Longitudinal Adaptive Pattern Analysis of Child Abuse Victims During Childhood and Adolescence

Abstract

Adopting a contextual, systematic perspective, the present study aimed to understand whether an individual’s adverse circumstances in one domain have a continuous effect on his or her maladaptation in another domain. The Korean Children and Youth Panel Survey data set, comprising a stratified sample of 1,932 students recruited from 95 schools from 16 cities in Korea, was used for pattern identification, latent class analysis, and latent transition analysis. Consistent with Masten’s resilience model, latent class analysis was used to identify the following four types of patterns: resilient, maladaptive, vulnerable, and competent-unchallenged. These four patterns were clearly identified at Time 1, Time 2, and Time 3. Latent transition analysis was used to identify the continuity and change patterns in the four groups across the developmental pathology. The transition probabilities from Time 1 to Time 2 were relatively unstable, with many variations observed among the latent classes; however, the transition probabilities from Time 2 to Time 3 remained relatively stable. An in-depth discussion of the findings and their implications is provided.

Keywords

Resilience child abuse school adaptation longitudinal pattern analysis latent class analysis latent transition analysis

Introduction

Although individuals are oriented toward well-being and happiness, their pursuit of happiness is often interrupted or forfeited by potential threats to their psychological well-being and development. Child abuse is one of the problems facing children and adolescents today. According to the report 2016 Child Abuse and Neglect Korea (Ministry of Health and Welfare, 2016), there has been a steady increase in the cases of child abuse in Korea. The number of child abuse reports filed nearly tripled from 2012 to 2016, with 6,403 child abuse cases reported in 2012, and 18,573 cases reported in 2016. Just as these national statistics demonstrate the increasing prevalence of child abuse, news outlets provide us with a growing body of alarming news reports of children who are in severe danger because of child abuse. Previous studies (e.g., Cicchetti & Rogosch, 2002) have demonstrated that maltreated children are at a high risk for maladaptive functioning and psychopathology across their life span because child abuse is associated with many detrimental effects on their biological, cognitive, social, emotional, representational, and linguistic development (Cicchetti & Lynch, 1995).

The present study used Masten’s resilience model (Masten & Reed, 2002), which provides a framework for understanding human beings and resilience from a contextual and systematic perspective. This model was chosen to identify resilient individuals from among those who reported having experienced child abuse because it is grounded in a person-focused approach that attempts to capture both the common and uncommon patterns of adaptation that naturally occur due to individual differences in the course of development. Person-focused approaches are also useful for identifying the change patterns that occur in the lives of real people over time due to multiple interactive processes and for identifying distinctive regularities and irregularities within these processes (Bergman & Magnusson, 1997). Masten’s model positions resilience as an inferential and contextual construct that requires two major kinds of judgment (Masten & Coatsworth, 1998; Masten et al., 1999). The two core components required in Masten’s resilience model are the presence of a demonstrable threat or risk and adaptation.

The first condition essential for one to be considered resilient is that there must be a significant, demonstrable risk that may pose a threat or have the potential to threaten one’s development. Similar to the approach employed in previous studies, the present study recognized child abuse as a demonstrable risk or threat to children’s development because parents, who are expected to be their children’s primary source of safety and security, can often be perpetrators of child abuse. In other words, the family unit, which has been found to have immediate and direct effects on child development, may pose a threat or act as an obstacle to one’s development. Child abuse, which violates the trust at the core of a child’s relationship with the world (Walker, 1994), creates an insecure attachment style that often leads to negative representations of the self and others (Muller, Sicoli, & Lemieux, 2000).

The second condition that is essential for one to be considered resilient is the quality of adaptation or developmental outcomes assessed as “good” or “OK” (Elder, 1998; Masten, 2001; Masten & Coatsworth, 1998; Waters & Sroufe, 1983). The present study adopted peer relationships and teacher–student relationships to measure the quality of adaptation because child abuse has been found to have negative associations with several stage-salient areas of development, including social difficulties and poor school performance (Aber & Cicchetti, 1984). Previous studies have shown an association between child maltreatment and interpersonal difficulties (Johnson et al., 2002), difficulties in peer relationships (Cicchetti & Lynch, 1993; Cicchetti & Toth, 1995), and peer rejection and avoidance behaviors in the classroom setting (Buhs, Ladd, & Herald, 2006). Within this framework, Masten’s model categorizes individuals into the following four groups: the maladaptive, resilient, competent-unchallenged, and vulnerable group. The first group is the Maladaptive. These individuals are characterized by the presence of adversity (exposure to child abuse—emotional and/or physical abuse) and maladaptive functioning (low school adaptability—peer relationships and teacher–student relationships). The second group is the Resilient. These individuals are characterized by the presence of adversity (exposure to child abuse—emotional and/or physical abuse) and functional adaptation (high school adaptability—peer relationships and teacher–student relationships). The third group is the Competent-Unchallenged. These individuals are characterized by absence of adversity (no exposure to child abuse—emotional and/or physical child abuse) and functional adaptation (high school adaptability—peer relationships and teacher–student relationships). The fourth group is the Vulnerable. These individuals are characterized by absence of adversity (no exposure to child abuse) and maladaptive functioning (low school adaptability—peer relationships and teacher–student relationships).

According to Bronfenbrenner (1986), human development can and does occur in several different settings, but majority of this development occurs in the family. He describes the family as a system that has an immediate and direct impact on child development. Similarly, the preamble to the United Nations Convention on the Rights of the Child (1989) describes the family as one of the foundational social units in which development takes place. In the early stages of development, children inhabit a world that revolves around their parents or primary caregivers. Parents and caregivers are the primary source of safety, security, love, understanding, nurturance, and support. Individuals are born into families, and they typically expect their families to provide a secure source of social support. Such support is a predictor of successful outcomes or adaptations over the life span. However, previous studies have found that many children grow up suffering from child abuse (Masten & Garmezy, 1985; Masten & Wright, 1998).

Second, the present study applied Bronfenbrenner’s ecological systems theory to identify the resilient group. In other words, the present study applied Bronfenbrenner’s ecological systems theory to understand human beings in a contextual, relational framework. Bronfenbrenner (1979) argued that every person develops in a unique environment—the system—that is distinct in its scope, content, and structure and that an individual needs to be understood within the interactive discourse or the context. This acknowledgment of inter- and intra-individual differences in systems enables researchers to understand resilience based on interactions across multiple domains and to examine core concerns about continuity and discontinuity. In other words, examining the self in relation to and in the context of different domains allows for a contextual understanding of resilience.

By integrating Bronfenbrenner’s ecological systems theory with Masten’s resilience model, resilient individuals can be described as those who have undergone the threat of disadvantage or adversity in one domain (such as the family or home) but have nevertheless managed to show functional adaptations in another domain (such as the school, or more specifically, in teacher–student or peer relationships). This study not only takes into account how a threat influences the self at a specific point in time but also how an experience in one domain (home) influences an individual when he or she is placed in another domain (school). It allows us to examine how an individual’s adverse circumstances in one domain (home) influence him or her in that domain, how risk influences the individual when placed in another domain (school), and how risk and adaptation interact to influence resilient individuals. Additionally, it enables us to track these changes at the latent class level over time.

Although only these two domains are considered in the present study, it importantly identifies resilient adaptive patterns among individuals who have experienced child abuse or those who have shown positive life adaptations despite the adversity they faced during childhood and adolescence. Therefore, we examined two research questions: (a) Among child abuse victims, can the resilient latent class be identified at each time point? and (b) How does each latent class change across time? To answer these questions, we examined whether resilient individuals could be identified among child abuse victims at each time point. We also examined the transition to resilient latent status. Finally, we considered how adverse circumstances in one domain (home) may have a discontinuous influence in another domain (school). The present results are especially meaningful because they suggest that psychological damage or failure in one domain does not necessarily lead to continuous maladaptive experiences in other domains. Many previous studies have proven that individuals manage to live adaptive lives despite adversity, attesting the presence of human resilience. Nevertheless, negative perceptions of individuals experiencing adversity as the vulnerable or the identification of survivors of adversity as outliers or as invulnerable—based on the prevalence of attempts to understand human beings facing adverse circumstances based on a linear, fixated perspective—continue to exist. Although the study has limitations in that it discusses only two among multiple domains, it reveals the presence of adaptive life patterns despite adverse circumstances, especially in terms of the ordinariness of resilience identified as a part of normative development. Furthermore, it provides practitioners, educators, and counselors insight into the most appropriate time for intervention to produce positive adaptations and offers suggestions to schools on ways to provide protective resources or assets to support individuals during the developmental process.

Method

Participants

The Korean Children and Youth Panel Survey (KCYPS) data set provided by the National Youth Policy Institute (2016) is a large, nationally representative, longitudinal data set available online for examining the growth and developmental process of Korean children and adolescents. The KCYPS data set comprised children in the 4th grade of elementary school in the beginning, who were followed up to the 10th grade of high school. A stratified sample of 2,378 students was recruited from 95 schools from 16 cities in South Korea. The present study used the KCYPS data set from 2011, 2013, and 2015 (namely, Waves 2, 4, and 6 of the original study; for convenience, hereafter, these will be referred to as Time 1, Time 2, and Time 3, respectively). In other words, the sample cohort group under the study included three different cohorts: elementary school fifth graders (Time 1), middle-school seventh graders (Time 2), high school ninth graders (Time 3), which the survey was conducted in the beginning of each year (National Youth Policy Institute, 2010–2016). While some variables were followed up on a yearly basis, the data used in the present study were collected and retrieved on two-year interval at the three time periods since emotional and physical abuse, and school adaptation ability were considered to be relatively stable and constant indicators.

The first time point (2011, when participants were in Grade 5, identified as late childhood) was chosen as the baseline to measure the change occurring during the transition from late childhood to mid-adolescence accurately. The second time point was meaningful in terms of normative development (2013, when the participants were in Grade 7, identified as the transitional period). This time point was chosen because Korean students undergo a transition from elementary school to middle school at this stage. Thus, they would not only experience a change in the school environment but also the transition from the late childhood to mid-adolescence so called puberty. The third time point (2015, when participants were in Grade 9, identified as mid-adolescence) was the identifiable and demonstrable period in adolescence before any changes would be influenced by environmental factors (e.g., school) due to transitions to high school.

Attrition and missing data

The data were first collected in 2010, with a stratified sample of 2,378 fourth-grade elementary school students. This initial sample included children who did not complete all three waves of the survey; 2,377 students completed Time 1 (4.8% of missing data), 2,264 completed Time 2 (7.6% of missing data), and 2,092 completed Time 3 (11.3% of missing data). The missing data were primarily due to two reasons, one being item nonresponse. Participants were given the survey, but some refused to respond to every single item. The other reason for the missing data was participant attrition due to death, illness, or refusal to continue to participate in the survey (see Table 1).

A common problem in panel surveys that may limit the analysis is that some respondents fail to respond at one or more panel waves, and such drop out of participants causes biases in all results that are influenced by these variables, resulting in an incomplete longitudinal record (Baltes, Reese, & Nesselroade, 1988; Little & Rubin, 2002). This incompleteness poses a problem that deletes observations with missing time points and analyzes only records with no missing values. In order to correct the potential problem, the missing data were managed using the Full Information Maximum Likelihood (FIML) procedure using Mplus 5.0 (L. K. Muthén & Muthén, 1998–2007). The total sample used after the FIML which uses the maximum likelihood estimators for latent class analysis (LCA) and latent transition analysis (LTA) was 1,932.

Sociodemographic characteristics

Approximately 52.5% of the participants were males and 47.5% were females. Participants were mainly from low- and mid-socioeconomic status (SES) groups, with 1,936 students (81.4%) in the lower third quartile range. The number of participants in the high-SES group was relatively small, with 385 students (16.2%), but the range of family income was largely dispersed from 6,001 to 50,000. As for family structure, most participants (79.7%) were raised by both parents, 69 (2.9%) were raised by the father only, 104 (4.4%) were raised the mother only, and 15 (0.6%) were raised by a step parent.

Measures

Adversity: Child abuse

To measure participants’ exposure to adversity, child abuse—physical and emotional abuse—was examined using a revised version of the Child Abuse Inventory (CAI; Huh, 2000; S. W. Kim, 2003), which is a four-item self-report scale. The physical abuse variable comprised the following two items: “If I do something wrong, my parents show extreme behaviors” and “My parents were often harsh on me, and I often had wounds or bruises from it.” The emotional abuse variable comprised the following two items: “If I make a mistake, my parents always punish me” and “My parents often scold, swear, or yell excessively at me calling me stupid, worthless, dumb, and finally, my least favorite, a failure.” The scale uses a four-point Likert-type scale with response options ranging from 1 (very likely) to 4 (very unlikely). It is reverse scored, with higher scores indicating exposure to child abuse. Conceptually, a demonstrable distinction of the presence or absence of risk or threat needs to be made to distinguish a resilient individual. Thus, this continuous variable was transformed into a dichotomous variable. In other words, each item was recoded as a dichotomous variable represented by 0 or 1 (ratings of 1 and 2 were recoded as 0 = no, and those of 3 and 4 as “1” = yes). In short, for the clear identification of the resilient group, we assumed the responses recoded to “no” to indicate that the participants had identified themselves as having had no exposure to emotional or physical abuse, while “yes” indicated that they identified themselves as having had exposure to emotional or physical abuse. According to S. W. Kim (2003), child abuse, as measured by the CAI, is significantly associated with withdrawal, depression/anxiety, delinquency, and aggression. Furthermore, in two longitudinal studies that measure child abuse using the CAI, child abuse is significantly associated with adolescent depression and delinquency and parental neglect (H. G. Kim, 2014; Son & Kim, 2017).

Adaptation: School adaptation

To measure one’s functional adaptation, participants’ school adaptability was measured using two variables, peer relationships and teacher–student relationships. School adaptation was measured using a revised version of the School Life Adaptability Inventory (Min, 1991), which is a 14-item self-report scale. The peer relationships variable was assessed using five items; for example, “I get along well with my classmates” and “I often disturb what my friends are doing.” The teacher–student relationships variable was assessed using four items; for example, “I say hello to my teacher” and “I feel comfortable talking to my teacher.” The questionnaire was scored on a four-point Likert-type scale, with response options ranging from 1 (never) to 4 (always). Several items were reverse scored such that higher scores indicated better school adaptability. Furthermore, each item on peer relationships was recoded as a dichotomous variable represented by 0 or 1 (ratings of 1 to 2 were recoded as “0” and those of 3 to 4 were recoded as “1”). Similarly, each item on teacher–student relationships was recoded as a dichotomous variable represented by 0 or 1 (ratings of 1 to 2 were recoded as “0” and those of 3 to 4 were recoded as “1”). For the clear identification of school adaptation, these recoded responses were labeled as “low” or “high” to indicate low and moderate to high levels of school adaptability, respectively.

Analysis

After the data screening process, descriptive statistics were computed using SPSS 21.0. Subsequently, LCA was applied to identify the class of resilient participants separately at each time point (Time 1, Time 2, and Time 3), and LTA was used to identify the transition of each subgroup from the previous time point to the next time point (i.e. Time 1 to Time 2, Time 2 to Time 3). The main objective of LCA is to identify the latent classes (the unobserved hypothesized categorical variables) in a population and to group individuals into homogeneous subgroups that are characterized by different response patterns within a sample that is represented by a few (imperfectly) measured observed indicators (Collins & Lanza, 2013). LCA is an appropriate method for the present study since it assumes a person-oriented approach (Magnusson & Torestad, 1993), through which it emphasizes the identification of a subsystem or the patterns relevant to the problem under study (Bergman & Magnusson, 1997; Bergman & Trost, 2006). In other words, LCA, like k-cluster analysis, can be used to identify clusters of a population under investigation (Burns & Burns, 2008). To examine transitions in the latent class across time, LTA was applied. In LTA, the transition probabilities of each status (class) at time T + 1 from those at time T are estimated. In this study, the latent classes were fixed across time as a result of LCA, which showed similar item-response probabilities across the three time points (B. O. Muthén & Asparouhov, 2011). Thus, each status (class) had the same characteristics (e.g., maladaptive or resilient) across the time points. After identifying the subgroups (classes or statuses) at each time point using LCA, we were able to determine the number of participants who moved into other subgroups or stayed in the same subgroup at the next time point by estimating their transition probabilities using LTA. LCA and LTA were conducted using Mplus 5.0 (L. K. Muthén & Muthén, 1998–2007). The LCA or LTA uses a combination of statistical information criteria, like Akaike’s Information Criterion (AIC; Akaike, 1987) and Bayesian Information Criterion (BIC; Schwartz, 1978) for simplification, and approximations are derived based on the most appropriate number of classes, which could range from a two- to five-class model. In general, according to the function of the models, those with lower AICs or BICs are thought to be more accurate, suggesting that lower AICs and BICs estimate a better model.

Results

The proportion of participants who reported having had exposure to adversity—emotional and/or physical abuse—decreased over time. Changes in the frequency of exposure to physical abuse remained relatively stable over time considering the attrition rate (448, 408, and 408 at Time 1, Time 2, and Time 3, respectively), in comparison to the variability in the frequency of emotional abuse found over time (865, 719, and 498 at Time 1, Time 2, and Time 3, respectively). As for school adaptation—peer relationships and teacher–student relationships—some variations were identified across time. Peer relationships (N = 1,369, 59.3% at Time 1; N = 1,450, 69.4% at Time 2; and N = 1,533, 74.5% at Time 3) showed an identifiable increasing trend over time, while there were some variations within teacher–student relationships. While students maintained good relationships with teachers at Time 1, there was a decrease in scores at Time 2, but a rebound at Time 3 (N = 1,834, 79.5% at Time 1; N = 1,527, 73.1% at Time 2; and N = 1,563, 75.9% at Time 3) (see Table 1).

LCA (measurement model)

The AIC and BIC were the two major indicators used to identify the measurement model. Both indicators were the lowest at Time 1, supporting the four-class model as the best solution. Therefore, the four-class model was selected for Time 1. At Time 2, while the BIC (5900.66) was the lowest, supporting the three-class model, the AIC (5807.25) supported the four-class model to be a better solution. At Time 3, the BIC (5582.30) was the lowest, supporting the three-class model. However, the AIC (5491.66) supported the four-class model to be a better solution. While the nonconvergence in the model fit in both Time 2 and Time 3 with the two model fit indices suggesting different subsets to be ideal, a four-class model was selected to testify the theoretical model, which comprises four different classes (Nylund, Asparouhov, & Muthén, 2007).

Latent transition analysis

In LTA, sometimes fitting a series of latent class models with different numbers of latent classes to data at a single time point can be helpful as a preliminary step in model selection. Doing this at each time point can be informative about the latent structure within times and how that structure changes across time points. Figure 1 displays the latent classes at each time point, which were quite similar across time.

Figure 1.

Item-response probabilities measurement for each latent class from Time 1 to Time 3.

Table 2 displays the proportion of students in each latent status in each time point. At Time 1, the competent-unchallenged latent status was the most prevalent (42.5%), followed closely by the resilient (22.5%) and vulnerable (19.5%) latent statuses. The maladaptive latent status was the least prevalent (15.5%). At Time 2, the overall pattern was similar, except for the prevalence of each latent status. As in Time 1, the competent-unchallenged (46.9%) latent status was the most prevalent, followed by the vulnerable (23.4%) and resilient (22.3%) latent statuses. The maladaptive latent status (7.4%) remained the least prevalent. At Time 3, the overall pattern was similar to that observed for Time 2, except for the prevalence of each latent status. Specifically, the prevalence of the resilient (22.0%) and vulnerable (21.1%) latent statuses declined from Time 2 to Time 3, while that of the competent-unchallenged (49.2%) and maladaptive (7.7%) latent statuses increased.

Table 1.

Descriptive statistics of the indicators.

	Time 1		Time 2		Time 3
	N	n (%)	N	n (%)	N	n (%)
Emotional abuse	2,264	865 (38.2)	2,092	719 (34.4)	2,061	498 (24.2)
Physical abuse	2,264	448 (19.8)	2,195	408 (18.6)	2,195	408 (18.6)
Peer relationships	2,308	1,369 (59.3)	2,090	1,450 (69.4)	2,058	1,533 (74.5)
Teacher relationships	2,308	1,834 (79.5)	2,090	1,527 (73.1)	2,058	1,563 (75.9)

Note: n = number of participants who were coded as experiencing child abuse or highly adaptive to school environment in terms of peer–teacher relationship.

Table 2.

Adaptive pattern prevalence in Time 1 to Time 3 from the latent transition analysis.

	Time 1	Time 2	Time 3
Latent status	N (%)	N (%)	N (%)
Maladaptive	300 (15.5)	143 (7.4)	148 (7.7)
Resilient	434 (22.5)	430 (22.3)	425 (22.0)
Competent-unchallenged	821 (42.5)	907 (46.9)	951 (49.2)
Vulnerable	377 (19.5)	452 (23.4)	408 (21.1)
Total	1,932 (100)	1,932 (100)	1,932 (100)

Overall, the competent-unchallenged latent status maintained the highest prevalence and increased slightly over time (N = 821, 42.5% at Time 1; N = 907, 46.9% at Time 2; and N = 951, 49.2% at Time 3). Furthermore, though the prevalence of the maladaptive status was the lowest at the beginning, there was a notable decrease at Time 2, which indicated transitional movements (N = 300, 15.5% at Time 1; N = 143, 7.4% at Time 2; and N = 148, 7.7% at Time 3). Meanwhile, the prevalence of the resilient status remained relatively stable from Time 1 to Time 3 (N = 434, 22.5% at Time 1; N = 430, 22.3% at Time 2; and N = 425, 22.0% at Time 3), while that of the vulnerable status increased from Time 1 to Time 2, and it decreased from Time 2 to Time 3 (N = 377, 19.5% at Time 1; N = 452, 23.4% at Time 2; and N = 408, 21.1% at Time 3). The estimated transition mean scores based on the model from Time 1 to Time 2 and from Time 2 to Time 3 have been presented in Table 3. The values along the diagonals of each matrix are indicative of stability in the specific groups, and those off the diagonals reflect movements from one class to another. This table presents several findings pertaining to the stability and change in profiles.

Table 3.

Transition probabilities of class changes over time.

	Time 2 (Time 3)
	Maladaptive	Resilient	Competent-unchallenged	Vulnerable
Time 1 (Time 2)
Maladaptive	0.22 (0.56)	0.23 (0.44)	0.15 (0)	0.41 (0)
Resilient	0.06 (0.12)	0.40 (0.88)	0.40 (0)	0.14 (0)
Competent-unchallenged	0.03 (0)	0.17 (0)	0.66 (0.94)	0.15 (0.06)
Vulnerable	0.18 (0)	0.10 (0)	0.22 (0.24)	0.50 (0.76)

Note: Values are the transition probabilities of class membership at Time 2 conditional on class membership at Time 1. Values in parenthesis are transition probabilities from Time 2 to Time 3. Shaded cells indicate stability probabilities. Bold face values are the highest probabilities of class membership at a subsequent time point, conditional on class membership at the previous time point.

Transition from elementary to middle school (Time 1 to Time 2)

Although some patterns remained unchanged, in general, the transitional period from elementary to middle school was most likely to have most transitions or variations. The highest stability probability was identified in the competent-unchallenged pattern (.66). The maladaptive (.22), resilient (.40), and vulnerable (.50) patterns were likely to show the same pattern at Time 2. However, interestingly, there were both positive and negative class changes identified during this period. Specifically, those who were identified as resilient at Time 1 were likely to change into competent-unchallenged at Time 2 (.40). Another interesting pattern identified was the probability of transition from the maladaptive status at Time 1 to the vulnerable status at Time 2 (.41).

From the transitional period to mid-adolescence (Time 2 to Time 3)

The patterns of transition probabilities of the class changes found from the transitional period to mid-adolescence were most likely to remain unchanged. These stable patterns were different from the earlier findings on the transition from Time 1 to Time 2. Specifically, individuals in the maladaptive (.56), resilient (.88), competent-unchallenged (.94), and vulnerable (.76) patterns in the transitional period (Time 2) were most likely to remain in the same pattern in mid-adolescence (Time 3). Unlike the changes identified in the transition from Time 1 to Time 2, a few pattern changes were identified in the transition from Time 2 to Time 3. While those who were maladaptive were likely to change into the resilient pattern (.44), those who were resilient were less likely to change into the maladaptive pattern (.12). Another contrasting pattern was found. While those who were vulnerable were likely to change into the competent-unchallenged pattern (.24), those who were competent-unchallenged were least likely to change into the vulnerable pattern (.06). Overall, while some transitional movements were found over time, the participants showed a greater tendency to remain stable.

Discussion

The three major purposes of this study were (1) to identify the resilient latent class or pattern, (2) to identify changes in the patterns—to follow the long-term resilience trajectory of child abuse victims using the KCYPS data set, and (3) to identify whether risk or threat in one domain (home) leads to continuous maladaptive ways of living in another domain (school). To accomplish these purposes, LCA and LTA were adopted. The salient findings were as follows.

First, to identify the resilient latent class using Masten’s resilience model, LCA was conducted to determine resilient patterns at each time point. Based on the results of the LCA, the following four types of latent classes were identified: resilient, maladaptive, vulnerable, and competent-unchallenged. As suggested by Masten’s resilience model (Masten & Reed, 2002; Masten & Tellegen, 2012), participants were classified into four different patterns. Thus, the present results are consistent with Masten’s resilience model (Masten & Reed, 2002). Second, LTA was used to identify the changes in patterns and development of the four groups at three time points (2011, 2013, and 2015). The transition probabilities from Time 1 to Time 2 were relatively unstable, albeit with many variations observed among latent classes. The transition probabilities from Time 2 to Time 3 remained relatively stable.

Transition from elementary to middle school (Time 1 to Time 2)

While most members of the competent-unchallenged group were likely to remain stable, those in the resilient, maladaptive, and vulnerable groups at Time 1 were more likely to make transitions to other groups at Time 2. Such variations in transitions indicate the fluctuation and instability that individuals experience during this developmental stage. The transition from elementary to middle school brings not only environmental changes but also salient changes in one’s role. Becoming a middle-school student, a salient role change, brings about a contextual change that involves specific expectations that are associated with specific positions in the society.

Among many variations, one of the most intriguing transitional movements identified was that of the probability of the maladaptive group at Time 1 transitioning to the vulnerable group at Time 2. One of the reasons behind this phenomenon may be the absence of child abuse at Time 2. According to Bronfenbrenner’s (1979) ecological systems theory, child development happens within the context of the system of relationships. The parent–child relationship is one in which the child is in his or her immediate surroundings and has direct interaction with the parents. This may constitute a bi-directional influence. In other words, a child’s parents may affect the child’s beliefs and behavior; however, the child also affects the behavior and beliefs of the parents. This implies that child development not only brings about changes in the perceptions, attitudes, and behaviors of children who have grown up into middle-school students, but that it also leads to changes in the surrounding individuals. This is especially true in the case of parents, whose behaviors and relationship with their children change (Berk, 2000). Thus, children who were once abused emotionally or physically may no longer be abused. Another reason behind such transition probabilities may be that salient role changes are stressful events that override or pose a greater threat or risk than the child abuse itself does. Abuse usually happens during childhood, and if it continues beyond childhood, it is considered a prolonged stressor that may not be as stressful or disturbing as school adaptation, which is one of the developmental tasks that is expected or required to be accomplished at a given period. In other words, those who were identified as maladaptive at Time 1 may still experience the same stressful parent–child relationship. However, a new stressful circumstance may make the individual less reactive to the chronic stressor but more reactive to the new stressor, in this case, school adaptation (Lazarus, 1966).

Another intriguing transitional movement identified was the probability of the resilient group at Time 1 to transition to the competent-unchallenged group at Time 2. One of the reasons for such a phenomenon may be the absence of child abuse at Time 2, as mentioned earlier (Berk, 2000). Another reason may be that an individual has become sufficiently resilient; in other words, the individual has developed his or her own support systems or patterns of regulation, stress reactivity, recovery, and coping strategies to handle the stressor (i.e., a stimulus or threat that disturbs his or her developmental pathway). According to Lazarus (1966), stress arises when individuals perceive that they cannot adequately cope with the demands being made on them or with threats to their well-being. Having experienced child abuse at a young age, an individual may develop his or her own coping strategy to adequately adapt to the circumstances (Knudsen, Heckman, Cameron, & Shonkoff, 2006; Shonkoff, 2006). Therefore, child abuse is no longer a significant stressor. Being able to cope with the situation, the individual may no longer identify as a child abuse victim. Finally, adolescence is a time to develop healthy and high quality relationships. Early adolescents devote a great amount of time and energy to relationships with peers than with parents. The peer group serves as a primary source of support, and contributes in important ways to adolescents’ self-concept and well-being. (Furman & Buhrmester, 1992; Prinstein, Boergers, & Vernberg, 2001; Youniss & Smollar, 1989).

Thus, it is possible that teacher–student or peer relationships—interpersonal relationships that individuals report as being healthy—may moderate or provide the assets or resources, an avenue for recovery for an individual to become competent in later developmental stages (Time 2).

From the transitional period to mid-adolescence (Time 2 to Time 3)

The transition probabilities of class changes from the transitional period to mid-adolescence showed stable patterns as compared to the change patterns identified during the transition from elementary to middle school. Most of the individuals were likely to remain stable and in the same classification as they were at Time 2. Moreover, minimal variations in transitions were identified. That is, other than the four identifiable patterns, no transitional movements were identified from Time 2 to Time 3. Among these four variations, one of the most intriguing transitional movements identified was the probability of moving from the maladaptive group at Time 2 to the resilient group at Time 3. This movement is important because it implies that, even if there was a threat at one point in time, there is probably room for a rebound. Furthermore, it implies that a threat in one domain does not necessarily lead to continuous maladaptive ways of living in other domains of one’s life. Adolescence is a period in which one’s peer group and peer relationships become especially important. Such transitional movement takes into consideration the developmental stages and implies that, though child abuse may continue, good peer relationships (or teacher–student relationships) may serve as secure boundaries so that enable an individual to make functional adaptations in the school setting and probably transition into the resilient group (Larson, Richards, Moneta, Holmbeck, & Duckett, 1996; Youniss & Smollar, 1989). Another possible factor in such transitional movements is that identity formation is one of the most salient developmental tasks in adolescence. It is often observed that adolescents have the tendency to detach from their parents to develop independence through the process of individuation and distancing. Thus, while the individual may continue to recognize child abuse as a risk or threat, his or her tendency to detach or create distance in relation to parents may render the abuse less meaningful or painful (Larson et al., 1996; Youniss & Smollar, 1989). Meanwhile, what happens in school, especially in terms of peer relationships, may become more important and meaningful in the process of self-definition.

Implications

The results of the present study have several implications. First, this study used Masten’s resilience model to identify patterns and changes in them among child abuse victims in Korea. Instead of using different instrumentation, the study replicated Masten’s resilience model using the KCYPS data set. The four latent classes in Masten’s resilience model—the resilient, maladaptive, vulnerable, and competent-unchallenged—were identified. This supports the validity of Masten’s resilience model (Masten & Reed, 2002). Second, the presence of the resilient group at each time point emphasizes the ordinariness of resilience, or the idea that resilience is a common phenomenon that results from the functioning of basic human operational systems. As the present findings illustrate, although there are times when individuals face adversity, if one’s basic human operational systems are well functioning, the individual tends to make positive movements and adaptations, which is the ordinary nature of human beings. Furthermore, consistent with earlier findings, the present observation of the resilient pattern at each time point implies that risk in one specific domain may not necessarily lead to poor functioning in another domain of life. In other words, it demonstrates that psychological damage or failure in one domain does not necessarily lead to continuous maladaptive ways of living in other domains over time. This supports the assumption underlying resilience, which is that human beings can maintain healthy, functional, and adaptive ways of living despite adverse life events.

Third, the change in patterns across time demonstrates that resilience is a dynamic process (Luthar, Cicchetti, & Becker, 2000) that is dependent on interactions among various factors surrounding the individual (Dyer & McGuinness, 1996). This implies the need to understand resilience in relation to an individual’s context. Fourth, unlike the changes or variations that were identified from Time 1 to Time 2, only a few transition pattern changes were identified from Time 2 to Time 3. Adolescence is a period of fluctuation during which both biological and ontological changes occur. Such a demonstration of change in patterns suggests that the transition from elementary to middle school is a more appropriate time for intervention than is the period between Time 2 and Time 3.

Finally, as Bronfenbrenner suggested, “Every child needs at least one adult who is irrationally crazy about him or her” (Brendtro, 2006, p. 165). Human development is linked to multiple contexts. Key contexts for adolescents include family, peers, school, and community (Hill, 1983). Developmental contextualism suggests that human development is embedded in the given social contexts and that it occurs in a reciprocal relationship between adolescents and these contexts (Lerner, 1978, 1979, 1991). While many studies have suggested that poor parent–child relationships are strongly associated with negative interpersonal relationships in other domains, the findings from the present study demonstrate that the insecure attachment that results from poor parent–child relationships may be improved through good peer and teacher–student relationships. In other words, adolescents are situated in relation to multiple, changing contexts, and within such continuously changing interactive and social contexts, adolescents are able to develop as individuals with unique qualities (Lerner, 1979, 1982, 1991). School practitioners—including teachers and counselors—should implement policies that place greater emphasis on peer engagement and fostering the teacher–student relationship so that there are protective resources and assets that can support and shape the individual during the developmental process (Lerner, 1979).

Limitations and recommendations for further study

This study was designed as a longitudinal analysis so that it would be possible to detect patterns of change between and across child abuse victims. Cicchetti and Garmezy (1993) noted that a cross-sectional study can examine the resilience process only at a certain point in time. Thus, they suggested that a longitudinal study would be a better means of identifying the nature of resilience. This study complements previous cross-sectional studies.

However, because all of the instruments in this study were self-reported surveys, the study may reflect several biases. One of the possible reasons for bias is the school-related measures assessed in the school setting. Students participating in the survey may have experienced the dilemma of social desirability (Krumpal, 2013); in other words, they may have felt pressure to answer in ways that conform to social norms and that fulfill the expectations of teachers, peers, and the school. Additionally, individuals may also not have been willing to identify themselves as child abuse victims. Because child abuse is a personal indicator that explains the bonding in the parent–child relationship, which is directly related to the core components of self-definition, one may decide not to disclose personal information, though he or she may continue to experience of abuse. As one undergoes the transition into mid-adolescence, during which one of the salient developmental tasks is identity formation, one may decide not to disclose family issues. One may develop the intuitive knowledge that the experience of having been exposed to child abuse is not considered valuable and may even pose a secondary threat; that of a stereotype or an unwanted, misrepresentation of the self. In such a case, one may disguise his or her actual experiences to avoid experiencing humiliation and stereotyping that often come with the failure to conform to social norms—what is considered to be “good.” Thus, it is likely that extreme cases of child abuse are not represented in the study. Even though we measured child abuse as one indicator of the parent/child relationship, we did not measure other indicators of the parent/child relationship (e.g., attachment). Therefore, future research should include those indicators to control for the nuisance effects of the results.

Moreover, while changes in patterns were observed in the present study, the study was limited in its ability to identify the direct causal relations explaining this phenomenon. One of the related limitations is the threat of chronology bias (Feinstein, 1977). In other words, the data set did not provide sufficient information to distinguish between or control the varying lengths of time and exposure to maltreatment or to determine whether the child abuse is an acute or prolonged chronic stressor. It also did not measure the acuteness—the varying degrees of severity of the risk. Moreover, it was limited in that it was unable to capture what happened beyond the time of measurement. According to Freud, early childhood experiences are important in the development of the self. Taking this Freudian perspective into account, it seems that the specific period during which child abuse occurred in an individual’s lifetime—whether in early childhood or in mid-adolescence—may exert a unique influence on the person’s development.

Finally, while the present study aimed to identify the adaptive patterns of individuals using Masten’s resilience model and Bronfenbrenner’s ecological systems theory, different contexts are likely to produce different conceptualizations of resilience (Southwick, Bonanno, Masten, Panter-Brick, & Yehuda, 2014). While the present study attempts to employ a contextual, relational framework, it remains an inferential study whose results may not accord with reality.

This study restricted its investigation of risk and adaptation to two domains, excluding protective factors. Researchers have argued that risks of specific and general problems in development usually co-occur and that the accumulation of these risks at one point in time or over time is strongly related to an increased risk of poor outcomes on multiple indicators of development (Masten & Wright, 1998; Rutter, 1979; Seifer & Sameroff, 1987). Future resilience studies might consider measuring resilience across multiple domains using Masten’s resilience model, thereby shedding light on the interaction between risks and the quality of adaptations despite the co-occurrence of multiple risks.

Footnotes

Acknowledgments

The authors thank the National Youth Policy Institute for conducting the Korean Children and Youth Panel Survey and granting permission for free access to the large, nationally representative, longitudinal database. Furthermore, the authors would like to thank the participants of the survey who have contributed to this study ( for a more comprehensive description of the original data).

Article Notes

References

Aber

J. L.

Cicchetti

(1984). Socioemotional development in maltreated children: An empirical and theoretical analysis. In Fitzgerald

H. E.

Lester

Yogman

(Eds.), Theory and research in behavioral pediatrics ( Vol. 2, pp. 147–205). New York, NY: Plenum Press.

Akaike

(1987). Factor analysis and AIC. Psychometrika, 52, 317–332.

Baltes

P. B.

Reese

H. W.

Nesselroade

J. R.

(1988). Introduction to research methods: Lifespan developmental psychology. Hillsdale, NJ: Lawrence Erlbaum.

Bergman

L. R.

Magnusson

(1997). A person-oriented approach in research on developmental psychopathology. Development and Psychopathology, 9, 291–319.

Bergman

L. R.

Trost

(2006). The person-oriented versus the variable-oriented approach: Are they complementary, opposites, or exploring different worlds? Merrill-Palmer Quarterly, 52(3), 10.

Berk

L. E.

(2000). Child development (5th ed.). Boston, MA: Allyn and Bacon.

Brendtro

(2006). The vision of Urie Bronfenbrenner: Adults who are crazy about kids. Reclaiming Children and Youth, 15(3), 162–166.

Bronfenbrenner

(1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.

Bronfenbrenner

(1986). Ecology of the family as a context for human development: Research perspectives. Developmental Psychology, 22(6), 723–742.

10.

Buhs

E. S.

Ladd

G. W.

Herald

S. L.

(2006). Peer exclusion and victimization: Processes that mediate the relation between peer group rejection and children’s classroom engagement and achievement? Journal of Educational Psychology, 98(1), 1–13.

11.

Burns

R. B.

Burns

R. A.

(2008). Business research methods and statistics using SPSS. London, England: SAGE.

12.

Cicchetti, D., & Garmezy, N. (1993). Prospects and promises in the study of resilience. Development and Psychopathology, 5, 497–502.

13.

Cicchetti

Lynch

(1993). Toward an ecological/transactional model of community violence and child maltreatment: Consequences for children’s development. Psychiatry: Interpersonal and Biological Processes, 56(1), 96–118.

14.

Cicchetti

Lynch

(1995). Failures in the expectable environment and their impact on individual development: The case of child maltreatment: Risk, disorder, and adaptation. In Cicchetti

Cohen

D. J.

(Eds.), Developmental psychopathology: Risk, disorder, and adaptation ( Vol. 2, pp. 32–71). New York, NY: Springer.

15.

Cicchetti

Rogosch

F. A.

(2002). A developmental psychopathology perspective on adolescence. Journal of Consulting and Clinical Psychology, 70(1), 6–20.

16.

Cicchetti

Toth

S. L.

(1995). Child maltreatment and attachment organization: Implications for intervention: Social, developmental, and clinical perspectives. In Goldberg

Muir

Kerr

(Eds.), Attachment theory: Social, developmental, and clinical perspectives (pp. 279–308). Hillsdale, NJ: The Analytic Press.

17.

Collins

Lanza

(2013). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. New York, NY: Wiley.

18.

Dyer

J. G.

McGuinness

T. M.

(1996). Resilience: Analysis of the concept. Archives of Psychiatric Nursing, 5, 276–282.

19.

Elder

G. H.

(1998). The life course as developmental theory. Child Development, 69, 1–12.

20.

Feinstein

A. R.

(1977). Clinical biostatistics. St. Louis, MO: Mosby.

21.

Furman

Buhrmester

(1992). Age and sex differences in perceptions of networks of personal relationships. Child Development, 63, 103–115.

22.

Hill

J. P.

(1983). Early adolescence: A research agenda. Journal of Early Adolescence, 3, 1–21.

23.

Huh

M. Y.

(2000). The study for the development and validation of parenting behavior inventory perceived by adolescent (Doctoral dissertation). Ewha Womans University, Seoul, Korea.

24.

Johnson

J. G.

Cohen

Gould

M. S.

Kasen

Brown

Brook

J. S.

(2002). Childhood adversities, interpersonal difficulties, and risk for suicide attempts during late adolescence and early adulthood. Archives of General Psychiatry, 59, 741–749.

25.

Kim

H. G.

(2014). The effect of perceived parental abuse and peer attachment on school life adjustment according to children’s gender. Family and Environment Research, 52(1), 11–19.

26.

Kim

S. W.

(2003). The effect of social support on abused children’s adjustment (Doctoral dissertation). Seoul National University, Seoul, Korea.

27.

Knudsen

E. I.

Heckman

J. J.

Cameron

Shonkoff

J. P.

(2006). Economic, neurobiological, and behavioral perspectives on building America’s future workforce. Proceedings of the National Academy of Sciences of the United States of America, 103(27), 10155–10162.

28.

Krumpal

(2013). Determinants of social desirability bias in sensitive surveys: A literature review. Quality & Quantity, 47(4), 2025–2047.

29.

Larson

R. W.

Richards

M. H.

Moneta

Holmbeck

Duckett

(1996). Changes in adolescents’ daily interactions with their families from ages 10 to 18: Disengagement and transformation. Developmental Psychology, 32(4), 744–754.

30.

Lazarus

R. S.

(1966). Psychological stress and the coping process. New York, NY: McGraw-Hill.

31.

Lerner

R. M.

(1978). Nature, nurture, and dynamic interactionism. Human Development, 21, 1–20.

32.

Lerner

R. M.

(1979). A dynamic interactional concept of individual and social relationship development. In Burgess

R. L.

Huston

T. L.

(Eds.), Social exchange in developing relationships (pp. 271–305). New York, NY: Academic Press.

33.

Lerner

R. M.

(1982). Children and adolescents as producers of their own development. Developmental Review, 2, 342–370.

34.

Lerner

R. M.

(1991). Changing organism-context relations as the basic process of development: A developmental contextual perspective. Developmental Psychology, 27, 27–32.

35.

Little

R. J. A.

Rubin

D. B.

(2002). Statistical analysis with missing data (2nd ed.). New York, NY: Wiley.

36.

Luthar

S. S.

Cicchetti

Becker

(2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71, 543–562.

37.

Magnusson

Torestad

(1993). A holistic view of personality: A model revisited. Annual Review of Psychology, 44, 427–452.

38.

Masten

A. S.

(2001). Ordinary magic: Resilience processes in development. American Psychological Association, 56, 227–238.

39.

Masten

A. S.

Coatsworth

J. D.

(1998). The development of competence in favorable and unfavorable environments: Lessons from research on successful children. American Psychologist, 53, 205–220.

40.

Masten

A. S.

Garmezy

(1985). Risk, vulnerability, and protective factors in developmental psychopathology. In Lahey

B. B.

Kazdin

A. E.

(Eds.), Advances in clinical child psychology ( Vol. 8, pp. 1–52). New York, NY: Plenum Press.

41.

Masten

A. S.

Hubbard

J. J.

Gest

S. D.

Tellegen

Garmezy

Ramirez

(1999). Competence in the context of adversity: Pathways to resilience and maladaptation from childhood to late adolescence. Development and Psychopathology, 11, 143–169.

42.

Masten

A. S.

Reed

M. J.

(2002). Resilience in development. In Snyder

C. R.

Lopez

S. J.

(Eds.), Handbook of positive psychology (pp. 117–131). New York, NY: Oxford University Press.

43.

Masten

A. S.

Tellegen

(2012). Resilience in developmental psychopathology: Contributions of the project competence longitudinal study. Development and Psychopathology, 24, 345–361.

44.

Masten

A. S.

Wright

M. O. D.

(1998). Cumulative risk and protection models of child maltreatment. Journal of Aggression, Maltreatment and Trauma, 2(1), 7–30.

45.

Min

B. S.

(1991). The influence of school life and adjustment and self-concept on academic achievement (Master’s thesis). Hongik University, Seoul, Korea.

46.

Ministry of Health and Welfare. (2016). Child abuse & neglect Korea 2016. Retrieved from http://www.korea1391.go.kr/

47.

Muller

R. T.

Sicoli

L. A.

Lemieux

K. E.

(2000). Relationship between attachment style and post traumatic stress symptomatology among adults who report the experience of childhood abuse. Journal of Traumatic Stress, 13, 321–332.

48.

Muthén

B. O.

Asparouhov

(2011). LTA in Mplus: Transition probabilities influenced by covariates. Mplus Web Notes: No. 13. Retrieved from http://www.statmodel.com

49.

Muthén

L. K.

Muthén

B. O.

(1998–2007). Mplus user’s guide fifth edition . Los Angeles, CA: Muthén & Muthén. Retrieved from http://www.statmodel.com

50.

National Youth Policy Institute. (2010–2016). Korea Children and Youth Panel Survey 1

–

7th survey of 4th grade [Data file and code book]. Retrieved from http://www.nypi.re.kr/

51.

National Youth Policy Institute. (2016). Korea Children and Youth Panel Survey 1

–

7th survey [User’s guide]. Retrieved from http://www.nypi.re.kr/

52.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569.

53.

Prinstein

M. J.

Boergers

Vernberg

E. M.

(2001). Overt and relational aggression in adolescents: Social-psychological adjustment of aggressors and victims. Journal of Clinical Child & Adolescent Psychology, 30, 479–491.

54.

Rutter

(1979). Protective factors in children’s responses to stress and disadvantage. In Kent

M. W.

Rolf

J. E.

(Eds.), Primary prevention of psychopathology: Vol. 3. Social competence in children (pp. 49–74). Hanover, NH: University Press of New England.

55.

Schwartz

(1978). Estimating the dimension of a model. The Annals of Statistics, 6(2), 461–464.

56.

Seifer

Sameroff

A. J.

(1987). Multiple determinants of risk and vulnerability. The Invulnerable Child, 45, 51–69.

57.

Shonkoff

J. P.

(2006). A promising opportunity for developmental and behavioral pediatrics at the interface of neuroscience, psychology, and social policy: Remarks on receiving the 2005 C. Anderson Aldrich Award. Pediatrics, 118, 2187–2191.

58.

Son

J. A.

Kim

S. H.

(2017). Longitudinal relationship between adolescent’s depression and delinquency: The effect of parents, peer and teacher factors. Korean Journal of Child Studies, 38(6), 1–15.

59.

Southwick

S. M.

Bonanno

G. A.

Masten

A. S.

Panter-Brick

Yehuda

(2014). Resilience deﬁnitions, theory, and challenges: Interdisciplinary perspectives. European Journal of Psychotraumatology, 5, 25338.

60.

United Nations Convention on the Rights of the Child. (1989). UN General Assembly. Retrieved from http://www.unicef.org/crc/

61.

Walker

A. M.

(1994). Normal early infant behavior patterns. Journal of Paediatrics and Child Health, 30(3), 33–46.

62.

Waters

Sroufe

L. A.

(1983). Social competence as a developmental construct. Developmental Review, 3(1), 79–97.

63.

Youniss

Smollar

(1989). Adolescents’ interpersonal relationships in social context. In Berndt

T. J.

Ladd

(Eds.), Peer relationships in child development (pp. 300–316). New York, NY: Wiley.