Child Abuse & Neglect 72 (2017) 247–257
Contents lists available at ScienceDirect
Child Abuse & Neglect journal homepage: www.elsevier.com/locate/chiabuneg
Full Length Article
Child maltreatment characteristics as predictors of heterogeneity in internalizing symptom trajectories among children in the child welfare system
MARK
Susan Yoon The Ohio State University, College of Social Work, 1947 College Rd. N., Columbus, OH 43210, United States
AR TI CLE I NF O
AB S T R A CT
Keywords: Child maltreatment Heterogeneity Internalizing symptoms Developmental trajectories Longitudinal
This study investigated heterogeneity in the developmental trajectories of internalizing symptoms among 541 children who were involved with the child welfare system and examined child maltreatment characteristics, including types, level of harm, and timing, as predictors of internalizing trajectory patterns. Secondary longitudinal research was conducted using data from the National Survey of Child and Adolescent Well-Being-I, collected from 1999 to 2007 in the United States. Three distinct trajectory groups were identified: high–decreasing; low–increasing; and low–stable Sexual abuse, emotional abuse, neglect, and more severe levels of harm from maltreatment predicted membership in two maladaptive groups compared to the low–stable group. The findings of the study suggest the importance of providing a thorough assessment of the type and severity of maltreatment experiences and continued monitoring of internalizing symptoms for children with child welfare involvement.
1. Introduction Substantial research has documented that children who experience maltreatment are at a greater risk of developing internalizing symptoms during childhood and adolescence (Appleyard, Yang, & Runyan, 2010; Bolger & Patterson, 2001). However, prior research reports that not all maltreated children develop internalizing symptoms and some children successfully achieve adaptive behavioral functioning despite their maltreatment experiences (Masten, Best, & Garmezy, 1990). These findings suggest that internalizing behavioral trajectories may be heterogeneous. Although studies have reported heterogeneous internalizing trajectories in the general population (e.g., Nivard et al., 2016), less is known about heterogeneity in patterns of internalizing trajectories among child welfareinvolved children as well as how early maltreatment experiences may shape distinct patterns of internalizing trajectories. A better understanding of heterogeneity in internalizing symptom trajectories and its predictors is critical to design and implement interventions that promote positive development among high-risk children. Various maltreatment characteristics, such as maltreatment types and level of harm from maltreatment, may influence the developmental trajectories of internalizing symptoms. Therefore, this study aims to investigate heterogeneity in internalizing symptom trajectories among child welfare-involved children and to examine maltreatment characteristics (i.e., type, level of harm, timing) as predictors of trajectory group membership. 1.1. Child maltreatment and internalizing symptoms Internalizing symptoms are characterized by problems within the self, such as anxiety, depression, social withdrawal, or somatic
E-mail address:
[email protected]. http://dx.doi.org/10.1016/j.chiabu.2017.08.022 Received 27 February 2017; Received in revised form 11 August 2017; Accepted 19 August 2017 0145-2134/ © 2017 Elsevier Ltd. All rights reserved.
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
complaints (Achenbach, Edelbrock, & Howell, 1987). A robust body of literature indicates a significant positive association between child maltreatment and internalizing symptoms (e.g., Bolger & Patterson, 2001; Robinson et al., 2009). Although the concurrent association between child maltreatment and internalizing symptoms has been clearly established, much less attention has been paid to the longitudinal effects of maltreatment on internalizing symptoms. In addition, the existing longitudinal studies have yielded mixed findings (e.g., Appleyard et al., 2010; Thompson & Tabone, 2010). Some studies have suggested a short-term effect of child maltreatment on children’s internalizing symptoms. In one study, for example, children’s early maltreatment experiences (ages 0–6) were significantly associated with their internalizing symptoms at age 6, but not at age 8 (Appleyard et al., 2010). Similarly, another longitudinal study reported that the impact of maltreatment on internalizing symptoms, although strong at the initial assessment, decreased over the course of eight years in boys at risk of maltreatment (Godinet, Li, & Berg, 2014). Other studies, in contrast, have found a long-term, enduring effect. For instance, Lansford et al. (2002) found that physical abuse in the first five years of life was associated with increased internalizing symptoms during adolescence. Additionally, a delayed effect of maltreatment was found in one study where early alleged maltreatment (< age 4) was not concurrently associated with children’s anxiety/depression, but was related to significant increases in anxiety/depression over time (Thompson & Tabone, 2010). The mixed findings from longitudinal empirical studies suggest that there may be individual differences in developmental trajectories of internalizing symptoms among maltreated children. 1.2. Heterogeneity in internalizing symptoms of maltreated children According to the developmental psychopathology perspective (Sroufe & Rutter, 1984), children’s early life experiences and subsequent environmental challenges affect their developmental outcomes over a life course. Developmental psychopathology emphasizes that multiple contributors affect one’s developmental outcomes and a myriad of pathways exist to adaptive or maladaptive behavior (Sroufe & Rutter, 1984). Therefore, similar negative experiences can result in various outcomes and pathways (Sroufe & Rutter, 1984). A growing body of child maltreatment literature identifies distinctive externalizing trajectories within maltreated children, exploring why some individuals are more susceptible to serious behavioral symptoms while others remain relatively unaffected (Tabone et al., 2011). However, only a small number of studies have investigated heterogeneity in internalizing symptom trajectories and yielded mixed results (Kim, Cicchetti, Rogosch, & Manly, 2009; Lauterbach & Armour, 2016; Proctor et al., 2010). Proctor et al. (2010) examined heterogeneity in caregiver reported internalizing symptom trajectories over an 8-year period (from age 6–14) for 279 children living in a Southwestern suburban city in the United States, and identified three subgroups: stable adjustment, mixed/ decreasing adjustment, and increasing adjustment. Using caregiver ratings of anxiety/depression symptoms at ages 4, 6, 8, 10, 12, and 14, Lauterbach and Armour (2016) found four distinct anxious/depressed symptom trajectories (low–stable, moderate–stable, moderate–increasing, high–decreasing) among 1354 U.S. children who have experienced or are at risk of maltreatment. In contrast to these studies, Kim et al. (2009) found no evidence of heterogeneity in internalizing symptom trajectories from age 6–10 among 249 maltreated and 200 non-maltreated children who attended a summer research camp in a Northeastern urban city in the United States. Using camp counselors’ ratings of internalizing symptoms, Kim et al. (2009) identified a single-class model as the best-fitting model for both maltreated and non-maltreated children. Possible explanations for the discrepancy in previous findings may include differences in study sample, developmental stage, study location (e.g., urban vs. suburban), informant (e.g., caregivers vs. camp counselors), operationalization of the outcome (e.g., anxiety/depression vs. internalizing symptoms), and other confounders affecting the outcome. The substantial differences in findings from prior studies suggest the need for further investigation of internalizing trajectories among maltreated children. 1.3. Maltreatment characteristics and internalizing symptoms Building on the developmental psychopathology perspective which highlights the importance of the nature and timing of the experience (Cicchetti & Toth, 1995), maltreatment experiences will have a different impact for an individual depending on the type, severity (i.e., level of harm), and timing of maltreatment. Some researchers have assessed how different forms of maltreatment may have differential impact on the development of internalizing and externalizing symptoms among children. Although study findings are not identical, some common themes have been revealed. In general, children who experienced sexual abuse (Tremblay, Hébert, & Piché, 1999) or neglect (Bolger & Patterson, 2001; Manly, Kim, Rogosch, & Cicchetti, 2001) were found to be at a higher risk of showing internalizing symptoms, whereas emotional or physical abuse were found to be more related to externalizing symptoms (Manly et al., 2001; Teisl & Cicchetti, 2008; Villodas, Litrownki, Newton, & Davis, 2016). Studies have also examined the role of the child’s age at the time of maltreatment and found contradictory findings. Some studies found no unique effect of the timing of maltreatment on internalizing symptoms (Jaffee & Maikovich-Fong, 2011; Robinson et al., 2009), whereas, others have indicated that children who experience maltreatment earlier in life exhibit greater levels of internalizing symptoms than children who experience maltreatment later in life (Keiley, Bates, Dodge, & Pettit, 2000; Kim & Cicchetti, 2010). For example, Kim & Cicchetti (2010) showed that earlier onset (0–36 months) of maltreatment, but not later onset (≥4 years), was associated with emotional dysregulation, which in turn led to later internalizing symptoms in school-aged children. In a sample of 578 children, early physical maltreatment (prior to age 5) was associated with greater levels of internalizing symptoms than physical abuse occurring at later periods (≥5 years) (Keiley et al., 2000). Preliminary evidence supports the impact of the severity of maltreatment on internalizing symptoms. Manly et al. (2001) found that severity of physical neglect, particularly when it occurred during the preschool period, was associated with greater internalizing 248
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
symptoms in 814 school-aged children. One study reported that more severe neglect was associated with higher levels of internalizing symptoms, although the severity of the other types of maltreatment did not have significant effects on these outcomes (Lynch & Cicchetti, 1998). To date, only a limited number of studies have examined the differential effects of diverse maltreatment dimensions beyond the global maltreatment effect and, thus far, have yielded inconclusive findings. Moreover, maltreatment characteristics have not yet been fully examined as predictors of internalizing trajectories of child welfare-involved children. 1.4. Other factors affecting patterns of internalizing symptom trajectories The relation between maltreatment and internalizing symptom trajectories may be influenced by multiple personal and contextual factors, such as risk and vulnerability and protective and promotive factors. The differential susceptibility to the environment framework suggests that various individual characteristics may make some people more vulnerable to the negative effects of adversity or, conversely, more likely to benefit from environmental resources and supports (Ellis, Boyce, Belsky, BakermansKranenburg, & Van IJzendoorn, 2011). A number of characteristics, including placement stability (Proctor et al., 2010), earlier onset of maltreatment (Kim et al., 2009), frequency of later physical abuse (Proctor et al., 2010), and behavioral health service utilization (Lauterbach & Armour, 2016) have been found to be predictive of membership in internalizing symptom trajectory groups and/or growth factors (initial point, rate of change). However, prior studies have produced mixed results for some of these factors in predicting internalizing trajectories. For example, chronic maltreatment significantly predicted membership in the moderate–stable, moderate–increasing, and high–decreasing depressive/anxious symptom classes in one study (Lauterbach & Armour, 2016), but did not predict the growth factors of internalizing symptoms in another study (Kim et al., 2009). Children’s sex predicted the rate of change in higher internalizing symptom classes in Lauterbach and Armour’s (2016) study, but did not predict internalizing trajectory growth factors in Kim et al.’s (2009) study. Finally, race/ethnicity has not been examined in prior internalizing symptom trajectory studies, but was found to be predictive of externalizing symptom trajectory patterns among child welfare-involved children, with Hispanic children having a higher probability of being in the worsening symptom group, when controlling for poverty and other covariates (Woodruff & Lee, 2011). 1.5. Present study Using data from a nationally representative longitudinal study of children who were referred to child protective services (CPS), this study addressed the following research questions: 1) To what extent is there heterogeneity in the patterns of internalizing symptom trajectories among children who were investigated by CPS for maltreatment?; and 2) How are maltreatment characteristics (i.e., type, level of harm, timing) associated with distinctive patterns of internalizing symptom trajectories? 2. Methods 2.1. Sample This study was a secondary data analysis of the National Survey of Child and Adolescent Well-Being (NSCAW-I), which is a nationally representative longitudinal study of children and families involved with the child welfare system. The full NSCAW-I sample (N = 6228) includes children who were between the ages of 0 and 14 at the time of sampling (October 1999 −December 2000). Data were collected in the United States from 1999 to 2007. Baseline (Wave 1) interviews occurred 2–6 months after the close of investigation and follow-up interviews occurred 12 months (Wave 2), 18 months (Wave 3), 36 months (Wave 4), and 59–96 months (Wave5) after baseline (Dowd et al., 2008). Wave 2 data were excluded from the analyses because the dependent variable was not collected at this point in time. Here after, Waves 1, 3, 4, and 5 are referred to respectively as Time 1 (T1), Time 2 (T2), Time 3 (T3), and Time 4 (T4). The analysis sample included 541 children who were between the ages of 4 and 5 years at baseline (T1). Children in the current sample were ages 4 and 5 years old (M = 4.47, SD = 0.50) at T1, 5–7 (M = 5.76, SD = 0.69) at T2, 6–8 (M = 7.09, SD = 0.70) at T3, and 8–13 (M = 11.04, SD = 0.95) at T4. Children in this sample were 51.2% male, 51.6% White/Non-Hispanic, 27.9% Black/ Non-Hispanic, and 20.5% Hispanic. Sample characteristics are presented in Table 1. Attrition analysis was performed to examine potential attrition bias produced by differential attrition. No significant differences were found in study variables between those who remained in the study versus those who dropped out. Missing data were handled using full information maximum likelihood (FIML). FIML allows respondents with missing data to be still included in the analyses for unbiased inference (Goldstein, 2009). The sample weights were not used in this study because the NSCAW-I weights are highly variant whole-sample weights and are not appropriate for use with small subsamples (Dowd et al., 2008). 2.2. Measures 2.2.1. Internalizing symptoms The Child Behavior Checklist for children age 4–18 (CBCL/4–18; Achenbach, 1991) was used to measure children’s internalizing symptoms. The internalizing symptoms scale (31 items) is composed of three subscales (withdrawn, somatic complaints, anxious/ depressed symptoms). Caregivers rated their child on the extent to which each item applied to their child, using the follow response 249
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
Table 1 Descriptives of Study Variables by Internalizing Trajectory Groups (N = 541). Internalizing trajectory group
Child gender (1 = female) Child race White/Non-Hispanic Black/Non-Hispanic Hispanic Below federal poverty level Out-of-home placement Behavioral health services Earlier onset of maltreatment More severe harm Physical abuse Sexual abuse Emotional abuse Neglect
Total %
High–decreasing %
Low–increasing %
Low–stable %
χ2
p
48.8
46.2
44.4
49.4
0.567 9.422
0.753 0.051
51.6 27.9 20.5 47.6 17.0 15.9 46.8 42.0 33.9 21.8 40.9 68.9
44.2 40.4 15.4 4.2 28.8 30.8 38.5 61.5 36.5 21.2 59.6 81.5
66.7 18.5 14.8 61.1 20.4 22.2 38.9 53.7 37.0 38.9 44.4 84.6
53.1 25.5 21.4 47.2 13.0 12.1 49.1 36.6 35.1 19.3 38.5 67.1
4.032 9.295 13.822 3.462 14.987 0.103 10.424 8.371 9.971
0.133 0.010 0.001 0.177 0.001 0.950 0.005 0.015 0.007
categories: 0 = not true, 1 = somewhat or sometimes true, 2 = very true or often true. Higher scores on the scales indicate greater internalizing symptoms. The CBCL/4–18 internalizing behavior raw score (summary score) was created by summing up the responses to the 31 items and raw internalizing behavior scores collected at T1 − T4 were used in the analysis. Raw scores of less than 15 are considered to be in the normal range, 16–19 in the borderline range, and scores greater or equal to 20 in the clinical range (Achenbach, 1991; Woodruff & Lee, 2011). Cronbach’s α ranged from 0.89–0.90. 2.2.2. Child maltreatment characteristics For the purpose of the study, three maltreatment characteristics were assessed: maltreatment type; level of harm from maltreatment (i.e., the harm to the child caused by the maltreatment); and the age at the onset of maltreatment (i.e., timing). Both unsubstantiated and substantiated maltreatment reports were used based on prior research which indicated that children with unsubstantiated cases exhibit similarly high levels of emotional and behavioral symptomatology as children with substantiated reports (Hussey et al., 2005). Other studies have also suggested that there are generally no distinct differences between unsubstantiated and substantiated reports, and substantiation is not a strong predictor of child welfare or child developmental outcomes (Kohl, JonsonReid, & Drake, 2009). For maltreatment type, four non-mutually exclusive variables—physical abuse, emotional abuse, sexual abuse, and neglect—were created using CPS reports and the Parent-Child Conflict Tactics Scales (CTS-PC; Straus, Hamby, Finkelhor, Moore, & Runyan, 1998). The child’s caregiver reported on the CTS-PC the frequency of physical abuse (severe and very severe physical assault scales: 8 items: e.g., beat the child, burned or scalded child on purpose), emotional abuse (3 items: e.g., swore or cursed at the child), sexual abuse (2 items: e.g., forced sexual contact), and neglect (5 items: e.g., unable to get child to a doctor when needed) in the past year, using the response scale ranging from 0 = never to 6 = more than 20 times. Caregiver responses were dichotomized (0 = no, 1 = yes) for each type of maltreatment and merged with the CPS maltreatment reports of each maltreatment type (0 = no, 1 = yes) in order to address the concerns of under-reporting of maltreatment in CPS records (Sedlak et al., 2010) and potential social desirability response biases in caregiver reports (Bennett, Sullivan, & Lewis, 2006). A child was coded as having the experience of maltreatment if either informant reported its presence. Past research has suggested that this approach tends to decrease the measurement error/bias (Horton, Laird, & Zahner, 1999). The level of harm from maltreatment was assessed using the CPS workers’ responses on a 4-point response scale (none, mild, moderate, severe) for the level of harm to the child item. None and mild levels of harm were coded as less severe harm (=0) and moderate and severe levels of harm were coded as more severe harm (=1). The timing variable was created using the caseworker–reported prior history of CPS reports. Children (the sample) in this study were 4 and 5 years old at baseline, which was 2–6 months following the close of the CPS investigation. However, it should be noted that the child’s age at baseline did not equate to the age of onset of maltreatment because they could have a prior history of CPS reports or involvement before their recruitment to NSCAW-I. Similar to guidelines from Jaffee and Maikovich-Fong (2011), the timing variable was created based on the developmental period in which maltreatment first occurred. The first CPS report during infancy/toddlerhood (age ≤ 3) was coded earlier onset (= 1) and the first CPS report after the preschool/school age period (age ≥ 4) was coded as later onset (= 0). 2.2.3. Control variables The control variables for this study were the child’s gender (0 = male and 1 = female) and race/ethnicity, placement status (0 = no and 1 = yes), receipt of behavioral health services (0 = no and 1 = yes), and family income reported by the caregiver at T1. Race was dummy coded (Black, Hispanic) using White as a reference group. Family income was reported by the caregiver and was recoded 1 = below the federal poverty level or 0 = above the federal poverty level. 250
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
2.3. Analysis plan Growth Mixture Modeling (GMM) was performed using Mplus version 7.3 to examine heterogeneity in the patterns of internalizing symptom trajectories in child welfare-involved children and to assess how maltreatment characteristics are associated with distinctive trajectory patterns. GMM was chosen as the primary analytical approach in the study because it is a powerful longitudinal statistical method that allows examination of heterogeneity in the changes of the outcome over time and identification of unobserved subgroups in the developmental trajectories of individuals (Jung & Wickrama, 2008). Following the recommendations of Jung and Wickrama (2008), the process of determining the number of latent trajectory groups began with latent class growth analysis (LCGA). LCGA, also known as group based trajectory modeling (GBTM), is an analytical technique for longitudinal data and represents a special form of GMM, in which the variance and covariance estimates for the growth factors are not allowed to vary across each latent class (Nagin, 2005; Jung & Wickrama, 2008). The CBCL/4–18 internalizing behavior raw scores collected at the four time points (T1, T2, T3, T4) were entered into a series of unconditional LCGA models with increasing class size. Determination of the optimal number of classes was based on fit statistics, parsimony, interpretability, and theoretical justification (Jung & Wickrama, 2008). Three measures of model fit were utilized: Bayesian Information Criterion (BIC); the Vuong-Lo-Mendell-Rubin likelihood ratio test (VLMRLRT); and the Parametric Bootstrapped likelihood ratio (BLRT). A smaller value of BIC indicates a better fitting model and statistically significant VLMR-LRT and BLRT values (p < 0.05) indicate a significant improvement in model fit (Jung & Wickrama, 2008). For models with more than one class, classification quality (i.e., precision of class assignment) was assessed by entropy (Muthen, 2004). An entropy statistic ranges from 0 to 1, with values closer to 1 representing higher classification accuracy (Jung & Wickrama, 2008). The proportions of individuals assigned to classes (i.e., class sizes) were also checked to ensure that each class has adequate class size (more than 5% of the total sample). Finally, graphic outputs were further assessed to check whether or not the identified classes were distinct and meaningful. Once the optimal number of classes was determined, additional model specification was made, such as constraining or freeing mean and variance parameters across latent classes, determining developmental shapes (e.g., linear, quadratic) within classes, and specifying variances of growth parameters (Jung & Wickrama, 2008). The models with quadratic terms were estimated based on empirical evidence that internalizing trajectories may be curvilinear (Bongers, Koot, van der Ende, & Verhulst, 2003). Starting with the invariant model, the parameters were freed step by step to examine if allowing them to vary across classes improves the fit or the interpretability of the model. Afterwards, the latent classes were interpreted and named. The next step involved specifying a conditional GMM by adding covariates (i.e., predictors and control variables) into the unconditional model. To examine how child maltreatment characteristics are associated with distinctive patterns of internalizing symptom trajectories (Research question 2), child maltreatment characteristics (i.e., type, level of harm, timing) and control variables (i.e., child’s gender, race, out-of-home placement, receipt of behavioral health services, and family income below the federal poverty level) were added as covariates into the GMM model using the ON statement in Mplus. In Mplus, the categorical latent class variable is related to the covariates by way of multinomial logistic regression (Jung & Wickrama, 2008; Muthen, 2004). Thus, adding covariates in the GMM model (i.e., conditional model) corresponds to multinomial logistic regression (Muthen, 2004), allowing the examination of child maltreatment characteristics in predicting membership in the identified internalizing symptom trajectories. 3. Results 3.1. Heterogeneity in internalizing behavior trajectories Table 2 displays model fit indices for the LCGA models of internalizing symptoms. The model fit indices provided mixed information regarding the optimal number of classes. The 5-class model had the lowest BIC score, but the VLMR-LRT suggested the 3class model as the best fitting model. BLRT statistics added no further information. Because the optimal number of classes was not clearly indicated by the model fit statistics, the models were carefully assessed for the quality and accuracy of the classification as well as the size and interpretability of the classes. The examination of graphic outputs (i.e., plots of the estimated latent trajectories) further confirmed the 3-class model to be superior to the 4-class or 5-class models. The classes in the 3-class model had distinctive and meaningful trajectories whereas the 4-class and 5-class models had poorly separated classes. Additionally, 4-class and 5-class models had one or more latent classes that had less than 5% of the sample. The 3-class model was selected based on VLMR-LRT, class size, classification accuracy, parsimony, and interpretability of the classes. Because adding a quadratic term significantly improved the model fit, Δc2(6) = 53.911, p < 0.001, a series of quadratic 3-class models were estimated to specify the best shape of the Table 2 Model Fit Indices for Internalizing Behavior LCGA Models. Model
BIC
VLMR
VLMR p-value
BLRT
BLRT p-value
Entropy
1-class 2-class 3-class 4-class 5-class
12056.361 11982.480 11945.972 11920.522 11909.904
– 88.095 52.602 42.101 28.014
– 0.6642 0.0438 0.5831 0.3552
– 92.761 55.389 44.331 29.498
– < 0.001 < 0.001 < 0.001 < 0.001
– 0.873 0.830 0.821 0.805
Note. BIC = Bayesian Information Criterion; VLMR = Lo-Mendell-Rubin Likelihood Ratio Test; BLRT = Parametric Bootstrapped Likelihood Ratio.
251
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
20
15
10
5
0 4
5
6
7
8
9
10
11
12
Age in Years High–Decreasing (9%)
Low–Increasing (16%)
Low–Stable (75%)
Fig. 1. Latent trajectories of internalizing behavior problems.
trajectories. In the final model, the variances and covariances of the linear and quadratic factors were set to zero. The 3-class internalizing symptom trajectory GMM model (Fig. 1) consisted of the following subgroups: high–decreasing, low–increasing, and low–stable. Approximately 9% (n = 49) of the sample was in the high–decreasing group. Children in this group exhibited clinical levels of internalizing symptoms during early childhood (ages 4–5), followed by a gradual decrease. The low–increasing group comprised 16% (n = 87) of the sample. Children in this group started in the normal range of internalizing symptoms but showed gradually increasing symptoms, peaking during early school age years (ages 6–8) with the clinical range of internalizing symptoms. Then, internalizing symptoms decreased over time after age 7, although stayed in the borderline range throughout the school age period. Children in the low–stable internalizing symptom group represented 75% (n = 405) of the sample. Children in this group exhibited low and normal levels of internalizing symptoms throughout the four data time points. 3.2. Key study variables by internalizing trajectory groups The bivariate relationship between study variables and the three internalizing trajectory groups are displayed in Table 1. Children who experienced more severe level of harm from maltreatment, χ2(2) = 14.987, p =0.001, and emotionally abused children, χ2(2) = 8.371, p =0.015, were more likely to be in the high–decreasing group compared to the low − stable group. Children who were placed in out-of-home care or receiving behavioral health services were also more likely to be in the high–decreasing group compared to the low–stable group. Sexually abused children, χ2(2) = 10.424, p =0.005, and neglected children, χ2(2) = 9.971, p =0.007, were more likely to be in the low–increasing group compared to the low − stable group. No other variables had significant bivariate relationships with the three internalizing trajectory groups. 3.3. Predictors of internalizing trajectory group membership 3.3.1. High–decreasing group compared with low–stable group Emotional abuse and the level of harm from maltreatment significantly predicted membership in the high–decreasing group compared to the low − stable group, when controlling for maltreatment timing, physical abuse, sexual abuse, neglect, child’s gender, race, placement status, receipt of behavioral health services, and family income. Children who were rated by CPS workers as experiencing more severe harm from maltreatment had 3 times higher odds (p = 0.035) of being in the high–decreasing group compared to the low–stable group. Emotionally abused children had 3.45 times higher odds (p = 0.017) of being in the high–decreasing group compared to the low–stable group. Children who were in out-of-home care and receiving behavioral health interventions had 4.75 (p = 0.017) and 3.92 (p = 0.017) times higher odds of being in this group, respectively, compared to the low–stable group. 3.3.2. Low–increasing group compared with low–stable group Sexual abuse and neglect significantly predicted membership in the low–increasing group compared to the low − stable group, when controlling for maltreatment timing, severity, physical abuse, emotional abuse, child’s gender, race, placement status, receipt of behavioral health services, and family income. Sexually abused children had 3.75 times higher odds (p = 0.017) of being in the low–increasing group compared to the low–stable group. Neglected children had 2.83 times higher odds (p = 0.033) of being in the low–increasing group in comparison to the low–stable group. Household income below the federal poverty level also predicted membership in this group compared to the low − stable group. 3.3.3. High–decreasing group compared with low–increasing group No child maltreatment characteristics significantly predicted membership in the high − decreasing group compared to the low − increasing group. Yet, children who were Black or living in higher-income families (i.e., above the federal poverty level) were 252
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
Table 3 Child Maltreatment as Predictors of Internalizing Trajectory Group Membership. Reference Group
Low–stable
Comparison Group (s)
High–decreasing
Earlier onset of maltreatment More severe harm Physical abuse Emotional abuse Sexual abuse Neglect Gender (1 = female) Racea Black/Non-Hispanic Hispanic Below federal poverty level Out-of-home placement Behavioral health services
Low–increasing Low–increasing
High–decreasing
OR
95% CI
OR
95% CI
OR
95% CI
0.69 2.93* 1.38 3.45* 1.27 1.87 1.13
0.31–1.53 1 0.08–7.92 0.54–3.52 1.25–9.49 0.51–3.55 0.86–4.09 0.50–2.53
0.75 1.67 1.70 1.19 3.75* 2.83* 0.71
0.35–1.58 0.64–4.36 0.69–4.16 0.55–2.55 1.26–9.36 1.09–7.35 0.36–1.37
0.92 1.12 0.81 2.91 0.34 1.04 1.60
.34 .37 .27 .87 .09 .29 .63
2.18. 0.96 0.95 4.75** 3.92**
0.98–4.79 0.34–2.68 0.43–2.09 1.49–15.15 1.49–10.79
0.48 0.58 2.71* 3.21 2.37
0.18–1.32 0.20–1.70 1.21–6.06 0.98–8.73 0.91–6.22
4.53** 1.65 0.35* 1.48 1.65
1.45 − 14.04 .43 − 6.40 .13 − 0.95 .33 − 6.58 .49 − 5.56
− − − − − − −
2.46 3.46 2.44 9.76 1.26 3.76 4.05
Note. a White/Non-Hispanic is the reference group; OR = odds ratio. * p < 0.05. ** p < 0.01.
more likely to be in the high − decreasing group compared to the low − increasing group. The results from the conditional GMM model are summarized in Table 3.
4. Discussion This study examined heterogeneity in patterns of developmental trajectories of internalizing symptoms among children who were investigated by CPS for maltreatment, and assessed the extent to which different maltreatment characteristics (i.e., type, level of harm, timing) predict distinctive patterns of internalizing symptom trajectories. This study found subgroups (i.e., low–stable, low–increasing, and high–decreasing) of children who displayed distinct patterns of internalizing symptoms over time. Although children in this study all had the experience of being involved with the child welfare system, their internalizing trajectory patterns varied greatly. The results are in line with developmental psychopathology (Sroufe & Rutter, 1984), which suggests diversity in developmental processes and outcomes among individuals with similar early adverse life events (Cicchetti & Toth, 1995). The findings are also consistent with previous empirical research that suggested heterogeneity in developmental trajectories of internalizing symptoms in maltreated children (Lauterbach & Armour, 2016; Proctor et al., 2010) and the normative/community sample (Nivard et al., 2016; Sterba, Prinstein, & Cox, 2007). In general population studies, three internalizing trajectories (elevated–stable, decreasing/increasing, and low) were identified among 1364 children followed from age 2–11 years (Sterba et al., 2007) and five internalizing trajectories (very low, low, decreasing, increasing, adolescent increasing) were found among 7202 children in the United Kingdom followed from age 7–15 years. Although the internalizing trajectory groups identified in the current study may not be directly comparable to the trajectory groups identified in previous studies due to differences in sample (e.g., high-risk vs. normative, USA vs. UK, different age ranges) and measures, some similar findings were observed across these studies. In all studies, the largest portion of the sample exhibited stable and low levels of internalizing symptoms throughout childhood. These results are in line with findings from the broader child development literature which depicts a stable internalizing symptom trajectory as normative behavioral development in childhood (Bongers et al., 2003; Keiley et al., 2000). This study’s finding of a substantially large proportion of children showing low–stable internalizing symptoms despite their involvement with CPS may suggest resiliency within these children. Children who experience early adversities—including stress and trauma—may still achieve successful adaptation when they have protective factors that buffer the negative effects of such adversities (Masten et al., 1990). However, the large portion of children in the low–stable group in this study should be interpreted with caution because simply treating these children as resilient individuals can be misleading. Internalizing symptoms tend to be internal, hidden, and unnoticeable, making it difficult for outside observers to detect (Tandon, Cardeli, & Luby, 2009). Therefore, low levels of internalizing symptoms found in the low–stable group may be, in part, due to caregivers’ failure to identify non-obvious internalizing symptoms. The other two subgroups (i.e., high–decreasing, low–increasing) together represented one quarter of the study sample. These two groups are of clinical concern and warrant mental health attention compared to the low–stable group, although the downward trajectories of internalizing symptoms (e.g., after age 7 in the low–increasing group) may suggest evidence of resiliency and recovery. The identification of these two subgroups (i.e., high–decreasing, low–increasing) indicates the importance of taking a person-centered approach, as opposed to a variable-centered approach, in accurately describing diverse patterns of longitudinal change in the development of internalizing symptoms in maltreated children. 253
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
The high–decreasing group (9%) included children who showed clinical levels of internalizing symptoms during early childhood (age 4–5), followed by gradual decrease in internalizing symptoms during school age. This group is similar to Proctor et al.’s (2010) increasing adjustment group (8%) that represented gradually decreasing internalizing behaviors during school age. Emotionally abused children and children who experienced more severe harm from maltreatment were more likely to be in the high–decreasing group relative to the low–stable group. This finding may suggest that emotional abuse and severe level of harm from maltreatment had strong associations with children’s internalizing symptoms at baseline, but the strength of these associations became attenuated over time. One possible explanation for this high–decreasing internalizing symptom pattern is that children were emotionally abused or severely harmed from maltreatment at baseline but no longer experienced maltreatment during the subsequent years. For instance, it is possible that these children were later placed in out-of-home care and the out-of-home placements prevented their continued experience of abuse, contributing to the decrease in internalizing symptoms over time. Research has indicated that victims of transitory maltreatment are less likely to show internalizing symptoms over time compared to victims of chronic maltreatment (Ethier, Lemelin, & Lacharite, 2004; Jaffee & Maikovich-Fong, 2011). Future research should include time-varying predictors, such as the chronicity of maltreatment and caregiver stability to see how the changes in children’s environment influence internalizing trajectories of young maltreated children. Another possible explanation is that these children received behavioral health services and interventions because CPS workers perceived them as a high-risk group and made referrals to behavioral health professionals. Such services and interventions may have reduced internalizing symptoms in these children over time. Supporting these speculations, outof-home placement and behavioral health service utilization significantly predicted membership in the high–decreasing group in the current study. The low–increasing group (16%) included children who showed gradually increasing internalizing symptoms until early school age years (age 6–8). This trajectory pattern aligns with Gilliom and Shaw’s (2004) study of 303 boys (age 2–6) that found a gradually increasing trajectory of internalizing problems. Interestingly, once peaked at age 7, internalizing symptoms substantially decreased over time during school age years (i.e., T3 and T4). This finding may be explained by children’s transition to school. Whereas only 40% of the children were in elementary school at T2, approximately 96% of the children were in elementary school at T3 and 100% of the children were in either elementary or middle school at T4. Although transition to school and peer/teacher support variables were not examined in this study, it may be that children benefited from being surrounded by teachers and peers, as teacher or peer support can decrease feelings of rejection or loneliness in children. Research has shown that higher peer support is significantly associated with lower levels of internalizing symptoms in maltreated children (Ezzell, Swenson, & Brondino, 2000). Sexual abuse and neglect predicted membership in the low–increasing group compared to the low–stable group. This finding may imply that sexual abuse and neglect have a delayed, long-lasting impact on children’s internalizing symptoms, suggesting the importance of ongoing monitoring for sexually abused and/or neglected children. Consistent with these results, prior research has suggested that sexual abuse has long-term negative effects on children’s internalizing symptoms (Swanston et al., 2003). Although psychological injuries and damages were not assessed in this study, they may explain the influence of sexual abuse on internalizing symptoms over time. Victims of sexual abuse are likely to experience psychological damages such as shame, self-blame, guilt, powerlessness, stigmatization, and low self-efficacy (Swanston et al., 2003), which may lead to increasing and ongoing internalizing symptoms (Bolger & Patterson, 2001). A possible explanation for neglect predicting an increase in internalizing symptoms during early childhood and early school age years may be that neglectful caregivers failed to identify their children’s early signs of internalizing symptoms at T1, allowing the symptoms to worsen. Neglectful caregivers—often characterized by their insensitive and unresponsive parenting—may have found it difficult to detect their children’s internalizing symptoms because internalizing symptoms tend to be quiet and unobservable (Tandon et al., 2009). The inability to detect internalizing symptoms may be especially true when children are young and have limited verbal skills to accurately describe their internal feelings, negative emotions, and stress (Tandon et al., 2009). It could be that neglectful caregivers’ recognition of their child’s internalizing symptoms increased over time as the symptoms grew more serious and the child began to verbally express their mood states, such as depressed or anxious feelings. It may also be that neglected children perceived themselves as helpless in controlling their life events or outcomes and developed internalizing symptoms over time (Bolger & Patterson, 2001). It is noteworthy that physical abuse was not significantly associated with any of the identified internalizing trajectory groups. The lack of association between physical abuse and internalizing trajectory groups may be because physical abuse was measured only at baseline in this study. The influence of early physical abuse on internalizing symptoms may not be as strong as that of more proximal incidents of physical abuse (Proctor et al., 2010). Another possibility may be that physical abuse, in fact, is not related to children’s internalizing behavioral adaptations over time. Physical abuse has been reported to be more strongly associated with externalizing symptoms rather than internalizing symptoms in prior research (e.g., Manly et al., 2001; Teisl & Cicchetti, 2008; Villodas et al., 2016). The timing of maltreatment was also not associated with membership in any of the identified internalizing trajectory groups. Previous research has produced mixed results regarding the role of maltreatment timing on children’s internalizing symptoms, with some reporting no effect of maltreatment timing (e.g., Jaffee & Maikovich-Fong, 2011; Robinson et al., 2009) and others reporting elevated internalizing symptoms in children with earlier onset of maltreatment (e.g., Keiley et al., 2000; Kim & Cicchetti, 2010). It could be that the operationalization of the timing variable in this study led to a false negative result. Other methods, such as examining the age at the onset of maltreatment as a continuous variable or using a younger age threshold to define an earlier onset of maltreatment, may have yielded different results. The lack of association between maltreatment timing and internalizing trajectory patterns in this study may also be because maltreatment timing alone is not sufficient to predict children’s internalizing behavioral adaptations over time; perhaps the interactions between the timing of maltreatment and other maltreatment characteristics need to 254
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
be considered. For example, it may be the interaction between earlier onset and chronic maltreatment that predicts poorer internalizing symptom trajectories, rather than earlier onset itself. Research has shown significant interaction effects between maltreatment timing and other maltreatment characteristics on children’s internalizing symptoms (Jaffee & Maikovich-Fong, 2011; Manly et al., 2001). For example, severity of neglect was associated with internalizing symptoms and withdrawn behavior when it occurred during the preschool period, as opposed to the school age period (Manly et al., 2001). Future research should investigate the interaction effects between the timing of maltreatment and other maltreatment characteristics on internalizing trajectories. 4.1. Strengths and limitations This study has several limitations. First, all study covariates were assessed only at baseline and was not able to examine how the changes in maltreated children and their surrounding environment may be related to children’s internalizing symptom trajectories. For instance, longitudinal maltreatment information—such as chronic maltreatment, changes in maltreatment types, and maltreatment occurring at different developmental periods—was not assessed in this study. Most notably, chronic maltreatment could not be examined in the current study due to limitations in the data, but warrants further attention given that previous studies have yielded inconsistent findings regarding the influence of chronic maltreatment on internalizing trajectories (Kim et al., 2009; Lauterbach & Armour, 2016). Future research should address how the chronicity of maltreatment shapes patterns of internalizing symptom trajectories. Second, some of the reported effects had relatively wide confidence intervals, suggesting potential poor precision of the effect estimates. Several factors such as heterogeneity in the sample (e.g. inclusion of both unsubstantiated and substantiated maltreatment cases) or covariates with sparse data (i.e., low variance, small cell counts) in latent classes may have contributed to the wide confidence intervals (Warner, 2013). There may also be some potentially important variables that were not assessed in the study. For example, emotional regulation skills are important correlates of internalizing symptoms (Kim et al., 2009), but were not examined in the study due to the lack of data. Additionally, broader environmental factors, including peer relationships (Lauterbach & Armour, 2016) and exposure to community violence (Goldner, Gross, Richards, & Ragasdale, 2015) may have significant influence on children’s symptom trajectories but were not included in the study. Third, this study is limited by the use of the caregiver-reported measure of children’s internalizing symptoms. It has been suggested that caregivers often underreport their children’s internalizing symptoms due to its hidden and unnoticeable nature (Kamphaus & Frick, 1996; Tandon et al., 2009). This concern may be amplified in child welfare sample given that maltreating parents tend to have a biased perception of their children (Widom, Czaja, & Dumont, 2015). Nonetheless, previous research with families of youth exposed to trauma found that caregiver reports in the CBCL/4–18 were consistent with clinical interviews by child psychiatrists/psychologists (Saigh, Yasik, Oberfield, Halamandaris, & McHugh, 2002). Furthermore, the parent and teacher ratings of maltreated children’s behavior problems were found to be similar in a study of child behavior in maltreating families (Salzinger, Kaplan, Pelcovitz, Samit, & Krieger, 1984). Future studies may benefit from using multiple informants to address the limitations of solely relying on caregiver reports (De Los Reyes & Kazdin, 2005). Fourth, consistent with prior research (e.g., Everson et al., 2008; Swahn et al., 2006), the concordance between CPS reports and caregiver reports were low to moderate in this sample. Nevertheless, despite the low to moderate concordance rates, researchers advocate for the need to use multiple sources in detecting child abuse and neglect (e.g., Swahn, 2006) because of the strengths and weaknesses of each reporting source. It should also be noted that some of the items (e.g., unable to ensure that the child gets food needed) in the CTS-PC neglect scale may not sensitively discriminate between caregivers’ intentional neglectful behaviors and poverty or behaviors related to poverty. On a related note, this study was limited in that the co-occurrence of maltreatment subtypes was not considered. For instance, children who experience neglect only may have different trajectories from children who experience multiple types of maltreatment. Future research should investigate the role of co-occurring maltreatment types in determining patterns of internalizing trajectories. Finally, the sample composition limits the generalizability of study findings. This study used a sample drawn from children who were referred to the child welfare system and thus, study findings may not be generalized to children who have not been referred to CPS. Despite these limitations, this study has several distinct strengths. The use of longitudinal outcome data collected at four time points over the course of eight years allowed the examination of the long-term effects of child maltreatment on internalizing symptoms over time. In addition, this study addressed methodological limitations of the conventional growth modeling approaches which assume that a single trajectory fits an entire population. Using GMM, this study further advanced our knowledge about distinctive trajectory patterns of internalizing symptoms. The findings of this study enabled the discussion and suggestion of clinical implications that can lead to long-term, positive behavioral adjustment in child welfare-involved children. 4.2. Implications The findings from this study provide several important research and practice implications. The study’s results indicate that children involved in the child welfare system represent heterogeneous populations that exhibit distinctive patterns of internalizing symptoms over time, and diverse maltreatment characteristics—including maltreatment type and level of harm from maltreatment—predict these trajectory patterns. Therefore, clinicians and practitioners should view child welfare-involved children as a heterogeneous population and thoroughly assess their maltreatment experiences, acknowledging that different features of maltreatment (e.g., type, level of harm, timing) may differently influence the child’s internalizing symptoms over time. In addition, 255
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
ongoing assessment and monitoring of behavioral development is needed for children investigated for child maltreatment. Children in the low–increasing group had similar initial points (i.e., low levels of internalizing symptoms) with the low–stable group, but substantially different trajectories (i.e., escalating internalizing symptoms over time), underscoring the need for continued behavioral assessment to identify emerging signs of future internalizing symptoms. Although this study’s findings suggest preliminary evidence of heterogeneity in internalizing symptom trajectories of children involved with the child welfare system, the findings need to be replicated and further verified in future research using other measures of children’s behavior, such as teacher-reports or observational measure of internalizing behaviors. Furthermore, future research should test the external validity of extracted latent classes by using class membership as a predictor for distal outcomes, such as adolescent suicidal behavior or substance use. Future research is also needed to more fully investigate child welfare-involved children’s continued behavioral development through later developmental stages (e.g., adolescence, young adulthood). Acknowledgments This document includes data from the National Survey on Child and Adolescent Well-Being, which was developed under contract with the Administration on Children, Youth, and Families, U.S. Department of Health and Human Services (ACYF/DHHS). The data were provided by the National Data Archive on Child Abuse and Neglect. Funded through the Department of Health and Human Services, Administration for Children and Families, Children's Bureau, Grant #90CA1817. The contents of this publication do not necessarily reflect the views or policies of the funders, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S Department of Health and Human Services. This information is in the public domain. Readers are encouraged to copy and share it, but please credit the author. References Achenbach, T. M., Edelbrock, C., & Howell, C. T. (1987). Empirically based assessment of the behavioral/emotional problems of 2-and 3-year-old children. Journal of Abnormal Child Psychology, 15(4), 629–650. Achenbach, T. (1991). Child behavior checklist/4-18. Burlington, VT: University of Vermont. Appleyard, K., Yang, C., & Runyan, D. K. (2010). Delineating the maladaptive pathways of child maltreatment: A mediated moderation analysis of the roles of selfperception and social support. Development and Psychopathology, 22(2), 337–352. Bennett, D. S., Sullivan, M. W., & Lewis, M. (2006). Relations of parental report and observation of parenting to maltreatment history. Child Maltreatment, 11(1), 63–75. Bolger, K. E., & Patterson, C. J. (2001). Pathways from child maltreatment to internalizing problems: Perceptions of control as mediators and moderators. Development and Psychopathology, 13(04), 913–940. Bongers, I. L., Koot, H. M., van der Ende, J., & Verhulst, F. C. (2003). The normative development of child and adolescent problem behavior. Journal of Abnormal Psychology, 112(2), 179–192. Cicchetti, D., & Toth, S. L. (1995). A developmental psychopathology perspective on child abuse and neglect. Journal of the American Academy of Child and Adolescent Psychiatry, 34(5), 541–565. http://dx.doi.org/10.1097/00004583-199505000-00008. De Los Reyes, A., & Kazdin, A. E. (2005). Informant discrepancies in the assessment of childhood psychopathology: A critical review, theoretical framework, and recommendations for further study. Psychological Bulletin, 131(4), 483–509. Dowd, K., Kinsey, S., Wheeless, S., Thissen, R., Richardson, J., Suresh, R., ... Smith, K. (2008). National survey of child and adolescent well-being (NSCAW) combined waves 1–5 data file user's manual- restricted release version. Ithaca, NY: National Data Archive on Child Abuse and Neglect, Cornell University. Ellis, B. J., Boyce, W. T., Belsky, J., Bakermans-Kranenburg, M. J., & Van IJzendoorn, M. H. (2011). Differential susceptibility to the environment: An evolutionary–neurodevelopmental theory. Development and Psychopathology, 23(01), 7–28. Ethier, L. S., Lemelin, J. P., & Lacharite, C. (2004). A longitudinal study of the effects of chronic maltreatment on children's behavioral and emotional problems. Child Abuse & Neglect, 28(12), 1265–1278 S0145-2134(04)00256-X [pii]. Everson, M. D., Smith, J. B., Hussey, J. M., English, D., Litrownik, A. J., Dubowitz, H., ... Runyan, D. K. (2008). Concordance between adolescent reports of childhood abuse and child protective service determinations in an at-risk sample of young adolescents. Child Maltreatment, 13(1), 14–26. Ezzell, C. E., Swenson, C. C., & Brondino, M. J. (2000). The relationship of social support to physically abused children’s adjustment. Child Abuse & Neglect, 24(5), 641–651. Gilliom, M., & Shaw, D. S. (2004). Codevelopment of externalizing and internalizing problems in early childhood. Development and Psychopathology, 16(2), 313–333. Godinet, M. T., Li, F., & Berg, T. (2014). Early childhood maltreatment and trajectories of behavior problems: Exploring gender and racial differences. Child Abuse & Neglect, 38(3), 544–556. http://dx.doi.org/10.1016/j.chiabu.2013.07.018 [doi]. Goldner, J., Gross, I. M., Richards, M. H., & Ragsdale, B. L. (2015). The relation of severity and type of community violence exposure to emotional distress and problem behaviors among urban African American adolescents. Violence and Victims, 30(3), 432–449. Goldstein, H. (2009). Handling attrition and non-response in longitudinal data? Longitudinal and Life Course Studies, 1(1), 63–72. Horton, N. J., Laird, N. M., & Zahner, G. E. (1999). Use of multiple informant data as a predictor in psychiatric epidemiology. International Journal of Methods in Psychiatric Research, 8(1), 6–18. Hussey, J. M., Marshall, J. M., English, D. J., Knight, E. D., Lau, A. S., Dubowitz, H., & Kotch, J. B. (2005). Defining maltreatment according to substantiation: Distinction without a difference? Child Abuse & Neglect, 29(5), 479–492. Jaffee, S. R., & Maikovich‐Fong, A. K. (2011). Effects of chronic maltreatment and maltreatment timing on children’s behavior and cognitive abilities. Journal of Child Psychology and Psychiatry, 52(2), 184–194. Jung, T., & Wickrama, K. (2008). An introduction to latent class growth analysis and growth mixture modeling. Social and Personality Psychology Compass, 2(1), 302–317. Kamphaus, R. W., & Frick, P. J. (1996). Clinical assessment of child and adolescent personality and behavior. Needham Heights, MA: Allyn & Bacon. Keiley, M. K., Bates, J. E., Dodge, K. A., & Pettit, G. S. (2000). A cross-domain growth analysis: Externalizing and internalizing behaviors during 8 years of childhood. Journal of Abnormal Child Psychology, 28(2), 161–179. Kim, J., & Cicchetti, D. (2010). Longitudinal pathways linking child maltreatment, emotion regulation, peer relations, and psychopathology. Journal of Child Psychology and Psychiatry, 51(6), 706–716. Kim, J., Cicchetti, D., Rogosch, F. A., & Manly, J. T. (2009). Child maltreatment and trajectories of personality and behavioral functioning: Implications for the development of personality disorder. Development and Psychopathology, 21(3), 889–912. Kohl, P. L., Jonson-Reid, M., & Drake, B. (2009). Time to leave substantiation behind findings from a national probability study. Child Maltreatment, 14(1), 17–26. Lansford, J. E., Dodge, K. A., Pettit, G. S., Bates, J. E., Crozier, J., & Kaplow, J. (2002). A 12-year prospective study of the long-term effects of early child physical maltreatment on psychological, behavioral, and academic problems in adolescence. Archives of Pediatrics & Adolescent Medicine, 156(8), 824–830 poa20065. Lauterbach, D., & Armour, C. (2016). Symptom trajectories among child survivors of maltreatment: Findings from the Longitudinal Studies of Child Abuse and Neglect
256
Child Abuse & Neglect 72 (2017) 247–257
S. Yoon
(LONGSCAN). Journal of Abnormal Child Psychology, 44(2), 369–379. Lynch, M., & Cicchetti, D. (1998). An ecological-transactional analysis of children and contexts: The longitudinal interplay among child maltreatment, community violence, and children's symptomatology. Development and Psychopathology, 10(2), 235–257. Manly, J. T., Kim, J. E., Rogosch, F. A., & Cicchetti, D. (2001). Dimensions of child maltreatment and children's adjustment: Contributions of developmental timing and subtype. Development and Psychopathology, 13(4), 759–782. Masten, A. S., Best, K. M., & Garmezy, N. (1990). Resilience and development: Contributions from the study of children who overcome adversity. Development and Psychopathology, 2(04), 425–444. Muthen, B. (2004). Latent variable analysis: Growth mixture modeling and related techniques for longitudinal data. In D. Kaplan (Ed.), Handbook of quantitative methodology for the social sciences (pp. 345–368). Newbury Park, CA: Sage. Nagin, D. (2005). Group-based modeling of development. Cambridge, MA: Harvard University Press. Nivard, M. G., Lubke, G. H., Dolan, C. V., Evans, D. M., Pourcain, B. S., Munafò, M. R., & Middeldorp, C. M. (2016). Joint developmental trajectories of internalizing and externalizing disorders between childhood and adolescence. Development and Psychopathology, 1–10. Proctor, L. J., Skriner, L. C., Roesch, S., & Litrownik, A. J. (2010). Trajectories of behavioral adjustment following early placement in foster care: Predicting stability and change over 8 years. Journal of the American Academy of Child and Adolescent Psychiatry, 49(5), 464–473. Robinson, L. R., Morris, A. S., Heller, S. S., Scheeringa, M. S., Boris, N. W., & Smyke, A. T. (2009). Relations between emotion regulation, parenting, and psychopathology in young maltreated children in out of home care. Journal of Child and Family Studies, 18(4), 421–434. Saigh, P. A., Yasik, A. E., Oberfield, R. A., Halamandaris, P. V., & McHugh, M. (2002). An analysis of the internalizing and externalizing behaviors of traumatized urban youth with and without PTSD. Journal of Abnormal Psychology, 111(3), 462. Salzinger, S., Kaplan, S., Pelcovitz, D., Samit, C., & Krieger, R. (1984). Parent and teacher assessment of children's behavior in child maltreating families. Journal of the American Academy of Child Psychiatry, 23(4), 458–464. Sedlak, A. J., Mettenburg, J., Basena, M., Peta, I., McPherson, K., & Greene, A. (2010). Fourth national incidence study of child abuse and neglect (NIS-4). Washington, DC: US Department of Health and Human Services. Sroufe, L. A., & Rutter, M. (1984). The domain of developmental psychopathology. Child Development, 55, 17–29. Sterba, S. K., Prinstein, M. J., & Cox, M. J. (2007). Trajectories of internalizing problems across childhood: Heterogeneity, external validity, and gender differences. Development and Psychopathology, 19(02), 345–366. Straus, M. A., Hamby, S. L., Finkelhor, D., Moore, D. W., & Runyan, D. (1998). Identification of child maltreatment with the Parent-Child Conflict Tactics Scales: Development and psychometric data for a national sample of American parents. Child Abuse & Neglect, 22(4), 249–270. Swahn, M. H., Whitaker, D. J., Pippen, C. B., Leeb, R. T., Teplin, L. A., Abram, K. M., & McClelland, G. M. (2006). Concordance between self-reported maltreatment and court records of abuse or neglect among high-risk youths. American Journal of Public Health, 96(10), 1849–1853. Swanston, H. Y., Plunkett, A. M., O’Toole, B. I., Shrimpton, S., Parkinson, P. N., & Oates, R. K. (2003). Nine years after child sexual abuse. Child Abuse & Neglect, 27(8), 967–984. Tabone, J. K., Guterman, N. B., Litrownik, A. J., Dubowitz, H., Isbell, P., English, D. J., ... Thompson, R. (2011). Developmental trajectories of behavior problems among children who have experienced maltreatment heterogeneity during early childhood and ecological predictors. Journal of Emotional and Behavioral Disorders, 19(4), 204–216. Tandon, M., Cardeli, E., & Luby, J. (2009). Internalizing disorders in early childhood: A review of depressive and anxiety disorders. Child and Adolescent Psychiatric Clinics of North America, 18(3), 593–610. http://dx.doi.org/10.1016/j.chc.2009.03.004 [doi]. Teisl, M., & Cicchetti, D. (2008). Physical abuse, cognitive and emotional processes, and Aggressive/Disruptive behavior problems. Social Development, 17(1), 1–23. Thompson, R., & Tabone, J. K. (2010). The impact of early alleged maltreatment on behavior trajectories. Child Abuse & Neglect, 34(12), 907–916. Tremblay, C., Hébert, M., & Piché, C. (1999). Coping strategies and social support as mediators of consequences in child sexual abuse victims. Child Abuse & Neglect, 23(9), 929–945. Villodas, M. T., Litrownik, A. J., Newton, R. R., & Davis, I. P. (2016). Long-term placement trajectories of children who were maltreated and entered the child welfare system at an early age: Consequences for physical and behavioral well-being. Journal of Pediatric Psychology, 41(1), 46–54. http://dx.doi.org/10.1093/jpepsy/ jsv031. Warner, R. (2013). Applied statistics: From bivariate through multivariate techniques (2nd ed.). Los Angeles, CA: Sage. Widom, C. S., Czaja, S. J., & DuMont, K. A. (2015). Intergenerational transmission of child abuse and neglect: Real or detection bias? Science, 347(6229), 1480–1485. Woodruff, K., & Lee, B. (2011). Identifying and predicting problem behavior trajectories among pre-school children investigated for child abuse and neglect. Child Abuse & Neglect, 35(7), 491–503. http://dx.doi.org/10.1016/j.chiabu.2011.03.007.
257