Developmental differences in the association of peer relationships with traumatic stress symptoms

Abstract

Background:

Although childhood trauma exposure has a high incidence, traumatic stress often goes untreated in children and youth. We investigate peer relationship quality as a prevention strategy for reducing traumatic stress across different developmental periods.

Methods:

We analyzed longitudinal data from the National Survey of Child and Adolescent Wellbeing (NSCAW I) using a time-varying effect model (TVEM) to investigate the association between peer relationship quality and traumatic stress symptoms across ages 8–17 years. We controlled for a robust set of confounders identified through a Directed Acyclic Graph (DAG).

Results:

The unique association between peer relationship quality and traumatic stress symptoms was negative and significant from ages 8 to 8.5 years, and again from ages 9.4 to 10.9 years and at age 16.4 to 16.8 years, with maximum associations of −1.45 T-score points at age 8.5 years (95% CI = [−2.87, −0.40]), −1.57 at age 9.4 years (95% CI= [−3.13,−0.01]), and −1.89 at 16.7 years (95% CI= [−3.70,−0.09]).

Conclusion:

Peer relationship quality protected against traumatic stress during specific times during adolescent development. Our results suggest that helping youth establish and maintain positive peer relationships may be a useful prevention approach for helping them cope with trauma experiences.

INTRODUCTION

Childhood trauma includes adverse events that occur early in life, such as exposure to neighborhood violence or experiencing abuse. Annually, around 38% of American youth witness serious episodes of community violence, 37% experience physical assault, 6% experience sexual victimization, and 6% witness interpersonal violence (Finkelhor et al., 2009, 2015; Zinzow et al., 2009). These events increase risk for traumatic stress, and this traumatic stress is linked to a variety of poor health outcomes later in life including depression, sleep disturbances, obesity, and substance use (Chapman et al., 2004; Edwards et al., 2003; Felitti et al., 2019; MacMillan et al., 2001; Molnar et al., 2001). Various mechanisms link childhood stress to later health, including neural responses to threat that alter brain structure and function, psychological distress, inflammatory responses, and limited psychosocial resources (Taylor, 2010). Although childhood trauma exposure has a high incidence, protective factors within the child’s environment can prevent traumatic stress symptoms and thwart the transmission from traumatic experiences to poor health—specifically the presence of psychosocial resources despite early trauma. We studied a specific protective factor—peer relationships—and evaluated how its protection against traumatic stress might vary across different developmental periods from childhood to late adolescence.

The Transactional Model of Stress and Coping (TMSC) suggests that social support can buffer the relationship between exposure to stressors—such as early life adversity or trauma—and stress symptoms (Ahrens et al., 2008; Lauterbach & Armour, 2016). Specifically, strong peer relationships—i.e., when an individual has trusting peers that he or she can turn to in times of need for support—can reduce feelings of anxiety and stress through the emotional, instrumental, appraisal, and informational support peers provide (Bokhorst et al., 2010; Glanz et al., n.d.; Lee & Goldstein, 2016; Richardson et al., 1995). Social support is a particularly important protective factor to examine because it is modifiable, cuts across various settings, and moves away from focusing on only individual-level factors (Cengiz & Ince, 2014; Eather et al., 2013; Fletcher et al., 2008; Steese et al., 2006). Moreover, clinically-based individual-level approaches to trauma treatment can be resource-intensive—especially in cases where trauma is not identified and intervened upon in a timely manner (Greer et al., 2014). Yet, the potential role that social support could play in mitigating traumatic stress after trauma exposure remains unexplored, particularly as the nature of social support may change across different developmental periods.

Normative adolescent development includes the increased desire for autonomy and movement towards independence combined with an increasing reliance on friends for emotional (e.g., turning to friends for support when coping with distress) and instrumental support (e.g., asking friends for help to meet a specific need) compared to younger children for whom friendships may primarily consist of engaging in shared activities (Bokhorst et al., 2010; DuBois & Silverthorn, 2005; Miller et al., 2020; Newcomb & Bagwell, 1995). Thus, peer relationship quality may become increasingly important source of social support and means for reducing traumatic stress symptoms as children move into adolescence. Peers may function as an important source of social support for children who experience maltreatment, as, support from parents may be lacking and/or unstable in the context of child maltreatment (Pickreign Stronach et al., 2011; Sousa et al., 2011). Yet, few researchers have studied the role of peers across development in the context of trauma. Therefore, we aim to identify ages when peer relationship quality is most strongly associated with reduced traumatic stress symptoms among youth. To do so, we analyze longitudinal data from the National Survey of Child and Adolescent Wellbeing (NSCAW I) using a time-varying effect model.

METHODS

Study Sample

We used data from the restricted-use National Survey of Child and Adolescent Wellbeing (NSCAW). NSCAW is a nationally representative, longitudinal survey of children and families who have been the subjects of investigation by Child Protective Services. The present study uses the first of two cohorts enrolled in NSCAW. We include NSCAW data drawn from children, parents, other caregivers, caseworkers, teachers, and administrative records (Dowd et al., 2004). The full NSCAW I sample consists of 6,228 children, ages 0–14 years at baseline, who first had contact with the child welfare system between October 1999 and December 2000 (Dowd et al., 2004). Data collection occurred in five waves over eight years: at the close of investigation (Wave 1), 12 months (Wave 2), 18 months (Wave 3), 36 months (Wave 4), and 59–96 months (Wave 5) post-investigation (Dowd et al., 2004). Presently, we exclude Wave 2, as this data collection period did not include child interview. Research Triangle Institute (RTI) Institutional Review Board provided approval of the original NSCAW studies, and NSCAW study staff obtained informed consent from adults and caregivers in the studies. The University of Michigan Institutional Review Board approved the current study.

Variables

Outcome.

At Waves 1, 3, 4, and 5, individuals 8–17 years old completed an adapted Trauma Symptom Checklist for Children (TSCC) post-traumatic stress scale. This 10-item measure assesses the effects of childhood trauma exposure through the child’s self-report of trauma symptoms (e.g., how often they remember things that happened that they did not like; Briere, 1996). The TSCC has demonstrated strong internal consistency and concurrent validity with the Child Behavior Checklist (CBCL; α=0.87, r=0.72–0.80)(Briere, 1996). For each item, the child recorded the frequency with which the statement pertains to her/him on a 4-point scale ranging from 0 (never) to 3 (almost all the time). We used standardized T-scores for analysis. T scores at or above 65 for the traumatic stress scale are considered clinically significant (Briere, 1996).

Predictor.

The same individuals also completed the Loneliness and Social Dissatisfaction for Young Children scale, which evaluated peer relationship quality for children enrolled in school (α = 0.89) (Asher & Wheeler, 1985; National Survey of Child and Adolescent Well-Being (NSCAW): Restricted Release Version Appendix-Volume II, 2008). The 15-item scale covered topics such as if it was easy for the respondent to make friends, if the respondent had friends to talk to at school, if the respondent could find a friend when s/he needed one, and if there were kids at school that the respondent could go to when s/he needed help. We scaled items so that higher values indicated stronger peer relationship quality. We summed all 15 items to create a total peer relationship quality score and standardized this measure for analyses.

Covariates.

We developed a Directed Acyclic Graph (DAG) to identify various time-invariant and time-varying variables that could confound the association between peer relationship quality and traumatic stress symptoms (Supplemental Figure 1, available online). A DAG is a conceptual tool traditionally used in epidemiological research to represent causal relationships between variables that are relevant to a research question (Greenland et al., 1999; VanderWeele et al., 2008; VanderWeele & Robins, 2007). This tool aids researchers in identifying the minimally sufficient adjustment set of variables that should be included as covariates in an analytic model to reduce confounding bias. Based on our DAG, we controlled for factors known to influence peer relationship quality and traumatic stress, but we did not include variables on the causal path between peer relationship quality and traumatic stress, as doing so would bias estimated effects. Covariates included: child’s biological sex; types of initial maltreatment; family risk factors; receipt of mental health services in the last 12 months; emotional and behavioral problems; depressive symptoms; receipt of any financial assistance; and a child’s placement setting.

Time-invariant covariates.

At Wave 1, the caseworker identified types of initial maltreatment that the child experienced; we categorized responses into physical abuse, sexual abuse, neglect, emotional abuse, and other maltreatment. At Wave 1, caseworkers also identified family risk factors based on the information available to them at the time of the case investigation but not based on a standardized measure. Risk factors that caseworkers assessed included: if a caregiver had alcohol abuse, drug abuse, mental health/emotional problems, intellectual/cognitive impairments, poor parenting skills, a history of domestic violence, or low social support at the time of investigation. The present analyses included these risk factors as dichotomous variables.

Time-varying covariates.

At all Waves, a child’s current caregiver provided information regarding whether the child received mental health services specific to emotional, behavioral, learning, attentional, and substance abuse treatment in the past year from the following sources: inpatient treatment, residential treatment, outpatient treatment, therapeutic nursery treatment, day treatment, in-home counseling, hospital emergency room treatment, school counseling, and primary care outpatient treatment. We collapsed all sources of mental health services, thereby creating a dichotomous variable for whether or not a child received any type of mental health treatment in the last year.

At all waves, caregivers and teachers completed the Child Behavior Checklist (CBCL; α = 0.95–0.96)(Achenbach, 1991) and the Teacher’s Report Form (TRF; α = 0.96)(Achenbach & Rescorla, 2001) to assess if a child had an emotional or behavior problem. To avoid multicollinearity issues related to emotional and behavioral problems, we created a dichotomous variable that designated if the child had an emotional and/or behavioral problem at each wave defined as any CBCL or TRF T-score > 64.

We also controlled for a child’s depressive symptoms as a sum score from the Children’s Depression Inventory completed by the child (α = 0.71–0.87)(Kovacs, 1991). This is a 27-item scale with a 0–2 rating scale for each item (total raw scores 0–54).

At all waves, current caregivers indicated whether anyone in the household was currently receiving any financial assistance, and we created a dichotomous variable that described whether or not financial assistance in the household was present. At all waves, the child, caseworker, and caregiver provided information on the home setting of the child (e.g., foster home, kin-care home). We created a dichotomous variable representing whether a child was at an in-home or out-of-home placement.

Analysis

We structured the data with multiple rows per respondent—a row per each age between 8 and 17 years reflecting mid-childhood and the transition into adolescence—each row known as a ‘person-year.’ We restricted the sample to this age range because the measurement tool used to assess our outcome—traumatic stress—was designed for this age group (Briere, 1996). Of the original 6,228 children, 3,746 children fell into the age range of 8–17 years at some point within the longitudinal study (i.e., between Waves 1–5). The first measurement period for each child within our analytic dataset, however, varied. For example, some children entered the analytic sample at Wave 1 if they fell within the age range of 8–17 years. Many children, however, were too young at Wave 1 to receive the TSSC measure, and these children did not enter into the present analytic sample until Waves 3, 4, or 5 (18 months-96 months post-investigation). Due to this staggering of age and entrance into the analytic sample, a child’s age did not necessarily reflect recency of trauma. Therefore, child age in years (rather than time since investigation) was our time-variable.

For the analysis, person-years from any measurement occasion when an individual had information on the outcome, predictors, and complex survey design variables (defined below), was included regardless of later attrition, as TVEM can accommodate different spacing of measurement occasions (Li et al., 2015). Of the 3,746 children in the correct age range, 2,392 had complete information on all analytic variables, which included 5,390 person-years (an average of around two observations per child over time).

We used a time-varying effect model (TVEM) to examine how peer relationship quality was associated with traumatic stress symptoms over the time period between childhood to adolescence (8–17 years). TVEMs estimate how associations between predictors and an outcome change over time without assuming associations follow parametric functions of time, unlike traditional linear models that require growth to follow a parametric function. TVEM does not, for example, force a relationship between two variables to be linear or quadratic with respect to time (Li et al., 2015). Whereas researchers often apply structural equation modeling-based and multilevel modeling-based growth models to characterize growth in a specific construct over time, we can apply TVEMs to characterize change in a specific relationship over time. Our analysis controlled for child’s biological sex, type of initial maltreatment, and child welfare worker’s initial risk assessment as time-invariant covariates with time-invariant effects. We controlled for a child’s receipt of mental health services in the last 12 months, emotional and behavioral problems, depressive symptoms, receipt of any financial assistance, and a child’s placement setting as time-variant covariates with time-invariant effects. Thus, the values of these variables could change between waves, but we restricted the magnitude and direction of the effect of these variables to be constant across waves. To address the concern that emotional and behavioral problems reflect traumatic stress—rather than confound the relationship between peer relationship quality and traumatic stress—we conducted sensitivity analyses that omitted this covariate. All TVEMs were fit using the %WeightedTVEM SAS macro, available at methodology.psu.edu. The %WeightedTVEM macro accommodates survey weights and clustering to accounting for the NSCAW complex study design.

RESULTS

Table 1 presents descriptive statistics for time-invariant variables, and Table 2 presents descriptive statistics for time-varying variables stratified by age. The analytic sample was 50% female, the most prevalent maltreatment type was neglect (55%), and the most prevalent family risk factors were poor parenting skills and low social support for parents (30%). Regarding mean trends in time-varying variables, traumatic stress and depressive symptoms decreased, peer relationship quality strengthened, and mental health service utilization increased over time.

Table 1.

Descriptive statistics for time-invariant variables of the analytic sample from the National Survey of Child and Adolescent Wellbeing (NSCAW I) at baseline.

Variable	% or $\overline{x}$ (se)
Biological sex
Male	50%
Female	50%
Initial maltreatment type
Neglect	55%
Physical abuse	36%
Sexual abuse	13%
Emotional abuse	12%
Other	11%
Family risk assessment at baseline
Poor parenting skills	30%
Low social support	30%
History of domestic violence	27%
Mental/emotional health problems	15%
Intellectual/cognitive impairments	7%
Alcohol abuse	7%
Drug abuse	7%
Age at baseline	8.4 (0.1)
N	2392

Note. Reported descriptive statistics account for complex survey design

Table 2.

Descriptive statistics (% or $\overline{x}$ [se]) for time-variant variables of the analytic sample from the National Survey of Child and Adolescent Wellbeing (NSCAW I), stratified by age.

					Age (years)
	8	9	10	11	12	13	14	15	16	17
Traumatic stress	51.1 (0.6)	50.9 (0.8)	49.5 (0.9)	47.6 (0.7)	47.4 (0.9)	48.8 (0.7)	47.4 (0.8)	47.6 (0.6)	48.0 (1.0)	45.4 (1.2)
Peer relationship quality	62.1 (0.9)	64.6 (0.8)	65.3 (1.0)	66.3 (0.7)	66.9 (0.8)	68.0 (0.8)	68.0 (1.1)	68.1 (0.9)	69.7 (1.0)	72.1 (0.8)
Depressive symptoms	10.0 (0.6)	9.0 (0.5)	8.3 (0.6)	7.1 (0.3)	8.1 (0.5)	8.5 (0.5)	8.6 (0.7)	8.5 (0.5)	7.3 (0.6)	6.7 (0.9)
Financial assistance	67%	60%	58%	59%	56%	56%	63%	57%	59%	65%
In-home placement	94%	92%	90%	93%	93%	91%	91%	88%	88%	92%
Mental health service receipt	12%	13%	11%	17%	19%	16%	21%	18%	14%	26%
Emotional/behavioral problems	39%	39%	42%	42%	48%	35%	46%	36%	43%	38%
n	669	608	540	680	624	615	626	501	349	178

Note. Reported descriptive statistics account for complex survey design. For Traumatic stress, a cut-off Traumatic Symptom Checklist for Children (TSCC) T-score of 65 represents clinical significance (Briere, 1996). Depressive symptoms according to the Child Depression Inventory with score around 17–19 have been suggested to represent clinical depression (Reynolds, 1998). For Emotional/behavioral problems, Child Behavior Checklist (CBCL)(Achenbach, 1991) or Teacher Report Form (TRF)(Achenbach & Rescorla, 2001) T-scores > 64 indicate problems.

Figure 1 presents TVEM results between peer relationship quality and traumatic stress symptoms. The solid black curve in the figure corresponds to coefficient estimates of the association between peer relationship quality and traumatic stress across time (i.e., age), and shaded regions correspond to the 95% CI of the coefficient estimates across time. CI’s not containing 0 (i.e., the dashed line) indicate an association between peer relationship quality and traumatic stress at a particular age. Shaded areas that fall entirely below the dashed line indicate age ranges where there is a negative association between peer relationship quality and traumatic stress symptoms (i.e., ages when peer relationships protect against traumatic stress).

Figure 1.

Plot of the association between peer relationship quality and traumatic stress symptom scores from age 8 to 17 years and 95% CI of the association, controlling for potential confounders.

The unique association between peer relationship quality and traumatic stress symptoms (TSSC T-score) was negative and significant from ages 8 to 8.5 years, and again from ages 9.4 to 10.9 years and at age 16.4 to 16.8 years, with maximum associations of −1.45 T-score points at age 8.5 years (95% CI = [−2.87, −0.40]), −1.57 at age 9.4 years (95% CI= [−3.13,−0.01]), and −11.89 at 16.7 years (95% CI= [−3.70,−0.09]). Taking the coefficient estimate for age 9.4 years as an example, this would mean that every additional one standard deviation increase in peer relationship quality is associated with a 1.45 point reduction in a child’s TSSC T-score. To put this in perspective, a T-score of 65 on the TSSC scale corresponds to clinical significance. Around 6% of 9-year olds in the analytic sample fell within +/− 1 points of a PTSD diagnosis. For this 6% of children, a one standard deviation improvement in their peer relationship quality could represent the difference between displaying and not displaying clinical PTSD.

The only covariates that had significant time-invariant effects on traumatic stress symptoms included being male (B = 1.70, s.e. = 0.52) and exhibiting depressive symptoms (B = 4.55, s.e. = 0.37). No other covariates demonstrated significance in the relationship between peer relationship quality and traumatic stress symptoms (Supplemental Table 2). We found the same pattern of significance in our sensitivity analyses that omitted emotional and behavioral problems as a covariate and analyses that controlled for child race/ethnicity instead of financial assistance (results not shown).

DISCUSSION

In this large, nationally representative, longitudinal sample of children with child welfare involvement, we found that quality peer relationships provided protection against traumatic stress symptoms at specific times during the transition into—but not consistently throughout—adolescence. Although researchers have extensively studied social support and mental health in adult populations, this study examines these constructs much earlier in life.

Because the adolescent period is typically characterized by movement towards independence and an increasing reliance on friends (Bokhorst et al., 2010; DuBois & Silverthorn, 2005), we hypothesized that in the context of trauma exposure, peer relationship quality would become more critical to reducing traumatic stress symptoms throughout the transition from childhood to adolescence. The evidence from the present study, however, suggests a more nuanced interpretation as peer relationships associate with mental health outcome at various periods in childhood and adolescence, but not necessarily when youth transition from childhood to adolescence. This finding is notable because it helps us understand the role of peers in protecting against the negative effects of trauma during a formative developmental phase.

One potential explanation for the non-association between peer relationship quality and traumatic stress during ages 11 to 16 years is that relationships during these years may serve different roles for individuals who have experienced early trauma compared to those who have not, thus expanding upon existing developmental theories of friendships (Bokhorst et al., 2010; DuBois & Silverthorn, 2005). While peers may serve as a source of social support for nontraumatized adolescents (Helsen et al., 2000; Stanton-Salazar & Spina, 2005), they may not fulfill this same role for adolescents who have experienced trauma. This could be because emotional support is critical for youth who have experienced trauma, and adolescents may not always be prepared to offer this to friends as their own emotional coping skills are still developing. Our findings are consistent with Helsen and colleagues’, who found that peer relationships improved mental health outcomes for adolescents with high parental support, but not for adolescents with low parental support (Helsen et al., 2000). Even if a traumatized adolescent has high quality peer relationships, this does not guarantee that the individual will receive needed support from their peers, as the degree of trust in these relationships may be insufficient and/or strong affiliations may be with deviant peers. Moreover, social stressors during the ages between 11–16 years are high (Lohman & Jarvis, 2000; Reres, 1980); these external demands on youth might limit the potential benefits of social support.

A strong foundation for peer relationships is often provided by a secure attachment to parents (Shulman, Elicker, & Sroufe, 1994; Sroufe, 2005; Sroufe & Fleeson, 1986). Without secure parental-attachment, even high-quality peer relationships may be unable to perform social support functions (i.e., emotional, instrumental, appraisal, and informational support) during certain developmental stages. In the case of early-life trauma, it is often true that the individuals that should provide a child with secure, trusting relationships (e.g., parents) are the ones that create trauma for the child, undermining the functions of early caregiving relationships to protect the child from harm and serve as a source of comfort when stressed. Mid-adolescence (i.e., ages 11–16 years) is a period of increased autonomy, exploration, and decision-making, which can result in inter- and intra-personal conflict. Without a strong foundation for relationships that can emerge from secure parental attachment, peer relationships may come under significant strain during typical adolescent development and the developmental transitions in friendships from childhood to adolescence. Such strain limits the ability for the peer-to-peer relationship to function as a restorative relational context or lend needed social support to an adolescent who has experienced trauma and thereby reduce traumatic stress symptoms.

Additionally, adolescents who have experienced trauma may develop strong relationships with deviant peers (rather than prosocial peers) due to impaired social relationship skills stemming from poor parental attachment (Akers & Lee, 1996; Elliott & Menard, 1996; Maschi et al., 2008; Warr & Stafford, 1991). Relationships with deviant peers may result in short-term coping strategies (e.g., substance use) that do not improve long-term well-being (Brook et al., 2011; Svensson, 2003). Given that mid-adolescence is already a time of exploring risk-taking behaviors, external pressures to engage in these short-term coping mechanisms may be particularly strong during this time.

Although we did not find that peer relationship quality protected against traumatic stress symptoms during the entirety of adolescence, peer relationships did protect against traumatic stress at specific intervals during the period from childhood into adolescence. Our results suggest that the role that quality peer relationships play across this transitional period exhibits developmental specificity. This may have implications for the role of peer relationships as a protective factor among adolescents who have experienced trauma. Youth may experience less peer relationship conflict during certain developmental stages, thereby allowing peer relationships to serve as effective sources of social support. For example, relationships of pre- and early-adolescents (i.e., 8 to 11 years) may be broader and mostly characterized by shared activities (e.g., sports, school-based interests)(Newcomb & Bagwell, 1995; Tesch, 1983), whereas those of mid-adolescents may be characterized by internal conflict regarding changes in social identity and roles (Claes, 1992), which may be exacerbated by pubertal onset (Laursen & Collins, 1994). In contrast, older adolescents (i.e., 16 years), although they face stressors regarding imminent transitions to adulthood, may be more competent in attaining instrumental and emotional support from friends or romantic partners and have more sophisticated emotional coping strategies that during early adolescence and late childhood.

Our study makes several unique and significant contributions to developmental science. First, this study is one of the first to evaluate the association between high quality peer relationships and traumatic stress across a broad age range spanning childhood and adolescence. Researchers have rarely evaluated age-related changes in the social support and their association with mental health outcomes. Second, we used a large, unique dataset that included children who had experienced trauma and assessments of multiple covariates that could otherwise explain the association between quality peer relationships and traumatic stress symptoms. Third, we investigated associations longitudinally across four waves representing 10 years of development, and used TVEM techniques. This allowed us to identify specific developmental periods when quality peer relationships were protective against stress symptoms; without this nuance, we may have erroneously concluded that the overall association between peer relationships and traumatic stress pooled across all ages was not significant.

Our results suggest that interventions post-trauma to improve peer relationships may be beneficial. Although developing specific interventions for improving peer relationships among youth is not without its challenges—especially for children who have been involved in the child-welfare system—using social and emotional learning as a framework for school-based youth development exhibits promise (Weissberg & O’Brien, 2004). For example, researchers have demonstrated that various different social-emotional learning curricula can result in increased prosocial behaviors and social-emotional competencies (including interpersonal relationships) among students (Ashdown & Bernard, 2012; Bierman et al., 2010; Caldarella et al., 2009; Merrell et al., 2007). Future work should consider the additional potential effects that social-emotional learning curricula may have on students’ traumatic stress symptoms.

Several study limitations also require attention. NSCAW I assessed traumatic stress symptoms with a self-reported trauma symptom checklist, and so we may have a biased report of traumatic stress in this population. Additionally, our measure of peer relationship quality did not tap into specific domains of social support that such relationships might offer. Because sample size was lower within the oldest age ranges, this may have precluded identifying significant findings for the oldest development period in our sample. Moreover, incomplete data for some respondents resulted in a reduced sample size; this loss has the potential to bias our results. An additional limitation involves the inability to disentangle multiple aspects of time—not only does time represent child development, but also it can represent time elapsed since trauma exposure. We expect that, because children entered the analytic sample at different times post-investigation (i.e., immediately, 18 months, 39 months, or 59–96 months post-investigation), our conceptualization of time represents developmental stage rather than proximity to trauma. Yet, because participants in the sample were—on average—8.4 years of age at baseline, results could also indicate that peer relationship quality serves as a protective factor in the years immediately following exposure to abuse and neglect.

Lastly, although we did apply an innovative method (i.e., Directed Acyclic Graph [DAG]) to help us control for confounding bias, additional confounders may be present. Therefore, we have provided our DAG in an effort to be transparent about the assumptions we made when interpreting the associations between quality peer relationships and traumatic stress symptoms. Our findings also suggest that future research needs to explore more deeply the roles that social relationships fulfill during different developmental stages for youth who have experienced childhood trauma.

CONCLUSION

Intervening upon child trauma has the potential to reduce later poor health outcomes. Rather than focusing on risks associated with trauma, assessing and intervening upon protective factors—such as peer relationship quality—would identify children’s strengths and opportunities to promote healthy development. Because higher-quality peer relationships protected against traumatic stress during specific time points during adolescence, trauma screening tools for these age groups could be improved by including questions regarding the quality of peer relationships. Such information could provide professionals in various child-serving settings with levers for reducing traumatic stress symptoms. In order to develop effective public health interventions to address traumatic stress, future research should study the emergence and function of peer relationships for youth who have experienced trauma, and evaluate how to establish and maintain positive peer relationships for traumatized youth of all ages.