Introduction
Adverse life experiences (ALEs) are some of the most robust environmental predictors of future psychological, social and economic outcomes [
31,
73]. The original Adverse Childhood Experiences (ACE) study performed by the Centers for Disease Control and Prevention in the mid-1990s demonstrated a marked dose–response relationship between the count of childhood adversities and increased problems in the affective, cognitive, somatic, aggression and substance use domains [
3]. This operationalization of counting the number of specific adversities was named the ACE score and it attracted substantial interest as both a tool for screening and for studying etiological factors of disease [
30,
60]. The use of the cumulative ACE score, however, has been challenged on both these fronts [
26,
55]. A person who experienced parental incarceration at age four could conceivably receive the same ACE score as a teenager who was subjected to sexual abuse at fifteen. This has produced criticisms that the current approach neglects critical dimensions of adversity, such as the type (e.g., parental deprivation versus threatening abuse) and timing (e.g., age, occurrence and frequency) of the adverse events. These aspects can significantly influence the impact of childhood experiences, yet are overlooked in the cumulative scoring method [
4,
54]. Attention has therefore shifted to investigating how type and timing of adversity can account for the development of emotional and behavioral problems.
Extensive research has been devoted to investigating the association between specific types of abuse and psychopathology [
10,
32,
45]. Isolating distinct adversity categories, however, is not straightforward as co-occurrence of multiple childhood events is overwhelmingly the norm rather than the exception [
14]. For instance, up to 63% of people that report one type of childhood maltreatment also report other abuse [
80] Thus, rather than replacing the original ACE list with individual adverse events, researchers have proposed several aspects on which cumulative approaches can be improved. Two will be highlighted here. First, a more comprehensive set of adversities guided by previous research has been suggested to expand the list of adversities [
27]. Less visible events such as peer and school difficulties, constant moving and a disorganized family environment all pose unique developmental challenges that can be incorporated in a broader operationalization of adversity [
25,
70,
75]. Authors have advocated for using different terminology to differentiate this broader operationalization from the original ACE score, like for example early adversity or adverse life events (ALEs) [
12]. Second, while individual events might be impossible to isolate, salient characteristics shared by groups of events can be investigated separately. One theoretical research line considers threatening events (such as physical or sexual abuse) separately from other adverse experiences which might be more depriving in nature (such as losing a parent, neglect, academic problems) [
54]. Empirical investigations using the latter approach are still in early stages and studies are yet to compare using a broad cumulative operationalization of ALEs including all events compared to ALEs consisting exclusively of physically threatening events such as sexual and physical abuse. Exploratory studies using these new operationalizations are needed to investigate the development of emotional, behavioral and cognitive problems similar to the tradition of the original ACE methodology.
Beside the issue of grouping events, the dimension of timing is also key in understanding the mechanisms through which adversity operates. According to theory, as the developing brain matures, different brain structures undergo periods of heightened neuroplasticity and accordingly periods of heightened vulnerability to stress as well [
47]. For example, the hippocampus roughly develops in the first 5 years of life, after which the amygdala goes through a period of increased sensitivity in the ages of 8 to 12, while the prefrontal cortex peaks later in the teenage years [
46]. Building on that knowledge, researchers started exploring the distal effects of stress in relation to how its timing affects psychopathological outcomes. The empirical findings have been mixed. Some researchers report that physical and sexual abuse in childhood were associated with depressive problems, anxiety symptoms, and suicidal ideation regardless of when they happened, as found in both large cohorts [
20,
22] and longitudinal designs [
18]. Other studies report adversity before adolescence (e.g. before age 11) to be more predictive of psychosis and depression when compared to adversity in adolescence [
2,
50,
66]. Yet others report adversity before age 6 as most predictive of anxiety and depression problems [
43]. Finally, timing of adversity has sometimes been reported not to be of consequence to transitioning into psychiatric disorders, as only cumulative lifetime scores were found to be significantly associated with psychopathology [
20,
22,
35]. These findings are difficult to compare and combine due to several design differences, which can be found in the sample populations; the outcomes under study and the analysis strategies used. To form a solid knowledge base, studies investigating a broad range of outcomes covering the internalizing/externalizing spectrum are necessary. This approach is analogous to the outcome-wide epidemiologic approach [
78]. Briefly, the outcome-wide approach argues that it is desirable to relate a single category of exposure to many health outcomes in a single study while controlling for shared sources of bias. This approach has rarely been applied in the context of childhood adversity. In particular the timing dimension of adversity has not been studied in relation to a variety of mental health outcomes using a single sample and analysis strategy.
Finally, childhood adversities affect men and women differently. Women are 1.5 to 5 times more likely to develop anxious or depressive responses to adverse events [
37,
69]. Beyond emotional problems, women are also more likely to report behavioral problems following childhood adversity [
24]. These differences are not present in early childhood, but emerge around ages 11 to 13 when girls also start developing depressive problems at an exponentially higher rate [
7,
63]. While biological differences in this early pubertal period have been a leading hypothesis for the mechanism behind observed sex differences, interactions between (neuro)hormonal factors and adversity can only account for up to a quarter of the variance in psychopathology [
13,
67]. Wider social factors clearly have an important role in this relationship, some of which could be captured by broader inclusion of adverse events. Regardless of how social or biological the etiology behind observed differences is, it is clear that sex is an important moderator when studying the relationship between adversity and psychopathology.
The current cross-sectional study investigates the dimensions of type and timing of ALEs and their association with psychopathology in a population-based cohort of adolescents. We take an outcome-wide approach to examine a wide range of emotional and behavioral psychopathology domains as outcomes. Those include anxious-depressed, withdrawn-depressed, somatic, rule-breaking, aggressive, social, attention, and thought problems, as well as psychotic experiences. To investigate the dimension of type of adversity, we scored ALEs in two ways: one which includes a wide array of adverse experiences (broad operationalization) and one which separately considers physically threatening experiences (e.g. sexual and physical violence) in line with distinctions made in the theoretical literature. To investigate timing, threatening and broad adverse life events were categorized according to when they occurred, namely in the first 3 years of life, between ages 4 to 8, ages 9 to 12, or after the age of 12. While we hypothesize a positive association between adversity and psychopathology, past literature is equivocal on specifying which periods of life should be particularly vulnerable to adversity. We therefore do not pre-specify which period we expect to be most associated with psychopathology. Third, we address the role of sex in moderating the association with psychopathology.
Discussion
The current study examined three dimensions of ALEs (type, timing and sex differences) in their association with psychopathology outcomes. Several clear patterns emerged. First, the lifetime associations with broad ALEs (which included hospitalization, parental divorce, repeating a class) showed a bigger effect size than the physically threatening operationalization (which only included sexual and physical abuse). However, the latter category of ALEs was found to be more robust to unmeasured confounding and thus associations were less likely to arise due to an unobserved variable. Second, adverse events in later years, particularly between ages 9 to 12, were found to be most associated with psychopathology regardless of type of ALE. Due to low numbers of physical ALEs in the age period up to 3 years old, we could not provide strong evidence for associations in this earliest period. Third, a markedly consistent pattern of sex moderation emerged. Whenever there were sex differences in effects, girls were always estimated to have a stronger association between ALEs and psychopathology than boys. These differences were most often found in the broad operationalization of ALEs, particularly for events happening after age of 12 and for chronic ALEs. The implications of each of these general patterns are discussed.
Differences between broad and physical adverse life events associations
Adolescents with more ALE in their lifetime history also self-reported more emotional, behavioral problems on all psychopathology outcomes measured. This positive lifetime association is solidly rooted in previous literature, which has reported similar links to internalizing/externalizing problems in children [
21], young adults [
68], and in later adulthood [
42]. As in the current study, these associations remain after adjustment for socioeconomic factors and are robust when using official records to measure abuse [
40,
49]. Going beyond previous literature, we found that using a broad operationalization of adversity produced bigger effect sizes on the associations with psychopathology than when limiting the definition only to physical and/or sexual adversity. This finding is perhaps not counterintuitive, considering that broad ALE predictor by definition contains more information about family dynamics, household dysfunction and the general environment growing up [
8]. The results from the current study show this information can be exploited to more effectively predict mental health symptoms of any type. This better predictive value of broad ALEs, however, comes at a cost. Although the effect sizes were bigger for broad ALEs, the associations were also vulnerable to confounding by unmeasured factors. This suggests that the same environmental factors (e.g. household dysfunction), which are closely linked to broad ALEs, can both increase its predictive power and simultaneously confound its associations [
79]. This has considerable implications for any theoretical model that attempts to causally map the effects of adversity. The physical ALE predictor, on the other hand, was found to have an association of smaller effect sizes with the psychopathology outcomes, but this time very large unmeasured confounders were needed to remove those associations. The current study therefore provides stronger evidence for physical and sexual abuse rather than broad ALEs as probable etiological contributors to psychopathology. The limited number of studies attempting to establish causal effects of adversity indeed support physical and sexual abuse as a causal factor for, for example, antisocial behavior and maladjustment [
41,
76]. Taken together, this information points to a fundamental trade-off when using broad versus physically threatening operationalizations of ALEs. Broad ALEs provide better predictive value above and beyond key sociodemographic and parental factors. Physical ALEs are more suited for an etiological study into the mechanism behind psychopathology.
In terms of timing, we found the developmental period in the ages 9–12 was overall most often associated with psychopathology outcomes. This corroborates some empirical findings reported so far in the literature on adolescent mental health. For instance, a study which looked into harsh physical parenting and its timing effects on psychopathology found that age 9 was a particularly sensitive period [
17]. Girls were most likely to develop internalizing problems at that age following harsh physical discipline, however boys were more sensitive at age 5. Another large cohort study focusing on children reports that middle childhood (around age 7) was when sexual and physical abuse produced most total psychopathology [
21]. Physically assaultive events were also more predictive of later depression for events experienced in childhood (age < 12) than adolescence or adulthood in a twin study design [
53]. There are, therefore, studies using children, adolescent and adult samples that report results similar to the present ones. Some researchers have made sense of these findings by positing that the beginning of puberty is a particularly sensitive period for adverse events due to accelerated hormonal and neurodevelopmental changes [
51]. Equally however, there are studies that do not detect sensitive age periods for psychopathology after interpersonal violence exposure [
22,
35]. Machine learning methods have also produced equivocal findings that adversity at ages 5–14 is most predictive of psychiatric symptoms [
65] and adversity around age 5 is most predictive of positive psychosis symptoms [
64]. These wide differences in designs, predictors, outcomes and methodology underline the importance of using outcome-wide approaches similar to the current study. The current results suggest that both internalizing and externalizing problems are most associated with adversity in the 9–12 age period.
Heightened vulnerability of girls to ALE
The current study found that girls were at higher risk for nearly all externalizing, internalizing and cognitive outcomes after exposure to broad chronic ALEs or ALEs after age 12. Additionally, girls were at higher risk for externalizing problems after chronic physical ALEs (e.g. physical and sexual abuse), and physical ALEs after age 12. Analyses for ALEs after age 12 were adjusted for prior chronicity. Therefore, most likely, different mechanisms underlie these two findings. There were less sex differences with smaller effect sizes for lifetime and middle-childhood ALEs. Chronic adversity has previously appeared to be associated with psychopathology only with very small effect sizes when not considering the role of sex [
23]. Studies that explicitly model sex differences, however, consistently report bigger effect sizes in girls exposed to adversity for depression [
36], psychosis [
29], PTSD [
37] and delinquent behavior [
48]. To our knowledge, this is the first study to show similar sex differences for chronic adversity and psychopathology. Notably, prior studies that investigated the timing component of sex differences report findings that parallel the present results. Breslau and colleagues (2017) used large-scale national survey data to study when girls surpass boys in reporting depression. Similar to the current sample, they estimate that age 12 is when statistically distinguishable differences emerge and continue to increase until around age 17. The present results suggest different responses to ALEs could be one explanation for that divergence. In a similar manner to the current analysis, Harkness et al. [
36] studied how age moderates the effect of various ALEs with depressive disorders. Unlike the present results, they found that ALEs after age 17 was when women were significantly more vulnerable to adversity than men. The authors themselves, however, note a major limitation in their adolescent group that they had 17 boys in total. With the current bigger sample, we were able to show that events experienced in adolescence after age 12 were more associated with psychopathology for girls. This sex difference again has been theoretically attributed to differences in pubertal hormonal changes and prescribed affiliative gender norms for girls [
13]. Although the mechanisms behind the sex differences remain ground for future research, the current investigation advances the knowledge about which ALEs could be prioritized for future study. Namely, chronic and late childhood experiences instead lifetime cumulative scores could be of particular research interest.
Strengths and limitations
A range of limitations should be noted about the present study. First, the chosen mode of outcome-wide analysis was strictly on the exploratory end of the spectrum. In practice this meant we could only interpret general trends in the data which are extremely unlikely to be Type I errors (e.g. sex interactions for all outcomes being of the same direction, across all outcomes association concentrating in the middle childhood age group). Therefore, a substantial limitation is that we could not confidently conclude anything about the specific associations between predictors and outcomes (e.g. definitively conclude physical abuse in ages 9–12 is positively associated with psychotic experiences). Those types of conclusions require rigorous adjustments for multiple hypothesis testing. Even limiting the scope of the investigation to a subsection of outcomes would require statistical power far beyond the reach of existing cohorts, especially for the high-risk adolescent population we report on here. We have argued, however, that the general conclusions offered here still meaningfully advance our understanding of how ALEs and psychopathology relate. An added strength is that the current high-risk sample is a particularly adequate population for this research question, as many studies focusing on childhood adverse experiences use cases from social services or other governmental referrals [
44]. In order for governmental intervention to occur, those cases selectively represent extreme and visible abuse, which do not capture the full range of less harsh adverse experiences [
16]. Conversely, using parent-reported events could miss important events in the later years of the child or abuse perpetrated by the parents themselves [
56]. Misclassifying abused adolescents as not abused could have conceivably biased associations towards the null. The current sample offers an advantageous balance of both heightened risk of reporting adversity, while covering a broad spectrum of experiences. There were physically threatening events we could study across all age and sex groups, except the earliest infancy period up to age 3 where events were reported only rarely. Due to the sampling strategy, the current high-risk cohort is not representative of the general population and thus the distributions of adversity and psychopathology are markedly distinct. However, the associations reported here can generalize to the general population due the inclusion of adolescents not at high risk of psychopathology [
62]. Limitations also emerge from the cross-sectional nature of the study. It is conceivable that some reciprocal relationship exists between current levels of psychopathology and experiences of academic hardship or even parental dynamics resulting in divorce. Such a case is harder to make, however, for physical and sexual abuse events. We also took into account a wide range of socioeconomic variables, parental style and history of psychopathology, and prior chronic experiences of adversity when reporting on each association. We also conducted sensitivity analyses for unmeasured variables, which provided information on which associations are robust and which are likely to disappear under inevitable residual confounding.
Conclusions
The current study shows when adversity may be most predictive of emotional and behavioral problems, especially in the middle-childhood years as adolescents transition into puberty. Furthermore, increased attention is warranted towards adolescent girls with chronic or recent experiences of broad adversity. Although physical or sexual abuse are striking events in an adolescent’s history, a girl with less conspicuous adversities might still be considered at a higher risk for developing emotional or behavioral problems. The operationalization of ALE as physically threatening, on the other hand, may be more useful in understanding etiological origins of psychopathology. As such, they are prime intervention targets to prevent future psychopathology of any kind. Future confirmatory studies could focus on any specific combination of ALEs and a psychopathology outcome to investigate other moderating mechanisms, changes over time and effectiveness of interventions like resilience training. Additionally, studies could investigate which protective factors attenuate the association with psychopathology in vulnerable age periods of childhood. A strong social support network or finding meaning in one’s experiences, for example, are some of the known factors that can lower or even reverse the effects of trauma into what is sometimes referred to as posttraumatic growth [
57,
59]. Finally, we make recommendations on how ALE should be operationalized in practice. A nuanced decision should be made depending on the purpose of measuring ALEs. For prediction purposes, a broad operationalization of ALEs, including family divorce, school difficulties, and parental health, is best suited, even if the adolescent socioeconomic status and their parents’ mental health is known. This could be particularly useful for screening purposes, as there is already a strong movement in pediatric care to identify adversity early and promote practices for a supportive family environment [
39]. Taken altogether, the current study provides information for which operationalizations of adversity are most associated with psychopathology, at what time point and in what sex.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.