Introduction
Childhood adversity (CA) is an umbrella term for environmental circumstances or events that deviate from the “expected” normative/typical environment such as childhood maltreatment and poverty [
1‐
3]. Collectively, these events robustly predict later internalizing and externalising outcomes in youth. However, rather than to consider the individual and additive effects of these events on later psychopathology, researchers have recently begun to identify common variance amongst CAs on outcomes. Notably, events that reflect physical or emotional
threat to an individual have been distinguished from events that reflect
deprivation (e.g. through neglect and poverty) [
4,
5] on the basis of their differential associations with cognitive functions in community samples of adults as well as adolescents [
4,
6]. Whether threat and deprivation exhaustively reflect adverse childhood events and also show differential associations with behavioural and emotional outcomes in those at higher risk for psychopathology remain outstanding questions. The present study aims to address these questions by assessing the dimensionality of a broad spectrum of early experiences, and their associations with externalising and internalizing symptomatology, amongst a high-risk child and adolescent sample in Japan. A secondary aim is to also explore the cross-cultural validity of these existing findings to enhance their generalizability more globally.
A longstanding challenge to studying the effects of CAs on later psychopathology has been how to capture their combined versus unique effects on different developmental outcomes. While earlier studies tended to ignore the interplay between CAs by summing their effects on outcomes [
7‐
9], more recent approaches have applied latent trait models (e.g., factor analysis, latent class analysis, principal component analysis) to assess common and distinct variance between CAs [
4,
5,
10]. Such approaches have found that within parental maltreatment, abusive events and neglect seem to cluster separately [
4,
5], providing support for a recently proposed theoretical framework [
11], which suggests that
threat (representing a series of experiences that are of threat to one’s physical integrity) should be distinct from
deprivation (representing the absence of expected cognitive and social input).
Although this emerging approach has informed the multidimensional nature of early adverse experiences, a number of outstanding questions remain. First, although threat and deprivation reflect key dimensions within parental maltreatment, they may not capture all aspects of family malfunctioning. Within this broader category, other CAs, for example, parental criminality, parental substance use and parental violence, seem to co-occur, and may be distinct from parental abuse and/or neglect [
4,
5]. Second, it may be that there are distinct forms of deprivation such that events representing circumstantial deprivation (e.g., parental loss) are distinct from physical and/or emotional neglect. Identifying other distinct dimensions of CA and further sub-dividing existing dimensions could be better informed by assessing a wide range of adversities, such as those present in a high-risk sample who experience more extreme forms and more combinations of CAs [
12,
13]. So far, as most existing studies have drawn on general population samples, co-occurrence between CAs may be limited due to floor effects of total number of CAs experienced by participants.
A second outstanding question is that most of this research of the specificity of associations between dimensions of CAs and outcomes has largely focused on finding specificity within neurocognitive functional domains, such as fear processing [
14], and social cognitive abilities [
4]. Little is known about whether these distinct dimensions of adverse experiences translate to psychopathology. Studies assessing broad categories of CAs (e.g., childhood maltreatment, extreme forms of deprivation such as early institutionalization) have reported both common [
15‐
18] and specific [
19,
20] associations with internalising and externalising problems. This mixed picture could reflect difficulties in disentangling individual dimensions of adversity within broad CA categories. While less research had focused on internalizing difficulties, studies using natural experimental designs have provided valuable insights on specific associations with externalising problems. Data from the English Romanian Adoptee Study, for example, showed that patterns of disinhibited attachment, impaired cognitive abilities, hyperactivity, and quasi-autistic behaviours reflect a constellation of deprivation-specific psychological consequences [
21]. Other studies have also reported a high prevalence of hyperactivity and disinhibited social engagement amongst children who have experienced severe deprivation from institutional rearing [
21‐
23]. Conduct problems, on the other hand, often arise following the experience of abuse [
16] or interpersonal violence [
24]. These studies, while insightful, are still limited by the co-occurrence of adversities, e.g., institutional deprivation is likely to co-occur with peer victimization even when parental abuse is absent [
25,
26]. These co-occurring adversities could confound findings. Given these limitations, a latent trait model can potentially clarify these mixed results by identifying common variance across CAs. For example, Copeland and colleagues reported that although both suffered from poverty, children exposed to single parenthood and/or parental crime exhibited elevated disruptive behaviour, whereas children with parental poor education (at least one parent left school before 11th grade) were at higher risk of emotional disorders [
10].
It is also worth noting that there are comparatively few studies assessing the dimensional nature and their impact conducted in countries such as those from the Far East compared to the rich evidence-base from Europe and North America. Nonetheless, the limited evidence suggests country-based differences in responses to early-life adversities. For example, compared to rates of adversities reported from general population based adult samples in the USA [
27], Japan reported fewer adversities and moreover, the association with psychopathology was less generalized and did not persist beyond adolescence [
28]. In contrast, countries such as Mexico [
1] and China [
29] showed more comparable data to the USA. Extending such evidence of how these differential pathways translate into symptom types can inform the generalizability of existing findings to a global context.
The present study aimed to address these gaps, by assessing the key dimensions underlying multiple CAs, as well as their association with symptomatology within a high-risk sample. Using data from an institutionalized children and adolescent sample from Japan, we tested two hypotheses. First, we expected that using Principle Component Analysis, the threat-deprivation theme would emerge in our sample together with other potential components of parental malfunctioning, such as circumstantial deprivation. Second, these dimensions were predicted to hold distinctive associations with symptomatology: after controlling for covariates. More particularly, we expected that threat would uniquely associate with both externalising and internalizing symptomatology, and deprivation with externalising symptoms. We investigated these questions in a sample that included adolescents, given that this developmental period is relatively under-studied in terms of the impact of CAs on psychological functioning. Yet, studying the impact of CAs on psychopathology at this juncture is crucial given that many persistent psychiatric problems first onset during adolescence [
30,
31], and thus could be a sensitive peirod for intervention [
32]. Notably, a secondary aim of the study was to extend the cross-cultural validity of the CA research to Eastern cultures.
Discussion
The aim of the current study was to investigate the underlying construct of childhood adverse experiences, and their associations with various domains of psychological and behavioural symptoms in a high-risk youth sample from Japan. Our data revealed this group of Japanese high-risk youth all experienced at least 1 type of adversity, with the majority experiencing 3 or more types of adversity. Our results indicated three key findings: (1) using principal component analysis, there were four principal components (PCs) that optimally accounted for the correlation among 18 types of adverse experiences—parental abuse, parental psychosocial risks, parental absence, and parental neglect—explaining a total of 35.1% of the variance. Each PC accounted for very similar proportions of the variance in the data (9.6%, 7.7%, 7.2%, and 6.9%); (2) Both PCs typically considered as childhood maltreatment—parental abuse and parental neglect—showed similar (conduct problems) as well as differential (parental abuse with peer problems, parental neglect with hyperactivity) associations with symptomatology; (3) parental absence did not significantly associate with any outcome.
These results partially supported the existing literature in terms of dimensionality of early experiences of adversities: two clusters emerged separately within the same category of childhood maltreatment: parental abuse versus parental neglect. In addition, parental psychosocial risks emerged as an independent component amongst the broad category of CAs. The consistency of our findings, which are based on a Japanese high-risk sample, with previous ones mostly based in community samples in Western societies suggests that the nature of adverse events and its occurrence pattern are similar across cultures, even when there are differences in prevalence rate, age of participants, or source of information. Furthermore, our results also extend previous findings of subtle differences between events comprising deprivation: distinguishing parental
absence from parental
neglect. The two components also held different associations with outcomes—parental neglect was associated with hyperactivity/inattention and conduct problems, whereas parental absence did not significantly associate with any outcome. Some studies have found that when impoverished living conditions are not accompanied with insecure mother–child attachment, children did not show atypical stress responses unlike children who experienced highly adverse circumstances
and insecure attachment [
41]. Others have found that after adjusting for interpersonal violence, the association between poverty and stress reactivity was diminished [
6]. These findings suggest that deprivation maybe a two-pronged construct with one “prong” more closely linked with negative outcomes.
Our results from the path model also confirmed previous findings, which suggested that regardless of type and nature of the event, dimensions related to childhood maltreatment appeared to be most robustly associated with 3 out of 4 outcomes measured (i.e., hyperactivity/inattention, conduct problems, peer problems). Aside from the general effect of maltreatment, our results also suggested that parental abuse and parental neglect, when separated, hold common but also specific associations with outcomes. That is, while conduct problems was associated with both parental abuse and parental neglect, hyperactivity/inattention only significantly associated with parental neglect, and peer problems only significantly associated with parental abuse. This result further adds to previous findings where hyperactivity seems to be a behavioural consequence of severe deprivation [
42], and conduct problems is more influenced by both threat and deprivation [
2]. Unexpectedly, the emotional symptoms subscale did not significantly relate to any of the dimensions of adversity in our data. Several reasons can be considered. First, the emotional symptoms subscale moderately correlated with the other three SDQ subscales, and this interrelatedness was accounted for when estimating the association between the four principal components and dependent outcome variables in the path model. It is possible that externalising problems have stronger associations with adversity exposure relative to internalising problems. In fact, in the current study, emotional symptoms were reported by caregivers, not participants themselves. Indeed emotional symptoms are more elusive, and may not be as visible to external observers as externalizing problems such as conduct problems and hyperactivity. Previous studies have demonstrated weaker predictive power of informant-report compared to self-report for internalising symptoms [
43]. The lower ability to accurately report internalising problems as an external observer may explain in part the absence of observed associations between PCs and emotional symptoms. Second, it is worth noting that emotional symptoms did significantly associate with parental abuse (
r = .13,
p < .01) and parental neglect (
r = .10,
p < .05) in a simple two-tailed correlation analysis, and boys and girls showed similar correlation patterns (Supplemental Table 1). However after we adjusted for covariates and clustering, a relatively conservative approach, this association no longer remained significant.
Finally, the parental psychosocial risk component supported previous findings that events reflect parental malfunctioning [
4] and maladaptive family functioning [
5] tend to cluster together. However, unexpectedly, in our data, this PC negatively associated with conduct problems. To follow up on this unexpected negative association, we ran a simple correlation between the PC and Conduct Problem subscale (
r = − .16,
p < .01), as well as for each subscale item (
r = − .11 ~ − .19,
ps < .05) (Supplementary Table 1). It may be that with the effects of maltreatment (i.e., abuse, neglect) removed by the two PCs of parental abuse and parental neglect, parental psychosocial risks not accompanied with violence against children, is no longer harmful within this high-risk sample. It is noteworthy that a negative association does not necessarily mean that parental psychosocial risks are protective against negative outcomes, given that this association is derived from an extremely high-risk sample. Thus, this same negative association may not generalise to the population, where the frequency and severity of adversities differs.
While the current study findings have exciting implications both theoretically and clinically, there are several limitations that should be noted. The first set of limitations relate to how CAs were assessed retrospectively. Although the use of self-reports, informant (e.g., parent) reports, and/or case records to code for and rate CAs are considered the gold-standard methods for assessing early-life adversity, these methods nonetheless have the potential for bias and under-reporting. Furthermore, in the case of retrospective self-reports, these could yield recall bias resulting in over- or under-reported events. For example, a longitudinal investigation of childhood maltreatment [
44] demonstrated that self-reports of sexual and physical abuse are highly unstable over time. In the current study, due to ethical and safeguarding restrictions, we did not have permission to ask participants themselves about their previous experiences of the CAs, nor could we access their case records directly. Therefore, relying on the participant’s key caregivers to provide information for the assessment of CAs using a standardized official checklist and two project-developed checklists was the only viable option. Although reports from child care professionals are more likely to rely on judgments based on objective events (via case records) relative to self- or parental-report of adverse experiences, similar to case records, there may also be a tendency to under-report adverse experiences due to unawareness. Moreover, even though some information may be elicited through daily interactions with children and young people, some participants may be less communicative than others. Although there are concerns about the way we rated the CAs, we applied dichotomous ratings for each item in the checklist since our objective was to assess the overlap between different types of CAs. Moreover, it was essential to obtain information as accurate and consistent as possible across caregivers while minimising task demands on them. Information such as CA’s age of onset, duration, and intensity were not included in the current study due to challenges of getting this information from case records in a consistent manner across participants. Future studies assessing early adverse experiences should ideally involve multiple sources and informants [
45], and use more refined measures of a broad range of early adverse experiences, as well as include age of onset, or length of exposure to adversities.
Another limitation of the study is the range of CAs that were explored. Although, we determined the items through multiple participatory meetings by institution caregivers, social workers, and institution directors, and used the option of ‘other’ to prompt any adverse events not included in the checklists, there may still be events that were excluded, such as exposure to peer victimization, which was difficult to assess based on case records. We also did not measure adversity associated with institutional care, or the effects of institutionalization per se, given that all participants had experienced this. In addition, we also had not considered participants’ adaptation to institutional life, which has been consistently identified as a risk factor for developmental outcomes such as internalizing and externalising outcomes. Instead, we controlled for total time spent in care in the path analysis, and correlated children’s total time in care with four SDQ difficulties subscales. Interestingly, we found that the longer children were in care, the more elevated the level of externalising problems, namely hyperactivity/inattention and conduct problems. This finding supports the association previously reported by other researchers between deprivation and externalising problems. This result remained significant even after we removed children who had been in the current institutional care for less than 6 months. Although it is not common for children to leave residential care in Japan once placed—for example in 2015 6.9% (
N = 2735) of institutionalised children had either gone back home (
n = 2597; 6.6%), were adopted (
n = 24; .06%), or moved to foster homes (
n = 114; .2%) [
46]—it is important to note that this association does not infer directionality. It is possible that children with higher levels of externalising problems would be more likely to remain in residential care for longer compared to those with milder or no such symptoms. Without longitudinal data, it is difficult to disentangle these possibilities.
Finally, we only used caregiver reports on participants’ internalising and externalising outcomes as nearly 40% of the participants were under the age of 11 years old, for which a self-report version of the SDQ is unavailable. As such, for consistency, we relied on caregiver reports for all participants. However, previous data has found that parental reports of symptomatology, compared to self-reports, showed weaker associations with some symptom outcomes, especially with internalizing problems, which may be more difficult to detect by external raters [
43]. Another potential issue is same-rater biases given that the same individual rated both the adversity and the symptom outcomes. Future research in this area, while promoting measurement development in different cultural/language settings, should also aim to collect multi-rated data to ensure maximal validity of the construct. Moreover, in the current study, the institutionalized youth are closely supervised by institutional caregivers, hence, certain behavioural indices (e.g., suicide and self-harm, risky sexual behaviour) are relatively rare, and were not included in outcome assessments, despite their known association with early adversity. Furthermore, ethical and safeguarding issues prevented us from asking young people about suicidal ideation. Future research should consider including a broader range of outcome measures.
The current study, aimed to examine temporal associations between ‘distal’ adverse events in childhood with later on problem-behaviours, demonstrated that early adverse experience is a multi-dimensional phenomenon. While different types of maltreatment holds unique associations with outcomes, maltreatment as a whole is the most robust predictor for psychological and behavioural symptoms. This emphasises the importance for practitioners in social services and policymakers to take concrete steps towards establishing and enforcing laws for the prevention, as well as intervention, for childhood maltreatment. Furthermore, since the two forms of deprivation, when distinguished, have fundamental differences in their impact on children’s developmental outcomes, it is extremely useful for intervention studies to reduce the negative effect of poverty, community violence, or other forms of adversity, by promoting a healthy relationship between the children and their caregiver. Moreover, for Japan specifically, where the child protection system is currently undergoing reform, service providers should emphasise the importance of assessing adverse experiences of children prior to protection, and use the information to guide the development of more effective individualized care plans, especially if these results hold are replicated. Future studies using longitudinal prospective design will be more robust and informative for drawing causal relationships between these dimensions of environmental input and outcomes. Follow-up studies on the same group of young people on how the effects of different dimensions of adverse experiences persist will also be useful for understanding the long-term consequences, and shed light in the role of culture in response to adversity, which has been over looked [
47].