Social cognitive mechanisms between psychological maltreatment and adolescent suicide ideation: Race/ethnicity and gender as moderators

Abstract

Childhood psychological maltreatment is a well-studied predictor of adolescent suicide ideation, while social support is a protective factor, but little is known about social-cognitive mechanisms that may link psychological maltreatment to suicide ideation. Further, given the impact that culture, race/ethnicity, and gender can have on social relationships and suicide-related risk factors, these mechanisms may differ across demographic groups.

Objective:

The current study examined whether psychological maltreatment predicts suicide ideation through self-perception of social competence, and whether this relationship differs depending on race/ethnicity and gender.

Method:

We analyzed a racially diverse, longitudinal sample of adolescents at risk for maltreatment (N = 765). Self-report measures of lifetime maltreatment were completed at age 12 and combined with data from Child Protective Services. Youth also completed measures of perceived social competence at age 12. Indicators of suicide ideation were taken at ages 8, 12, 16, and 18.

Results:

Perceived social competence scores differed significantly between children who were psychologically maltreated and those who were not, but these differences were nonsignificant for physical and sexual abuse. Self-perception of social competence fully mediated the relationship between childhood psychological maltreatment and suicide ideation for White girls and boys.

Conclusions:

Our findings suggest that poor perceived social competence is uniquely associated with psychological maltreatment, and White adolescents may develop suicide ideation through specific mechanisms involving social cognition. Certain youth may benefit from interventions improving social cognitions and promoting healthy relationships to prevent suicide ideation during adolescence.

In 2019, roughly one in five American high school students reported having seriously considered suicide (Ivey-Stephenson et al., 2020). Suicide ideation (SI) is strongly associated with suicide attempts; about one third of adolescents reporting SI attempt suicide (Nock et al., 2013). Given the high prevalence of SI and its strong association with attempts, understanding why and how it develops in children and adolescents is a crucial preventative measure against suicide attempts and deaths. SI, even in the absence of suicidal behavior, is painful and distressing (Jobes & Joiner, 2019). A deeper understanding of long-term predictors of SI through a developmental lens (see Oppenheimer et al., 2022) may inform more effective interventions through tailored, mechanism-based approaches.

Childhood maltreatment is a robust predictor of SI (Barbosa et al., 2014), and evidence shows that childhood physical, sexual, and psychological maltreatment (PM) are independently associated with adolescent SI (Miller et al., 2013). However, existing suicide research has focused primarily on physical and/or sexual abuse. A 2013 meta-analytic review on the association between childhood maltreatment and SI and attempts found that 52 studies examined sexual abuse and 34 studies examined physical abuse; only eight studies examined PM (Miller et al., 2013). The lack of research on PM is concerning, considering it is more common and often co-occurs with physical and sexual abuse (Finkelhor et al., 2014; Merrick et al., 2018).

A poor understanding of how social cognitive mechanisms relate to SI presents another gap in the suicide literature. Studies on social factors and SI focus on constructs such as relationship quality or social support, rather than how social relationships impact cognition and self-perception and vice versa. The first goal of the current study is to examine the role of social competence in the relationship between childhood maltreatment and adolescent SI, with a specific focus on PM. Our second goal is to explore whether the relationships between PM, social cognition, and SI vary by gender and race/ethnicity. Cultural and gender norms significantly impact social relationships, meanings of suicide, and risk and protective factors for suicide (Chu et al., 2010), and SI rates vary across demographic groups (Ivey-Stephenson et al., 2020). It is possible that risk for SI is conferred through differential pathways for different populations.

Psychological Maltreatment, Self-Perception of Social Competence, and Suicide Ideation

The nomenclature of maltreatment and abuse are often used interchangeably, though they measure different constructs. Briefly, maltreatment can refer to both abuse and neglect, while psychological maltreatment encompasses emotional or verbal forms of maltreatment. For consistency, we use the term “psychological maltreatment” (PM) throughout this paper, as psychological abuse and neglect are frequently measured together. Physical and sexual abuse are referred to as “abuse” where neglect is not measured.

PM may uniquely contribute to SI in adolescence by negatively impacting children’s developing social cognition. In a psychologically abusive or neglectful relationship, children may internalize parents’ negative perceptions of them. This may lead to the formation of maladaptive relational schemas that impact social functioning and heighten baseline suicide risk, as conceptualized when considering both social cognitive theories of the relational self (Andersen & Chen, 2002; Chen et al., 2011) and the fluid vulnerability model of suicide (Rudd, 2006). While PM and social cognition can also be linked to depression and other negative mental health outcomes through similar pathways (e.g., Marroquín, 2011), there has been less research on this topic within the suicide literature. Therefore, we examined how these factors relate specifically to adolescent SI. We present our theoretical framework in detail below.

PM is the most prevalent form of child maltreatment in the United States (Finkelhor et al., 2014). PM may occur on its own, or simultaneously with other forms of maltreatment. Compared to physical or sexual abuse, PM is as good a predictor or an even better predictor of mental health problems, including SI (Barbosa et al., 2014), yet it remains understudied in the suicide literature. Further, PM may confer risk for poor mental health and SI through unique pathways compared to other forms of childhood adversity. For example, PM (but not physical or sexual abuse) appears to be uniquely associated with maladaptive responses such as rumination and negative urgency (O’Mahen et al., 2015; Valderrama & Miranda, 2017), which are, in turn, associated with SI (Rogers & Joiner, 2017). This suggests that PM may have unique effects on individuals’ cognitive and affective functioning that other forms of maltreatment may not. We suggest social competence may be another mechanism that mediates the relationship between childhood PM and SI.

The mechanisms associating PM with SI through social competence may be explained by social cognitive theories of the relational self. According to these theories, consistent interactions with significant others can shape long-term beliefs about the self and others (Chen et al., 2011). The parent-child relationship has influential ramifications throughout the lifespan (Doyle & Cicchetti, 2017). Childhood PM may contribute to negative relational schemas—e.g., mental representations of relationships and the self in relation to others (Andersen & Chen, 2002). For example, psychologically maltreated youth may internalize their parents’ vocalized or implied negative perceptions of the youth, thereby developing persistent, global, negative beliefs about themselves. The effects of PM on relational schemas may be critical during adolescence because peer relationships become salient at this developmental stage (Brown & Larson, 2009). A negative relational schema formed due to parental PM in childhood may result in perceptions of or actual low social competence among peers in adolescence, ultimately heightening SI risk.

Maladaptive social cognition resulting from childhood PM can heighten baseline suicide risk, as described in the fluid vulnerability theory (Rudd, 2006). This cognitive model of suicide posits that suicidal episodes are time-limited and that stable, baseline risk factors interact with acute, dynamic risk factors to activate the suicidal mode (see also Beck, 1996). Maladaptive social cognition is a baseline risk factor that may be particularly salient for adolescents, considering the increasing importance of peer relationships during this developmental time period. Perception of low social competence and negative relational schemas can become associated with an adolescent’s suicidal belief system, with such associations growing stronger over time as the suicidal mode is repeatedly activated by these relational schemas. Indeed, social support is a well-studied protective factor against SI (Miller et al., 2015), and low social competence is significantly associated with SI among children and adolescents (King et al., 2001). We theorize that psychologically maltreated adolescents develop stable, long-lasting negative views about themselves and their relationships, which contribute to baseline vulnerability to SI.

Gender and Race/Ethnicity as Moderators

Disparities in childhood maltreatment and SI exist among gender and racial/ethnic divides. In the United States, Black, Hispanic, and multiracial children are at disproportionate risk of experiencing adverse childhood experiences (Merrick et al., 2018). Such experiences can also confer SI risk differentially to different populations; for example, among Puerto Rican young adults, adverse childhood experiences are associated with higher SI rates among young women only, not young men (Polanco-Roman et al., 2021). Overall, rates of SI and attempts among adolescents differ by gender and race/ethnicity. In 2019, a lower proportion of Black students seriously considered suicide compared to White students, but Black students attempted suicide at higher rates than White and Hispanic/Latinx students. High school girls were significantly more likely than boys to seriously consider and attempt suicide (Ivey-Stephenson et al., 2020). Different racial groups maintain different SI and suicide attempt (SA) risk trajectories from adolescence to young adulthood (Erausquin et al., 2019), suggesting unique underlying mechanisms across populations.

Social factors might differentially affect SI for boys and girls and for youth of different racial and ethnic groups. In a 2006 study, adolescent girls hospitalized for a suicide attempt or ideation perceived more peer support than did boys. Family support protected against hopelessness, depressive symptoms, and SI for girls, but not boys. Moreover, peer support was associated with greater depressive symptoms and SI, but only for boys (Kerr et al., 2006). For adolescent girls but not boys, social interactions (specifically venting) were protective against SI (Kim et al., 2014). Race/ethnicity also plays a role in how social factors impact SI, attempts, and deaths. For example, different racial/ethnic groups have unique risks and protective factors in the antecedents of suicide death (Lee & Wong, 2020). Further, acculturation and parent-child conflict are associated with suicidal behaviors among Asian American youth (Lau et al., 2002), while community- and church-based support are protective factors against suicide attempts among Black American adults (Compton et al., 2005). Black American college students have also reported greater endorsement of moral objections to suicide and motivation to survive and cope than White American college students, which may serve as another protective factor (Morrison & Downey, 2000). Further, SI and attempts may be associated with different mechanisms—such as relational schemas—for different racial/ethnic and gender groups.

According to the cultural theory of suicide, (1) culture affects which kinds of stressors lead to suicidal thoughts and behaviors; (2) cultural meanings around suicide can impact how SI develops; and (3) culture impacts how those thoughts and behaviors are expressed (Chu et al., 2010). In other words, race, ethnicity, religion, and national background shape cultural beliefs and norms around suicide. For example, Black students were less likely than their White peers to attribute causes of suicide to personal or interpersonal issues, but rather to systemic stressors rather than personal psychological factors (Joe, 2006). Gender socialization, or the process of developing beliefs about the roles, expectations, and behaviors typically associated with different sex and gender groups (Stockard, 2006), is another factor that may also impact population-specific stressors and cultural meanings related to suicide. For example, masculine norms (e.g., toughness, stigmatization of emotion expression, tolerance for aggression) contribute to poor emotion regulation and disproportionate rates of externalizing disorders among men and boys, compared to women and girls (Berke et al., 2018). Thus, race/ethnicity and gender socialization may moderate how adolescents interpret problems in social relationships and whether they develop SI.

The Current Study

The current study draws from a large, nationally representative, longitudinal dataset to investigate how childhood PM is related to SI in adolescence through a social cognitive mechanism, and how these relationships vary by gender and race. Our aims for this study were to:

1) Test the associations between PM, physical abuse, and sexual abuse and a social cognitive risk factor. We predicted perceived social competence (SC) would be significantly lower among youth who experienced PM (vs. those who did not) but would not significantly differ due to physical or sexual abuse.

2) Test the indirect effects of PM on SI via perception of social competence. We hypothesized that PM would predict lower levels of SC, which would, in turn, predict SI at ages 12, 16, or 18, adjusting for prior SI, race, and gender.

3) Explore the mediating effect of SC in the relation between PM and SI and the moderating effects of race and gender.

Methods

Participants

Children enrolled in the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN) consortium were born between 1986 and 1994 and were recruited from Baltimore, Chicago, San Diego, Seattle, and a statewide site in North Carolina. Each site had different inclusion criteria for recruitment (see Runyan et al., 1998). Of the 1354 children recruited for the original study, 1074 (79% retention) participated at age 8, 895 (66% retention) participated at age 12, 831 (61% retention) participated at age 16, and 919 (68% retention) participated at age 18.

The sample for our analyses included 765 children (51% female) who completed measures of SI (at ages 8, 12, 16, and 18), PM, and SC. Racial/ethnic distribution (reported by the caregiver at the first interview) was 55% non-Hispanic Black, 27% non-Hispanic White, 5% Hispanic/Latinx, 11% mixed race, and 1% other, and distribution by site was as follows: 21% Baltimore, 14% Chicago, 25% San Diego, 21% Seattle, and 19% North Carolina. Note that the LONGSCAN only included two options for gender.

Procedure

The study was initiated in 1991 and completed data collection in 2011. Children completed self-report measures of current SC and lifetime physical, sexual, and PM at age 12. Additionally, Child Protective Services (CPS) records were reviewed for allegations of maltreatment either on a continuous basis (San Diego and Seattle) or on the same year as a face-to-face visit (Baltimore, Chicago, and North Carolina).

Each project site obtained approval for the study procedures through local Institutional Review Boards (IRBs). Each site also had a Federal Certificate of Confidentiality, which ensured that most research data remained confidential, excluding any suspected maltreatment. Depending on state law, alleged maltreatment was subject to be reported to CPS. Informed consent was obtained from the children’s caregivers. Children provided assent for participation in interviews from ages 8 to 16 and informed consent for the interview at age 18. This secondary data analysis was exempt from review by the Hunter College Human Research Protection Program.

Measures

Childhood maltreatment.

At age 12, maltreatment was measured by CPS file reviews and self-report. Administrative reviews of CPS files used the Modified Maltreatment Classification System (MMCS) (English et al., 2005). A team of LONGSCAN abstractors reviewed CPS files and coded each allegation if it met the definition of each type of maltreatment, based on the MMCS definitions, with good interrater reliabilities for physical (κ = .84), sexual (κ = .71), and PM (κ = .72) (Everson et al., 2008). Allegations were coded as substantiated physical abuse when a caregiver or other responsible adult intentionally caused physical injury to a child. Sexual abuse was defined as attempted or actual sexual contact between an adult (usually a caregiver or household member) and a child, or sexual exploitation of the child. PM involved constant or severe thwarting of children’s emotional needs by harming a child’s psychological safety and security, acceptance and self-esteem, and age-appropriate autonomy.

The present study also incorporated self-report measures of maltreatment developed by the LONGSCAN consortium. The Self Report of Physical Abuse and Assault (see Runyan et al., 2014, for details) is a measure, based on Barnett et al. (1993), in which children reported on physical abuse by adults within and outside the household. The measure includes 15 stem items, such as “Has any adult ever kicked or punched you?” and “Has any adult ever done something else that badly physically hurt you or put you in danger of being hurt?” Children completed the Self Report of Sexual Abuse and Assault (Runyan et al., 2014), based on Barnett et al. (1993), assessing lifetime history of sexual abuse at age 12. The measure includes 11 stem items, such as “Has anyone ever touched your private parts or bottom in some way?” and “Has anyone ever forced you to look at their sexual parts?” Youth completed the Self Report of Psychological Maltreatment (Runyan et al., 2014), which is based on definitions of maltreatment by Barnett et al. (1993) and Hart and Brassard (1996) and assesses lifetime history of PM and neglect by a parental figure at age 12. The measure includes 18 stem items, such as “Have any of your parents ever threatened to hurt you badly?” and “Have any of your parents ever called you names/teased you that made you feel really bad about yourself?” Although information on validity and reliability is not available for these measures, which were developed for the LONGSCAN, a small sample of outpatient adolescents was tested to verify comprehension of the self-report questions (Runyan et al., 2014).

To accurately represent the absence or presence of maltreatment for each child, we created a dichotomous variable in which a child was coded for having experienced psychological, physical, or sexual maltreatment if they endorsed at least one item of the respective self-report, had at least one coded substantiation of the specific type of maltreatment from the MMCS, or both. We integrated CPS and self-report data based on recommendations by prior researchers working with the LONGSCAN dataset (Everson et al., 2008) after considering the substantive concerns with inaccuracy, bias, and lack of concordance for both CPS and self-report data. Dichotomizing the variable was necessary, as the two sets of data were incompatible in terms of types of questions asked and the scoring systems used.

Perception of social competence.

At age 12, an interviewer assessed children’s perception of their own social competence using the Social Competence scale of the Youth Self Report (YSR) (Achenbach, 1991). Children compared their participation frequency and performance of hobbies, games, sports, jobs, chores, friendship, and activities to others their age, which were coded on a 3-point Likert scale, except for grades, which was coded on a 4-point Likert scale. Examples of questions include, “Compared to others your age, about how much time do you spend in this sport?” and “Compared to others your age, how well do you get along with other kids?” The total raw score was converted to a t-score standardized by age and gender (see YSR manual; Achenbach, 1991). Test-retest reliability over seven days for children ages 11 to 14 years is .68 for the total social competence scores (Achenbach, 1991). The YSR has demonstrated acceptable criterion-related validity and content validity, such that the total quantitative scores and most of the individual YSR scores could be used to successfully discriminate between referred at-risk youth and non-referred, demographically matched youth (Achenbach, 1991).

Suicide ideation.

At ages 8 and 12, one item of the Trauma Symptom Checklist for Children (TSCC) assessed SI (Briere, 1996). Possible responses to an item inquiring how often youth wanted to kill themselves include “never,” “sometimes,” “lots of times,” or “almost all of the time.” While psychometric information is currently unavailable for the SI item, the TSCC demonstrates good concurrent validity with similar measures, such as the YSR (Achenbach, 1991), and children with histories of abuse have generally demonstrated high TSCC scores (Knight et al., 2008). For our SI variables at ages 8 and 12, we created a dichotomous variable that reflected the presence of any SI (at least “sometimes”). At ages 16 and 18, two items from the Youth Risk Behavior Surveillance System (YRBSS) (Brener et al., 2004), assessed SI (Knight et al., 2008). Possible responses were “yes” or “no” to the following questions: “During the last 12 months, did you ever seriously consider attempting suicide?” and “During the last 12 months, did you make a plan about how you would attempt suicide?” We created a dichotomous variable that classified children that considered and/or planned suicide in the last 12 months at ages 12, 16, or 18 as having experienced SI in the previous 12 months. We computed one dichotomous SI variable for ages 12, 16, and 18, due to the low base rates at each of these intervals. Further, given that SI during childhood and adolescence is a serious cause for concern, we were interested in its occurrence at any of these ages due to the implications for future psychological distress. If children reported SI at these times, they were coded as yes (1). We coded missing values as missing, rather than as 0, as not to underestimate SI prevalence. In a sample of 1,679 7^th-12^th graders in the U.S., most YRBSS question items demonstrated acceptable to high test-retest reliability over fourteen days (Brener et al., 2004). A later study found the two SI items from the YRBSS to have good convergent validity (r = .57–.71) with suicide-related items from other self-report measures (e.g., Patient Health Questionnaire) and good discriminant validity with measures of non-suicide-related measures (e.g., drug use) among a sample of 386 high school students (May & Klonsky, 2011).

Analytic Plan

We used a dichotomous variable for gender (as the LONGSCAN measured gender dichotomously), with boys as the reference group. The sample was predominantly composed of Black and White children (82%), so the remaining race/ethnicity groups, which only made up between 1% to 11% of the total sample, were coded as “Other” for the current analyses. Black children were the reference group. Eight hundred forty-six children reported on social competence. Of these, 81 children (10%) were excluded due to missing data on abuse history, SI at age 8, or SI at all follow-up periods. Chi-square tests of independence revealed no systematic difference between children with missing data and those with complete data on any of the variables of interest. Therefore, we used a complete-case analysis for the remaining analyses.

We conducted chi-square tests of independence to examine whether children who reported abuse were more likely than chance to report SI at age 8, and at ages 12, 16, or 18 (Table 1). Independent samples t-tests were conducted to identify differences in perceived social competence by abuse type (aim 1). For the mediation and moderated mediation analyses, we combined reported SI at follow-up periods, such that the outcome variable was a dichotomous measure of SI. To test hypothesis 2 that PM would predict SI at age 12, 16, or 18 via lower perceived SC, we used Model 4 of the PROCESS Macro for SPSS (Hayes, 2017) with 95% bootstrapped confidence intervals and 5000 resamples. Hypothesis 3, that this relationship would be moderated by race/ethnicity and gender, was tested using moderated mediation regression analyses. We also explored the effects of physical abuse and sexual abuse on SI through social competence, moderated by gender and race (Table 2). These exploratory analyses were conducted to ascertain whether psychological abuse was a unique predictor of SI through social competence and the moderators, compared to physical and sexual abuse. We tested this hypothesis using Model 17 of PROCESS. SC was centered around its mean before computing interaction terms.

Table 1.

Proportion of children reporting suicide ideation by maltreatment type, gender, and race/ethnicity

	SI age 8			SI age 12 through 18
	n (%) [Z]	χ2(1)	p	n (%) [Z]	χ2(1)	p
Maltreatment
Psychological	43 (12) [2.0]	3.57	.06	57 (16) [1.9]	3.16	.08
Physical	30 (13) [2.2]	4.22	.04	47 (21) [3.9]	14.39	.001
Sexual	21 (15%) [2.3]	4.65	.03	33 (23) [3.9]	13.86	.001
Gender
Girls	39 (10) [0.4]	0.07	.79	65 (17) [2.8]	7.42	.01
Boys	34 (9) [−0.4]			36 (10) [−2.8]
Race/ethnicity
Black	30 (7) [−2.7]	7.20 (2)	.03	47 (11) [−2.0]	4.96 (2)	.08
White	25 (13) [1.7]			35 (18) [2.1]
Other	18 (13) [1.6]			19 (14) [0.3]

Note: values for type of maltreatment and gender were calculated using Yates’ Correction.

Table 2.

Moderated Mediation Model of Psychological, Physical, and Sexual Abuse Predicting Suicide Ideation at Ages 12, 16, or 18

	Model 1 (Psychological Maltreatment)				Model 2 (Physical Abuse)				Model 3 (Sexual Abuse)
	Log Odds	95% CI (low)	95% CI (high)	Odds Ratio	Log Odds	95% CI (low)	95% CI (high)	Odds Ratio	Log Odds	95% CI (low)	95% CI (high)	Odds Ratio
Psych Maltreatment	0.04	−0.79	0.87	1.04	-	-	-	-	-	-	-	-
Physical abuse	-	-	-	-	1.05	0.20	1.91	2.86	-	-	-	-
Sexual abuse	-	-	-	-	-	-	-	-	0.53	−0.41	1.46	1.70
Perceived SC	0.02	−0.02	0.07	1.02	0.02	−0.02	0.07	1.02	0.02	−0.02	0.07	1.02
Gender	1.10	−1.25	3.44	3.00	1.48	−0.87	3.83	4.39	1.03	−1.27	3.33	2.80
Gender × PM	0.03	−0.86	0.92	1.03	-	-	-	-	-	-	-	-
Gender × Physical abuse	-	-	-	-	−0.78	−1.69	0.14	2.18	-	-	-	-
Gender × Sexual abuse	-	-	-	-	-	-	-	-	−0.12	−1.12	0.88	1.13
Gender × Perceived SC	−0.01	−0.06	0.04	1.01	−0.01	−0.06	0.04	1.01	−0.01	−0.06	0.04	1.01
White	3.57	0.89	6.24	35.52	3.96	1.37	6.56	52.46	3.88	1.28	6.48	48.42
Other	1.41	−1.61	4.44	4.10	1.25	−1.85	4.35	3.49	1.48	−1.49	4.45	4.39
PM × White	0.60	−0.41	1.61	1.82	-	-	-	-	-	-	-	-
PM × Other	0.14	−1.03	1.31	1.15	-	-	-	-	-	-	-	-
Physical abuse × White	-	-	-	-	0.31	−0.71	1.32	1.36	-	-	-	-
Physical abuse × Other	-	-	-	-	0.35	−0.84	1.54	1.42	-	-	-	-
Sexual abuse × White	-	-	-	-	-	-	-	-	0.79	−0.34	1.92	2.20
Sexual abuse × Other	-	-	-	-	-	-	-	-	0.40	−0.87	1.67	1.49
Perceived SC × White	−0.08	−0.13	−0.02	1.08	−0.08	−0.14	−0.03	1.08	−0.08	−0.14	−0.03	1.08
Perceived SC × Other	−0.03	−0.09	0.03	1.03	−0.03	−0.09	0.04	1.03	−0.03	−0.10	0.03	1.03
SI age 8	0.50	−0.15	1.14	1.65	0.48	−0.18	1.13	1.62	0.42	−0.23	1.08	1.52
Site	0.10	−0.06	0.26	1.11	0.08	−0.09	0.24	1.08	0.08	−0.08	0.24	1.08
Indirect effect of psychological, physical, and sexual abuse on suicide ideation via perceived social competence moderated by gender and race/ethnicity
	Log Odds	95% CI (low)	95% CI (high)		Log Odds	95% CI (low)	95% CI (high)		Log Odds	95% CI (low)	95% CI (high)
Boys × Black	−0.03	−0.13	0.04		−0.02	−0.09	0.03		−0.02	−0.11	0.04
Boys × White	0.08	0.001	0.22		0.05	−0.05	0.18		0.06	−0.05	0.22
Boys × Other	0.01	−0.07	0.11		0.01	−0.05	0.08		0.01	−0.07	0.10
Girls × Black	−0.02	−0.09	0.05		−0.01	−0.06	0.03		−0.01	−0.08	0.03
Girls × White	0.10	0.01	0.25		0.06	−0.06	0.20		0.07	−0.06	0.25
Girls × Other	0.03	−0.06	0.16		0.01	−0.04	0.11		0.02	−0.07	0.13
Indices of Partial Moderated Mediation
	Index	SE	95% CI (low)	95% CI (high)	Index	SE	95% CI (low)	95% CI (high)	Index	SE	95% CI (low)	95% CI (high)
Gender	0.02	0.04	−0.06	0.11	0.01	0.03	−0.05	0.08	0.01	0.03	−0.05	0.10
White	0.12	0.07	0.01	0.29	0.07	0.07	−0.06	0.22	0.08	0.09	−0.06	0.27
Other	0.04	0.06	−0.05	0.18	0.02	0.04	−0.04	0.13	0.03	0.05	−0.06	0.16

Note: PM = Psychological Maltreatment; SC = Social Comeptence; Boys are the reference group for gender, and Black is the reference group for race/ethnicity. Bold = statistically significant at p < .05. Model 1 includes Psychological Maltreatment as a predictor, Model 2 includes Physical Abuse, and Model 3 includes Sexual Abuse.

Given that the current study is a secondary data analysis, an a priori power analysis was not conducted. Furthermore, there is limited information among the preexisting literature in calculating power for complex mediation models, particularly for the moderated mediation logistic regression model in the current study (Thoemmes et al., 2010). Fritz and MacKinnon (2007) have laid out sample size guidelines for conducting mediation analyses with sufficient power (i.e., 0.8). Based on their criteria, our sample size of 765 children exceeds their estimates for achieving 0.8 (80%) power in detecting at least a small mediation effect. Notably, these guidelines may not consider additional covariates and moderated mediation analyses. The authors also emphasize that these sample size estimates should be used as a lower limit of the number of participants required to reach 80% power (Fritz & MacKinnon, 2007).

Results

Preliminary analyses

Of the 765 children in the present sample, 365 (48%) experienced PM at age 12, 229 (30%) experienced physical abuse, and 143 (19%) experienced sexual abuse. One child had no data on sexual abuse from self-reports or from MMCS. Overall, 41% of children did not report maltreatment and were not determined to be maltreated. Among the children who experienced physical abuse, 76% (n = 175) also experienced PM; and of the children who experienced sexual abuse, 74% (n = 106) also experienced PM.

Seventy-three (10%) children reported SI at age 8; 38 (5%) children reported SI at age 12 (20 missing values); 48 (6%) children reported SI at age 16 (202 missing values); and 38 (5%) adolescents reported SI at age 18 (157 missing values). Overall, 101 children (13%) reported SI either at age 12, 16, or 18. Thirty-two (4%) participants reported onset of SI at age 12; 32 (4%) participants reported onset at age 16; and 8 (1%) participants reported onset of SI at age 18. Further, 24 (3%) participants reported recurrent SI on at least two consecutive time points; however, no one reported SI at every time point from age 8 to age 18.

Chi-square tests of independence revealed that children who experienced any form of abuse were more likely than chance to report SI at age 8; further, children who reported physical or sexual abuse were more likely than chance to report SI at ages 12 through 18 (Table 1). There were no gender differences in physical, psychological, or sexual maltreatment; neither were there racial/ethnic differences in psychological or sexual maltreatment. However, White children were more likely than chance (Z = 2.2), and Black children were less likely than chance (Z = −3.2) to experience physical abuse, χ²(2) = 11.29, p < .01.

Further, chi-square tests revealed demographic differences in SI (Table 1). While there was no gender difference in SI reported at age 8, there was a gender difference in endorsement of SI at ages 12 through 18, with girls more likely and boys less likely to report SI than expected by chance (Z = 2.8, p < .05). There were also racial/ethnic differences in SI at age 8, with Black children being less likely than chance to report SI (Z = −2.7, p < .01). Table 1 presents the distribution of children who reported SI at 8 or ages 12 through 18 by type of maltreatment, gender, and race.

Psychological Maltreatment and Perceived Social Competence

Independent samples t-tests revealed significant group differences in SC by PM: children who experienced PM scored lower on SC (M = 45.79, SD = 9.58) than did children with no history of PM (M = 47.50, SD = 8.81), t(763) = 2.58, p = .01. However, there were no significant group differences in SC scores among children who were sexually abused (M = 45.65, SD = 9.32) compared to those who were not (M = 46.89, SD = 9.16), t(762) = 1.45, p = .15. There were also no differences between children who were physically abused (M = 45.97, SD = 9.28) vs. not (M = 46.99, SD = 9.19), t(763) = 1.40, p = .16.

Psychological Maltreatment Predicting Suicide ideation via Perceived Social Competence

We conducted mediation analyses to test our second hypothesis that history of PM would predict lower levels of SC, which would, in turn, predict SI at ages 12 through 18, adjusting for gender, race, site, and SI at age 8. The direct effect of PM on SI was not significant (b = 0.29, SE = 0.22, p = .19, 95% CI = −0.14, 0.72), when adjusting for covariates; however, greater PM predicted lower levels of SC (b = −1.43, SE = 0.67, p = .03, 95% CI = −2.74, −0.12). SC did not predict SI at follow-up (b = −0.02, SE = 0.01, p = .18, 95% CI = −0.04, 0.01), nor was the indirect effect significant (b = 0.02, SE = 0.02, 95% CI = −0.01, 0.08).

We replicated these mediation analyses using physical abuse and sexual abuse as predictors. When adjusting for covariates, physical abuse predicted future SI (b = 0.78, SE = 0.22, p < .01, 95% CI = 0.34, 1.22). However, physical abuse did not predict SC (b = −0.75, SE = 0.73, p = .30, 95% CI = −2.19, 0.68), nor was the indirect effect of physical abuse on SI through SC significant (b = 0.01, SE = 0.02, 95% CI = −0.01, 0.06). The model with sexual abuse revealed a significant effect of sexual abuse on SI (b = 0.78, SE = 0.24, p < .01, 95% CI = 0.31, 1.26) when adjusting for covariates. However, the effect of sexual abuse on SC was not significant (b = −0.84, SE = 0.85, p = .33, 95% CI = −2.51, 0.84). The indirect effect of sexual abuse on SI via SC was not significant either (b = 0.01, SE = 0.02, 95% CI = −0.02, 0.06).

Moderating Role of Gender and Race in the Indirect Effect via Perceived Social Competence

For the exploratory aim, we tested the moderating role of gender and race in the relation between SC and SI and the indirect effect of SC in the relation between PM and SI. The index of partial moderated mediation was nonsignificant for gender and youth classified as “Other” compared to Black youth; however, the index of partial moderated mediation was significant for White youth (Table 2). Notably, there was a significant interaction of gender and race in the direct relation between perceived SC and SI: White boys (b = −0.06, SE = 0.03, 95% CI = −0.11, −0.001) and White girls (b = −0.07, SE = −0.02, 95% CI = −0.11, −0.02) with lower perceived SC were more likely to report future SI. This was not the case for other racial/ethnic groups. Although the direct effect of PM on SI was not moderated by race and gender, findings revealed a significant moderated indirect effect; PM predicted subsequent SI via lower SC for White girls and boys (Table 2). We replicated the moderated mediation analyses using physical abuse and sexual abuse as predictors (see Table 2). The moderated mediation analyses were nonsignificant for these predictors. The indices of partial moderated mediation were also nonsignificant for both models.

Discussion

This study investigated the role of PM in the development of SI through the mediating pathway of perceived social competence, moderated by race/ethnicity and gender, in a large, geographically and racially diverse sample of high-risk adolescents from across the U.S. Youth who experienced PM rated their social competence significantly lower than those without PM. However, there were no group differences in social competence scores for youth who experienced physical abuse (vs. not) and sexual abuse (vs. not). These findings suggest that social cognition in early adolescence is uniquely impacted by psychological forms of maltreatment, perhaps due to a more direct influence on self-perception and relational schemas. Further, over 70% of youth in the current sample who experienced physical or sexual abuse also experienced PM. Thus, PM commonly co-occurs with other forms of maltreatment, highlighting the need to better understand how it relates to adverse outcomes such as SI.

Among the full sample, perceived social competence was not a significant mediator in the relationship between PM and SI. However, social competence fully mediated this relationship for White girls and boys. For White adolescents, PM did not directly predict SI but increased risk indirectly through lower perceived social competence. We did not find any significant indirect pathways from physical or sexual abuse to SI for any of the examined demographic subgroups, suggesting that this pathway is unique for PM. Overall, different forms of childhood maltreatment may confer risk for SI through differential pathways, and these pathways may further diverge based on race/ethnicity.

Several factors may explain why social cognitive factors are more salient to SI risk for White adolescents. Notably, we found this significant relationship despite half as many White youths in our sample as Black youths and roughly the same number of other race/ethnicity youth as White youth. It is unlikely that there was not enough power to detect effects among Black or other race/ethnicity youth. While racial and ethnic minority youth may face disproportionate external stressors relating to SI, such as minority stress, acculturation, and discrimination, they may also benefit from additional protective factors such as ethnic identity (Chu et al., 2010). White youth may also be more vulnerable to SI because of a susceptibility to maladaptive cognitions or negative perceptions about themselves and their social relationships, given White Americans are more likely than other racial groups to endorse intrapersonal perception motives for suicide (Chu et al., 2017). Further, considering Black Americans are also more likely to attribute reasons for suicide death to external factors than White Americans (Joe, 2006), the effect of PM on a child’s social competence could be stronger among White than Black children. More research is needed to draw firmer conclusions and propose alternative explanations for why PM predicts SI through perceived social competence for White youth.

Notably, we found White youth were significantly more likely to report SI than Black youth in all three maltreatment regression models. One explanation for this difference is that Black youth who experience SI may not disclose it. Considering prior research suggests ethnic minority youth were more likely to be “hidden ideators” (Morrison & Downey, 2000), a single-item assessment of SI may not adequately measure SI among Black youth.

Chi-square analyses indicated that physical and sexual abuse were directly associated with prospective SI, but PM was not. These results are inconsistent with prior studies showing PM is a robust predictor of SI and attempts, with even stronger effects than physical or sexual maltreatment (Barbosa et al., 2014). However, upon closer examination, studies are mixed regarding which form of maltreatment poses the greatest risk for suicide attempts (Zatti et al., 2017). Variation in the effect of different maltreatment subtypes may be due to methodological inconsistencies. For example, self-report and CPS records show little concordance, and concordance levels were lowest for PM (Everson et al., 2008). Disparate ways of measuring and determining the presence of maltreatment may contribute to biased recruitment methods, inaccurate identification of maltreatment cases, and poor operationalization of variables. These methodological inconsistencies may partially explain discrepancies in the direct relationship between different forms of childhood maltreatment and SI.

The lower rates of SI in adolescence than in childhood were also inconsistent with prior literature, which indicates that suicide risk increases in adolescence (Bilsen, 2018). Black children aged 5 to 12 have significantly higher rates of suicide than their White peers (Bridge et al., 2018). Considering our sample is primarily Black, this could explain the higher rates of SI at age 8 than later in adolescence. Perhaps, youth at the highest risk may have discontinued the study, resulting in lower rates of SI in adolescence.

Limitations

Major study limitations center around the operationalization of constructs. First, the measure used for perception of social competence is not commonly used. The self-report questionnaire is also a proxy for social cognition rather than a direct measure of social cognitive distortions. Thus, we could not distinguish objective difficulties in social functioning from a perception of poor social competence assessment tool. Second, SI was measured using one or two items from a larger behavioral checklist; thus, we were unable to capture the severity of SI. The single or two-item SI questions likely limited the sensitivity of the measure. Lastly, self-report measures of maltreatment could not be reliably combined with CPS data while remaining continuous, so final maltreatment variables were dichotomized, losing severity and frequency scores. Finally, while the CPS records of childhood maltreatment specify that the perpetrator must be a parent or caregiver, the self-report measures used for physical and sexual abuse do not specify this, introducing a potential confounding variable into our analyses. LONGSCAN was designed as an epidemiological study of child maltreatment, and its measures were less nuanced due to the nature of large-sample, longitudinal data collection.

The relatively narrow scope of this study can also be considered a limitation. On one hand, suicide is a transdiagnostic mental health problem, and many relevant intervening pathways were not included in this study. On the other hand, childhood maltreatment and social competence are associated with many outcomes other than suicide ideation, including many externalizing and internalizing symptoms. With this in mind, the current study improves our understanding of transdiagnostic mechanisms such as social cognition, contributing to a more fluid and dimensional model of psychopathology (e.g., Kotov et al., 2017).

Other limitations include issues relating to the study sample. First, we could not include nuanced analyses about racial/ethnic groups in the “other” category. We had to combine Hispanic/Latinx, mixed race, and other race/ethnicity (e.g., Asian, Native American) children into a single category due to their low representation in the sample. In addition, gender was measured in binary form, which does not account for additional gender identities. Another concern includes the low retention rate over the 10+ year course of the LONGSCAN study. The sample is a very high-risk group, having been pre-selected for risk of childhood maltreatment. This limitation is particularly challenging in suicide research, as the youth at highest risk may have dropped out of the study, contributing to the lower rates of SI with increasing age in our sample. Despite these limitations, LONGSCAN still contributes a large, longitudinal dataset that provides valuable insight on suicide-related outcomes in this hard-to-reach population.

Future research directions

To better understand the effect of social cognition on SI and suicide-related outcomes, future research should consider using cognition-specific measures, such as social reasoning problems. Additionally, future work should examine race/ethnicity with greater specificity, including stratifying groups rather than combining subgroups into an “Other” category. Further, skewed recruitment methods and a lack of emphasis on race/ethnicity result in a large body of knowledge about mostly White samples, limiting our understanding of variability in health-related outcomes among racial/ethnic groups. Race/ethnicity is often treated as a control variable, leading to imprecise conclusions tailored to no specific racial/ethnic group, minority or otherwise.

Prevention, Clinical, and Policy Implications

Interventions for youth with SI should not only emphasize the importance of building healthy social relationships but also prioritize helping youth correct maladaptive relational schemas. Increased understanding of social cognitive mechanisms underlying suicide risk will inform more effective interventions—especially when tailored for specific populations.