Conduct problems refer to behaviors which violate the rights of others and often include aggression, violence, and rule-breaking behaviors. Extant research has identified a link between childhood adversity and conduct problems in childhood and adolescence (e.g., Docherty et al., 2018). Childhood adversity refers to a broad range of negative experiences in childhood, which may include food insecurity, child maltreatment (i.e., abuse, neglect), parental incarceration, and exposure to intimate partner violence (IPV; Felitti et al., 2019). Within the childhood adversity literature, research has consistently found that using a cumulative adversity index is a robust method in predicting negative child outcomes, yet studies using this approach have failed to identify mechanisms that underlie these associations (as thoroughly discussed in McLaughlin et al., 2015). To address limitations of the cumulative adversity approach, McLaughlin et al. (2014) developed the dimensional model of adversity and psychopathology (DMAP), which proposes two dimensions of childhood adversity: experiences which deprive children of environmental stimulation and resources to promote child development (Deprivation) and experiences that threaten the child’s physical and emotional safety (Threat; McLaughlin et al., 2014). The present study examined the processes through which dimensions of childhood adversity differentially impact cognitive (e.g., lower working memory) and affective (e.g., anger dysregulation) risk factors that may contribute to the development of child conduct problems.

Dimensional Model of Adversity and Psychopathology

The DMAP posits that differentiating between adverse experiences is crucial in understanding underlying neurobiological factors that contribute to the development of later outcomes (McLaughlin et al., 2014). This model hypothesizes that deprivation is associated with neurobiological indicators of higher order cognitive processes, including reductions in grey matter volume and thickness globally (McLaughlin et al., 2010). In contrast, threat is hypothesized to be associated with neurobiological indicators of emotional processing, including hyperreactivity of the amygdala in response to facial displays of emotion (Pollak et al., 2000) and lower hippocampal volume, which has been shown to be implicated in verbal memory and emotional processing (as reviewed in Carrion et al., 2010).

Several studies have utilized the DMAP framework in understanding the associations between childhood deprivation and threat in predicting child outcomes (e.g., Miller et al., 2018; Miller et al., 2020; Milojevich et al., 2019). Specifically, research has found that childhood deprivation, but not threat, is linked to lower child verbal ability (Miller et al., 2018, 2020), lower performance on a cognitive control task (Machlin et al., 2019), and lower executive functioning (EF; Schäfer et al., 2022; Johnson et al., 2021). In contrast, threat, but not deprivation, has been found to be associated with emotional processing (e.g., avoidant emotional coping; Milojevich et al., 2019) and more accurate threat discrimination (Machlin et al., 2019).

The DMAP framework has been most frequently used to identify unique neurobiological mechanisms underlying broad psychopathology. A recent study conducted by Miller et al. (2020) utilized the DMAP framework to examine the link between deprivation and threat and child psychopathology within the The Future of Families and Child Wellbeing Study (FFCWS) dataset (the same dataset utilized in the present study). Miller et al. (2020) examined deprivation and threat (ages 1 and 3) in predicting internalizing and externalizing problems at child ages 5, 9, and 15, while examining child verbal abilities at age 5 as a mediator. They found higher deprivation, but not threat, was associated with lower child verbal abilities at age 5, which was linked with higher internalizing and externalizing symptoms at ages 5, 9, and 15. Further, the links between deprivation and both child outcomes at ages 5, 9, and 15 were mediated by age 5 verbal ability. Threat was not associated with child verbal abilities but demonstrated significant positive associations with child internalizing and externalizing problems at ages 5, 9, and 15.

The present study aims to expand upon research exploring the DMAP by including both cognitive and affective risk in a model examining the impact of adversity on developing conduct problems. This approach aims to contribute to our understanding of unique pathways that predict conduct problems in adolescence. Though, it is important to note that cognition and affective risk, although examined as separate dimensions within the model, are closely related and often impact one another. Below, extant research exploring the link between dimensions of adversity and adolescent conduct problems and their underlying mechanisms, are reviewed.

Childhood Social/Environmental Deprivation and Adolescent Conduct Problems

Indicators of deprivation including neighborhood disadvantage (Galán et al., 2017), food insecurity (Jackson et al., 2018), material hardship (Shelleby, 2018), and neglect (Sheridan et al., 2017), have consistently been associated with higher child conduct problems. The DMAP highlights that deprivation uniquely impacts cognitive abilities. One component of cognitive abilities is EF, which is a broad term that refers to several higher order cognitive functions that are important for decision-making, planning, monitoring behaviors, cognitive control, and emotional control (Miyake et al., 2000). Findings have indicated that deprivation is associated with lower EF, even while controlling for threat (Sheridan et al., 2017). Lower EF (Morgan & Lilienfeld, 2000) has been consistently shown to be associated with higher adolescent conduct problems. One facet of EF that may be particularly important to consider is working memory, which refers to skills evaluating and manipulating information in the mind (Miyake et al., 2000). Youth with low working memory may be uniquely at-risk for conduct problems as they may demonstrate difficulties with social decision making, increasing the use of maladaptive forms of social problem solving (e.g., aggression).

Although the pathways between early deprivation and lower EF (e.g., Fishbein et al., 2019) and between lower EF and adolescent conduct problems (e.g., Morgan & Lilienfeld, 2000) are well-established, few studies have identified EF or facets of EF as a mechanism underlying the association between deprivation and child conduct problems (for exceptions see Fatima et al., 2016; Fishbein et al., 2019). The findings of these studies point to the potential influence of lower child EF as a risk-factor for developing later conduct problems following experiences of deprivation, though the cross-sectional nature of these examinations limits our understanding of temporal precedence. The present study aims to address this gap by utilizing a longitudinal model to examine the association between a cumulative deprivation index and adolescent conduct problems mediated by lower child working memory while controlling for threat.

Threat and Adolescent Conduct Problems

Consistently, research has found that indicators of childhood threat, such as physical, emotional, or sexual abuse (Docherty et al., 2018), exposure to IPV (Steketee et al., 2021), and community violence (Galán et al., 2017) are linked with later adolescent conduct problems. As reviewed, the DMAP highlights that threat is uniquely associated with neurobiological indicators of emotional processing, such as stress response system dysfunction (Peckins et al., 2020). In line with these findings, research has consistently identified a link between threat and components of emotional processing (Milojevich et al., 2019). Specifically, research has consistently identified associations between emotional dysregulation following trauma and externalizing problems in adolescence (e.g., Kerig & Becker, 2012). In a review of the link between posttraumatic stress disorder (PTSD) and later delinquency, authors outline that childhood adversity may be associated with later antisocial behaviors through chronic levels of hyperarousal, negative cognitions about themselves/the world, and negative affect (Kerig & Becker, 2012). Emotional dysregulation may make it difficult for youth to engage in adaptive processing of external emotional cues (e.g., recognizing emotional cues; Kerig & Becker, 2012), which is associated with more conduct problems (Miller & Marsee, 2019).

The hypothesis that emotional processing difficulties may mediate the link between threat and conduct problems is well-supported (e.g., Hitti et al., 2019; Cooley et al., 2022; Kerig & Becker, 2012; Miller & Marsee, 2019); however, additional research exploring complex longitudinal designs are needed. Specifically, little extant research focusing on threat has controlled for deprivation, which limits our understanding of the way in which distinct dimensions of adversity may differentially confer risk to developing conduct problems. Addressing this gap, a recent study identified that emotional negativity (e.g., negative affect/reactivity to emotional experiences) at ages 10–12 uniquely mediated the link between maltreatment (during the first ten years of life) and externalizing problems eight years later, whereas it did not mediate the link between maltreatment and comorbid internalizing and externalizing problems. However, it is important to note that both deprivation and threat were associated with higher emotional negativity, highlighting that this facet of reactivity may not be distinct for experiences of threat (Duprey et al., 2022). Additional longitudinal research is needed that integrates the DMAP framework and theories highlighting the role of disrupted emotional processing following trauma in predicting externalizing problems. As such, the present study aims to examine whether anger dysregulation is a mechanism through which threat is associated with conduct problems. Anger dysregulation is a facet of emotional reactivity that reflects intense and frequent experiences of and difficulty managing anger (Sappington et al., 1997) and has been uniquely linked to externalizing problems (Hitti et al., 2019; Cooley et al., 2022).

It is important to note that emotional dysregulation and EF are distinct but related constructs. Emotional regulation or reactivity can be understood as a bottom-up process, such that emotional reactivity occurs without conscious processes. Top-down processes, such as EF, refer to behaviors that require conscious use of cognitive resources. These processes may be related to one another such that bottom-up emotional reactivity may affect the efficiency of the top-down cognitive processes (Ogilvie et al., 2011). As such, it is important to consider that these constructs are likely related and may affect one another across development.

The Present Study

The DMAP adversity dimensions (i.e., deprivation and threat) reflect both proximal (e.g., directly experienced by the child such as physical abuse) and distal experiences of adversity (e.g., neighborhood violence). The present study aims to examine multiple indicators of both childhood deprivation and threat across the social ecology, including community-, household-, and child-level indicators utilizing structural equation modeling. This allows for the examination of a comprehensive adversity construct that considers the importance of both the child’s direct experiences (e.g., physical abuse) and broader contextual factors that shape the child’s environment (Goetschius et al., 2020; Peckins et al., 2019; McLaughlin & Sheridan, 2016). As such, similar to Peckins et al.’s (2020) approach to assessing childhood deprivation, the present study estimated a latent construct reflecting deprivation including the following indicators: neglect as the child-level indicator, food insecurity and material hardship as the household-level indicators, and neighborhood lack of collective efficacy as the neighborhood-level indicator. Additionally, to examine threat, the present study included physical and emotional abuse as the child-level indicators, IPV in the home as the household-level indicator, and community violence as the community-level indicator. Further, the present study aims to examine child working memory at age 9 as a mechanism underlying the association between deprivation and adolescent conduct problems and anger dysregulation at age 9 as a mechanism underlying the link between threat and adolescent conduct problems.

We hypothesized that 1) both higher deprivation and higher threat during the first five years of the child’s life would be associated with higher conduct problems at age 15; 2) higher deprivation, but not threat, during the first five years of the child's life would be associated with lower working memory at age 9, which would mediate the association between higher early childhood deprivation and higher adolescent conduct problems at age 15; and 3) higher early childhood threat, but not deprivation, during the first five years of the child’s life would be associated with higher anger dysregulation at age 9, which would mediate the association between higher early childhood threat and higher adolescent conduct problems at age 15.

Method

Participants and Procedures

The present study utilized data from the FFCWS, a longitudinal birth cohort study that followed 4,898 families from their child’s birth until the child was age 15. Families were recruited between 1998 and 2000 following the birth of their child in 75 hospitals within 20 different U.S. cities (Reichman et al., 2001). Data were collected at the child’s birth, and at child ages 1, 3, 5, 9, and 15. Due to one of the primary study goals being to better understand families of single/unmarried parents experiences during child rearing, a larger proportion of single/unmarried families (75%) were recruited. 48% of mothers reported identifying as African American and 27% as Hispanic. Further, 58% of mothers reported having at least a high school diploma at baseline, and the average baseline household income was $31,900.

Measures

Demographic Questionnaire

Mothers completed a questionnaire at each time point that inquired about demographic variables, including household income, household size, ethnicity, and education. Further, youth completed a demographic questionnaire at age 15 that inquired about their race and ethnicity.

Indicators of Childhood Deprivation

Neglect

A sum of the mother-reported five-item neglect subscale of the Parent Child Conflict Tactics Scale (CTSPC; Straus et al., 1998) was used to examine a proxy for neglect at child ages 3 (α = 0.54) and 5 (α = 0.35). Mothers reported on the frequency of neglectful parenting on a 0 (never) to 6 (> 25 times in the past year) scale. Higher scores indicated higher neglect. Example items include: “had to leave your child home alone, even when you thought some adult should with him/her” and “were so drunk or high that you had a problem taking care of your child.” This scale displayed poor internal consistency at child age 5, α = 0.35; this low internal consistency may be due to low levels of endorsement in the population (Straus et al., 1998).

Material Hardship

Material hardship at child age 1 (α = 0.70), 3 (α = 0.64), and 5 (α = 0.72) was measured utilizing a sum score of 10 items derived from the Ability to Meet Expenses section of the Survey on Income and Program Participation (SIPP; U.S. Census Bureau, 2014). Items assessed access to resources (e.g., received free food) and housing (e.g., got evicted for not paying rent/mortgage); at ages 3 and 5, two additional items related to resource availability were included. A higher score represents higher material hardship.

Food Insecurity

Food insecurity at child ages 3 and 5 was measured utilizing a maternal-report on the 18-item scale developed by the U.S. Department of Agriculture (USDA). Items examined the quality and quantity of the availability of food in the household over the past 12 months (e.g., worried food would run out) on a 0 (did not experience) and 1 (did experience) scale. Items were summed and standardized as described in Nord and Bickel (2002) to create a sum of food insecurity ranging from 0 (unlikely to be food insecure) to 10. Sample items include: worried food would run out, food bought didn’t last, couldn’t afford to eat balanced meal and cut size of children’s meal, children skipped meals, and adult cut size of meal.

Neighborhood Lack of Collective Efficacy

At child age 3 (α = 0.85) and 5 (α = 0.86), neighborhood lack of collective efficacy was assessed using a sumscore of the mother-report five-item social control subscale (e.g., would individuals in the neighborhood intervene if children were showing disrespect to an adult) and five-item social cohesion and trust subscale (e.g., people around here are willing to help their neighborhoods) adapted from the Project on Human Development in Chicago Neighborhoods (PHDCN; Sampson et al., 1997). Mothers responded to items on a 1 (very unlikely/strongly disagree) to 4 (very likely/strongly agree) scale. Higher scores represent lower neighborhood collective efficacy.

Indicators of Threat

Spanking

Spanking at age 1 was assessed using primary caregiver report on the frequency of spanking their child, 0 (never) to 4 (every day or nearly every day).

Physical Abuse

A sum of the mother-reported five-item physical assault subscale of the CTSPC (Straus et al., 1998) was used for a proxy for physical abuse at child ages 3 (α = 0.57) and 5 (α = 0.60). Example items include: “pinched him/her” and “slapped him/her on the hand, arm, and leg”. Mothers reported on the frequency of parenting behaviors using a 0 (never) to 6 (> 25 times in the past year) scale. Higher scores indicate higher levels of physical abuse.

Emotional Abuse

A sum of the mother-reported five-item psychological assault scale of the CTSPC (Straus et al., 1998) was used as a proxy for emotional abuse at child ages 3 (α = 0.62) and 5 (α = 0.62). A sample item includes: “shouted, yelled, or screamed at him/her.” Mothers reported on the frequency of parenting behaviors on a 0 (never) to 6 (> 25 times in the past year) scale. Higher scores indicate higher levels of emotional abuse.

IPV in the Home

A sum of six items adapted from the Conflict Tactics Scale (CTS; Straus, 1990) was used to IPV at years 1, 3 and 5. Mothers were asked how often her romantic partner abused her physically (e.g., “how often does your partner slap or kick you”), sexually (e.g., “how often does your partner force you to have sex/do sexual things”), and/or controlled her behavior (e.g., “how often does he try and prevent you from going to work or school “) using a 0 (Never), 1 (Sometimes), or 2 (Often) scale.

Community Violence

Community violence was assessed using seven-items reported on by primary caregivers modeled after the My Exposure to Violence interviews (Selner‐O'Hagan et al., 1998) at child ages 3 (α = 0.70) and 5 (α = 0.73). Caregivers reported the number of times they encountered seven experiences (e.g., “saw someone get hit, slapped, punched, or beaten up by someone”) in the past year (i.e., never, once, 2–3 times, 4–10 times, more than 10 times).

Child Outcomes

Working Memory

As working memory is a facet of EF, the forward and backward children’s scores on the digit span task, a working memory task, from the Weschler Intelligence Scale for Children (WISC-IV; Weschler, 2003) were used to measure EF at age 9. Standardized scores were examined (M = 10, SD = 3); higher scores indicated better working memory.

Anger Dysregulation

A sum of six items (i.e., child argues a lot, destroys their own things, destroys other’s things, stubborn, sullen, or irritable, and has sudden changes in mood or feelings) reported on by primary caregiver-reports on the Child Behavior Checklist/6–18 (CBCL/6–18: Achenbach & Rescorla, 2001) at child age 9 were used to assess anger dysregulation (α = 0.77). These items reflect the anger dysregulation subscale, which has been shown to be distinct from other scales on the CBCL (Pardini et al., 2018). Primary caregivers reported on these symptoms on a 1 (not true) to 3 (very often or often true) scale. A sum of these six items from the caregiver reported Child Behavior Checklist 2/3 (CBCL 2/3; Achenbach, 1992) at child age 3 (α = 0.75) was included to control for continuity of anger dysregulation.

Child Conduct Problems

Adolescent conduct problems were measured at age 15 using the sum of 13-items (e.g., deliberately damage property that didn’t belong to you; sell marijuana or other drugs) modeled after the delinquency items in the National Longitudinal Study of Adolescent Health (Add Health Self-Report Delinquency [AHSRD]; Harris & Udry, 2014; α = 0.75). Adolescents reported on a 0 (Never) to 2 (Often) scale regarding how often they engaged in different behaviors (e.g., deliberately damage property that didn’t belong to you).

Covariates

Covariates include: maternal race/ethnicity measured at baseline (dummy coded, White as reference group), child-reported race/ethnicity measured at age 15 (dummy coded, White as reference group), maternal education at baseline (0 = high school or greater, 1 = less than high school), maternal marital/cohabitation status at baseline (0 = not married/cohabitating, 1 = married/cohabitating), child sex assigned at birth (0 = male, 1 = female), poverty ratio at baseline (household income divided by the poverty threshold designated by the U.S. Census Bureau), and prenatal exposure to alcohol and drugs (0 = no exposure, 1 = exposure). Working memory at age 3 was not controlled for because working memory data was not available for that age group.

Results

Analytic Plan

The proposed study utilized structural equation modeling (SEM) using the lavaan package in R (Rosseel, 2012) to examine the primary study hypotheses. For all models, adequate model fit was determined if the RMSEA is less than 0.05, the CFI is greater than 0.95, the TLI is greater than 0.90, and the SRMR is less than 0.08 (Hox & Bechger, 1998). First, measurement models of each latent construct were examined. Next, a model that examined the direct effects from both deprivation and threat to adolescent conduct problems at age 15 was estimated. The last model examined the mediators (i.e., working memory and anger dysregulation at age 9) in the relationship between deprivation, threat, and adolescent conduct problems at age 15. In both the direct effect and full model, pathways were drawn from the following covariates to deprivation, threat, and child conduct problems at age 15: mother race (Black vs. Non-Black; Asian vs. Non-Asian), mother ethnicity (Hispanic vs. Non-Hispanic), child-reported race (dummy codes for Black, Multiracial, other race), child-reported ethnicity (Hispanic vs. Non-Hispanic), child sex assigned at birth, baseline maternal education, baseline maternal marital status, baseline poverty ratio, prenatal exposure to alcohol or drugs. Further, two separate pathways were models (1) a pathway from year 3 externalizing problems to year 15 adolescent conduct problems and (2) a pathway from year 3 anger dysregulation to year 9 anger dysregulation were included in the model. The goal of including these pathways was to examine the role of adversity on outcome variables above and beyond baseline-levels of risk. Indirect effects were examined utilizing the bootstrap method within the lavaan package (Rosseel, 2012).

Preliminary Analyses

All values presented in the document are standardized unless otherwise noted. Normality statistics indicated that several variables were positively skewed and leptokurtic. To address non-normality, maximum likelihood estimation with robust standard errors (MLR) was used (Hox et al., 2010). Descriptive statistics are represented in Supplementary Table 1 and bivariate correlations in Supplementary Table 2. Little’s MCAR test (Little, 1988) was significant (χ2 = 8488.43, p < 0.001), suggesting that data were not missing completely at random (MCAR). However, t-tests identified covariates associated with missingness, suggesting data were missing at random (MAR). As such, the following covariates were included in all models: race, ethnicity, marital status, maternal baseline education. The present study utilized full information maximum likelihood (FIML) to address missingness (Newman, 2003).

Measurement Model

The first CFA estimated deprivation and included the following observed variables: neglect at ages 3 and 5; material hardship at ages 1, 3, and 5; food insecurity at ages 1, 3, and 5; and neighborhood lack of collective efficacy at ages 3 and 5. The model demonstrated a significant chi-square (χ2 (df = 33) = 66.79, p < 0.001), which is not atypical for large samples (Hooper et al., 2008). All other fit indices suggested good model fit, CFI = 0.98, TLI = 0.97, RMSEA = 0.03, and SRMR = 0.02. Covariances between same type of adversity across different ages were included (e.g., year 3 and 5 neglect, year 3 and 5 material hardship). Loadings ranged between 0.12 (lack of collective efficacy at age 5) and 0.76 (year 3 food insecurity).

The second CFA represented threat and included the following observed variables: spanking at age 1; physical abuse at ages 3 and 5; IPV in the home at ages 1, 3, and 5; and community violence at ages 3 and 5. This model demonstrated good fit, χ2 (df = 27) = 94.08, p < 0.001, CFI = 0.98, TLI = 0.97, RMSEA = 0.03, and SRMR = 0.02. Covariances between variables of the same type of adversity across ages were included. Modification indices highlighted that covarying emotional abuse and physical assault at each time point significantly improved model fit, consistent with research suggesting high co-occurrence of emotional and physical abuse (e.g., Brown et al., 2019). Covariances between year 3 emotional and physical abuse as well as year 5 emotional and physical abuse were included in analyses. Loadings were statistically significant ranging from 0.05 (year 3 IPV and year 5 IPV) to 0.76 (year 3 physical abuse).

Finally, a measurement model including both latent constructs (covarying each dimension) was examined. This model demonstrated good model fit, χ2 (df = 137) = 636.32, p < 0.001, CFI = 0.93, TLI = 0.92, RMSEA = 0.04, and SRMR = 0.05. Loadings are represented in Table 1. Some variables across the measurement models demonstrate loadings that are lower than the typical 0.3 threshold; however, each indicator significantly loaded onto the latent construct and the models demonstrated good model fit. Knekta et al. (2019) suggested that an appropriate factor loading for latent constructs should be based on the theoretical association between the observed variable and the overall construct. Regarding the present study, we may expect that indicators of adversity that are more distal to the child (such as community-level factors) would demonstrate lower factor loadings compared to adversity that is more proximal (such as physical or emotional abuse). Removing observed variables from the latent construct based on low factor loadings despite significantly loading onto the overall construct and being theoretically important to the construct may decrease content validity (Bandalos & Finney, 2010). Low factor loadings have been maintained in several other studies following similar theoretical justification (e.g., Knekta et al., 2019; Vogel et al., 2021; Byrd et al., 2020). As such, all indicators were retained.

Table 1 Standardized loadings for latent outcome variables in dual CFA model
Full size table

Structural Models

Direct Effect Model

The direct effect model demonstrated good model fit: χ2 (df = 561) = 2079.73, p < 0.001, RMSEA = 0.03, CFI = 0.93, TLI = 0.91, and SRMR = 0.06. In line with hypotheses, findings demonstrated that higher threat was significantly associated with higher adolescent conduct problems (b = 0.11, se = 0.08, p = 0.001). Contrary to hypotheses, deprivation was not significantly associated with adolescent conduct problems (b = 0.01, se = 0.08, p = 0.80). Deprivation and threat significantly covaried with one another (b = 0.48, se = 0.04, p < 0.001). A post hoc model was conducted to explore whether deprivation was associated with child conduct problems at age 15 without controlling for threat. This model demonstrated good fit: χ2 (df = 178) = 757.54, p < 0.001, RMSEA = 0.03, CFI = 0.97, TLI = 0.95. In this model, deprivation demonstrated a trend-level association with child age 15 conduct problems (b = 0.05, p = 0.06).

Full Model

The second model included two mediators (i.e., child working memory and anger dysregulation at age 9) in addition to the direct effects examined in the first model (Fig. 1; Table 2). This model demonstrated good model fit: χ2 (df = 666) = 2616.13, p < 0.001, RMSEA = 0.03, CFI = 0.92, TLI = 0.90, and SRMR = 0.06. In partial support with hypotheses, deprivation demonstrated a trend-level association with lower working memory (b = -0.07, se = 0.09, p = 0.051). Threat was not significantly associated with working memory (b = 0.02, se = 0.09, p = 0.53). In line with hypotheses, higher threat was significantly associated with higher anger dysregulation at age 9 (b = 0.19, se = 0.07, p < 0.001). Contrary to hypotheses, higher deprivation was also significantly associated with higher anger dysregulation (b = 0.09, se = 0.07, p = 0.01).

Fig. 1
figure 1

Full structural equation model depicting pathways from dimensions of adversity to adolescent conduct problems mediated by child working memory and anger dysregulation. Note. Pathways from additional covariates to independent and dependent variables were also included. Standardized values included. Model fit statistics: Chi-square (df = 666) = 2616.13, p < .001, RMSEA = .03, CFI = .92, TLI = .90, and SRMR = .06. ________________ Significant

Full size image
Table 2 Pathways for full model
Full size table

Analyses of indirect effects demonstrated that, contrary to hypotheses, working memory at age 9 did not mediate the association between either deprivation or threat and child conduct problems (95% CI -0.002, 0.004, se = 0.0002, p = 0.88). Consistent with hypotheses, higher anger dysregulation partially mediated the association between threat and higher adolescent conduct problems (95% CI 0.01, 0.04, se = 0.01, p < 0.001). Contrary to hypotheses, anger dysregulation at age 9 also mediated the association between deprivation and conduct problems (95% CI 0.01, 0.02, se = 0.01, p = 0.02). The direct effect of threat on adolescent conduct problems remained significant in this model (b = 0.10, se = 0.08, p = 0.01). Deprivation was not directly associated with adolescent conduct problems in this model (b = -0.002, se = 0.08, p = 0.95). Deprivation and threat significantly positivity covaried (b = 0.48, se = 0.04, p < 0.001). Further, working memory and anger dysregulation at age 9 were negatively correlated (b = -0.07, se = 0.10, p = 0.003). Several covariates were significantly related to primary study variables (Table 2).

Discussion

Findings suggest that threat is particularly meaningful in understanding the link between adversity and conduct problems. In addition, findings of the present study provided support for the DMAP such that higher deprivation, but not threat, in early childhood life demonstrated a trend-level association with lower working memory at child age 9. Anger dysregulation at age 9 emerged as a risk factor of developing conduct problems at age 15 following both early childhood deprivation and threat. Further, anger dysregulation mediated both the association between higher threat and higher conduct problems and the association between higher deprivation and higher conduct problems. working memory did not mediate the link between deprivation and adolescent conduct problems.

Direct Associations Between Adversity and Conduct Problems

The present study’s findings expand upon existing research by identifying a link between higher threat in early childhood and higher adolescent conduct problems above and beyond experiences of deprivation, demographic factors (i.e., maternal education, poverty ratio, race/ethnicity), and prenatal risk factors (i.e., prenatal exposure to alcohol and drugs). Inconsistent with our hypothesis, higher deprivation was not associated with adolescent conduct problems at age 15 while controlling for threat. This is inconsistent with research finding associations between indicators of deprivation, including food insecurity (Jackson et al., 2018), material hardship (Shelleby, 2018), and neglect (e.g., Sheridan et al., 2017) and higher adolescent conduct problems. However, it is important to note that the large majority of extant research did not control for threat. In line with the present study’s findings, two studies examining the DMAP (Miller et al., 2018, 2020) similarly found that threat demonstrated a direct link with higher externalizing (and internalizing problems) whereas deprivation did not. Despite not identifying a direct relationship, both these studies identified an indirect effect such that higher deprivation was associated with higher externalizing symptoms in adolescence through lower verbal abilities. In the present study, a post-hoc model that did not include threat indicated that higher deprivation demonstrated a trend-level association with higher adolescent conduct problems. Collectively, findings suggest that threat may be driving the effect between adversity and adolescent conduct problems; however, additional research is needed to understand these associations. It is also possible that given threat was experienced more often than deprivation, increasing the power to identify an effect within our model.

DMAP Model

Consistent with the DMAP, the present study identified that higher deprivation, but not threat, was associated with lower child working memory at trend-level. These findings are consistent with other longitudinal examinations of the DMAP model linking deprivation with aspects of EF (Vogel et al., 2021; Schäfer et al., 2022; Johnson et al., 2021). Our finding provides support for the DMAP framework and the longitudinal association between higher deprivation and lower working memory, above and beyond the effect of threat, prenatal risk, and demographic covariates. Further, findings demonstrated that both higher threat and deprivation were associated with higher anger dysregulation at child age 9. It is important to note that although both were significant, the effect of threat on anger dysregulation (standardized b = 0.19) was observably stronger than the effect of deprivation on anger dysregulation (standardized b = 0.09). Findings are consistent with research highlighting the link between indicators of deprivation and emotional reactivity (Bogdan et al., 2012). Our finding is in line with Duprey et al. (2022), which found that deprivation was associated with higher emotional lability/negativity (ability to recover from distress) and lower emotional regulation (labeling emotions).

It is possible that our finding that both violence and deprivation were associated with higher anger dysregulation reflects limitations in examining behavioral measures of complex, neurobiological processes. Working memory (as well as broader EF) and emotional regulation are related, such that EF include processes important for emotional regulation (Ogilvie et al., 2011). Higher-order cognitive processes, such as EF, and processes engaged in regulating emotional experiences, involve similar systems. Regulating emotional experiences involves the ventral system of the brain (i.e., amygdala and ventrolateral prefrontal cortex, orbitofrontal cortex, medial prefrontal cortex); higher-order cognitive processing involves the frontoparietal network, including the dorsolateral prefrontal cortex, dorsomedial prefrontal cortex, superior parietal cortex and dorsal anterior cingulate cortex (Dolcos & McCarthy, 2006). These systems are interactive (Beauregard et al., 2001) and, as such, it may be useful for future research to explore how coupling between the affective emotional control and executive control areas of the brain are implicated following exposure to dimensions of adversity.

Indirect Effects: Child Adversity and Conduct Problems

Inconsistent with hypotheses, working memory did not mediate the link between higher deprivation and adolescent conduct problems. This is surprising given the well-established link between lower EF, including lower working memory, and higher conduct problems (Morgan & Lilienfeld, 2000). The dissimilarities in our findings compared to previous research are likely due to methodological differences, as previous research have primarily been cross-sectional. A longitudinal cross-lag model exploring the associations between conduct problems and working memory may be informative to further elucidate how these constructs develop over time.

It is also possible that the present study did not find a link between working memory and conduct problems due to utilizing a single indicator of EF. EF is a multi-faceted construct and working memory alone does not adequately capture the diversity of EF deficits that may confer risk for developing adolescent conduct problems. In addition, working memory was examined six years before conduct problems. It may be that exploring more proximal associations between working memory and conduct problems would enhance our understanding of how difficulties with cognitive control impact social information processing and, subsequently, contribute to the development of conduct problems. Further, anger dysregulation, may tap into the emotional control dimension of EF, was included in the model. It is possible that cognitive control is less predictive of conduct problems when indicators of behavioral/emotional control are considered in the same model. Future research should explore whether components of EF differentially relate to conduct problems to further understand this link.

The present study found that the association between both threat and deprivation and child conduct problems was mediated by higher anger dysregulation. Few studies have examined this mechanistic pathway while controlling for deprivation and cognitive outcomes in the same model. Our finding is in line with models highlighting trauma-based reactions, which propose emotional dysregulation as one pathway through which conduct problems develop following child adversity (e.g., Bennett & Kerig, 2014). Research suggests that emotional dysregulation may make it difficult for youth to engage in adaptive processing of external emotional cues (e.g., taking perspective; Kerig & Becker, 2012) and youth may engage in risky or antisocial behaviors in an effort to reduce their distress (Hayes et al., 1996). Of note, our findings highlighted that this pathway was significant for both deprivation and threat, highlighting a mechanism through which adversity is linked with conduct problems.

Findings also identified that the pathway between threat and child conduct problems at age 15 remained significant even in the full model, which included both working memory and anger dysregulation. This suggests that, although anger dysregulation partially explains the pathway between early threat and adolescent conduct problems, additional mediational mechanisms are important to consider. In addition to examining neurobiological processes and individual factors, the development of conduct problems has been found to be heavily influenced by contextual factors, such as lower parental monitoring and engagement with deviant peer groups (e.g., Chung & Steinberg, 2006). Future studies should consider contextual factors, along with neurobiological indicators, as mechanisms underlying the link to threat and conduct problems to explore a comprehensive explanatory model.

Limitations and Future Directions

The present study expanded upon research exploring the DMAP by including both dimensions of adversity and both affective and cognitive outcomes within the same model. Additional strengths of this study included utilizing a longitudinal design in a large cohort study, including a standardized measure of working memory, and exploring outcomes with varying reporters (e.g., caregiver report and self-report).

Despite these strengths and contributions to the literature, the present study should be interpreted considering several limitations. First, working memory is a single indicator of one component of EF (working memory) and anger dysregulation examined the ability to regulate anger; as such, future research should consider utilizing more comprehensive measures of EF and other types of emotional dysregulation to provide a more robust examination of these pathways. Further, there were several years between measurements of adversity and working memory/anger dysregulation (four years) and working memory/anger dysregulation and conduct problems (six years). Exploring these pathways at more than one timepoint may be useful in elucidating the complex way in which adversity impacts conduct problems. In addition, caregivers reported on child anger dysregulation. Additional reporters, including child’s self-report on their emotional and/or anger dysregulation may be beneficial, especially given some aspects of dysregulation may be less observable by caregivers (e.g., emotional numbing). Future research should also consider using physiological indicators of emotional processing (e.g., emotional reactivity to negative and/or threatening stimuli) to examine objective measures of dysregulation.

As reviewed, EF and anger dysregulation demonstrate overlap and may influence one another. A more nuanced examination of these processes in the context of the DMAP may elucidate the differences between dimensions of adversity and components of EF and emotional processing. For example, future research should consider examining cognitive, behavioral, and emotional control utilizing paradigms that allow for the isolation of top-down and bottom-up regulation. Additionally, although the present study controlled for an earlier measure of anger dysregulation, we did not include an earlier measurement of working memory. This limits our ability to assert whether deprivation was associated with lower working memory above and beyond working memory in early childhood.

Additional limitations of the present study include potential model-level characteristics that impacted our ability to identify significance. Specifically, several loadings for the CFAs in the present study were lower than the typical threshold of b = 0.3. In addition, we controlled for an earlier measure of anger dysregulation (at age 3) and an earlier measure of externalizing problems (at age 3). These variables were derived from the same measure (the CBCL) and include some overlapping items. Statistically, these variables were highly correlated at the bivariate level (r = 0.89) and reflect high overlap between variables. It is also important to note that the present study summed scale scores and utilized those scores within latent constructs. However, there is emerging statistical research that recommends utilizing latent constructs rather than sum scores within analyses to maintain the structure of the validated measure (McNeish & Wolf, 2020). Future research should consider more complex models that allows for this nuance.

Further, although exploring dimensions of adversity provides greater specificity in line with empirically supported theories, there is still a substantial amount of variability in our conceptualization of adversity, including developmental time period, proximity to the child, and perceived impact/intensity of the adverse experience. Future research should consider additional characteristics of adversity, including the time period, whether it is experienced chronically or episodically, while examining differential outcomes based on dimensions of adversity. In addition, there is a high correlation between deprivation and threat (r = 0.49), suggesting that these experiences of adversity likely co-occur. As such, it may not represent ecologically valid profiles of adversity. Future studies should examine the interaction between adversity dimensions.

Conclusions/Implications

Few studies have examined the DMAP including both cognitive and affective outcomes in the same model. This approach strengthens support for the DMAP. In practice, identifying relevant behavioral correlates of neurobiological processes is important as it may increase efficacy of interventions developed for youth who have experienced adversity. For example, our findings identify that working memory may be uniquely impacted for youth who experience deprivation. Interventions for youth who have experienced deprivation may benefit from integrating a focus on developing cognitive control. Relatedly, we identified anger dysregulation as a shared outcome for both dimensions of adversity, suggesting that interventions for youth who experience any adversity should target managing emotional control. This is consistent with some trauma-informed therapeutic approaches that focus on developing emotional regulation and coping skills (trauma-focused cognitive behavioral therapy; Cohen et al., 2012).

Findings demonstrated that parsing dimensions of adversity is important, as threat appears to be uniquely associated with the development of conduct problems in adolescence above and beyond other indicators of adversity. In addition, the present study focused on identifying mechanisms through which adversity is associated with child conduct problems. Few studies have examined these pathways utilizing a longitudinal approach capturing the mechanisms that underlie adversity in early childhood, cognitive and affective outcomes in middle childhood, and adolescent conduct problems. In the present study, the association between threat and conduct problems operated through higher anger dysregulation, further strengthening established theoretical frameworks highlighting that early threat may confer risk for later conduct problems through dysregulation and other trauma-related symptomology.