment occurred prior to, concurrent with, or following justice-system involvement or the emergence of mental health and behavioral problems. Third, due to the small proportion of female participants, we were unable to explore possible gender-specific maltreatment classes and associations. However, this is representative of the gender distribution in Australian incarcerated juvenile populations. Finally, the prevalence of psychiatric disorder in the sample will be underestimated due the data source used. Young people with severe mental illness (e.g., psychotic illnesses) are highly likely to have contact with public psychiatric services in Victoria and thus appear on the database used. However, higher prevalence disorders, without any primary psychotic illness, can often be managed in the private sector or by a general practitioner, and may be underestimated. Relatedly, we could not reliably examine whether emerging personality pathology was associated with identified classes, given the reluctance to diagnose young people under the age of 18 with a personality disorder.
The heightened rates of serious mental health disorders, internalizing and externalizing symptoms, impulse-control deficits, (poly)substance abuse, suicide risk, and aggression problems seen among poly-victimized youth pose significant policy and practice challenges for youth justice, child protection, and mental health sectors (Charak et al., 2019). These youth are likely to attract a wide range of mental health diagnostic labels (D’Andrea et al., 2012); however, as Ford et al. (2013) argue, their pervasive self-regulatory deficits and complex “clinical” needs may be more effectively diagnosed and treated through a developmental trauma perspective that recognizes their potentially shared origins. This does not negate the need for risk-reducing interventions targeting “criminogenic” needs, as focusing solely on addressing clinical needs cannot be expected to reduce the likelihood of further offending. In line with preventative healthcare and early intervention, findings suggest that promoting health service utilization among children and youth at risk of entering, or in early contact with, the justice system is important, as is making sure appropriate trauma-informed interventions for mental-health, substance use, and behavioral problems are provided in a variety of accessible contexts (Meurk et al., 2019). For deep-end justice-involved youth, like those in this study, results point to the possible benefit of screening and assessment procedures to identify poly-victimized youth potentially in need of more intensive monitoring and evidence-based mental health supports (Charak et al., 2019). Further research is needed to understand how some of the challenges faced by poly-victimized youth may affect their responsivity to criminogenic treatments. Such knowledge may lead to more direct recommendations about how to best meet the complex needs of poly-victimized youth whilst reducing reoffending.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This data utilized in this research was collected as part of a larger study that received funding from an Australian Research Council (ARC) Discovery Project grant (DP1095697).
