Nonsuicidal Self-Injury in Preadolescents

Given the established risk associated with NSSI, and recent evidence highlighting the elevated risk of a preadolescent onset, research is needed to thoroughly characterize the nature and scope of preadolescent engagement in NSSI at the national level. Identifying factors associated with NSSI onset and history among preadolescents is vital to improve the detection of at-risk youth and establish a reference point for future efforts to monitor progress in addressing this public health issue. The current study aimed to analyze data from the Adolescent Brain and Cognitive Development (ABCD) Study, an epidemiologic study of youth aged 9 to 10 years. One previous ABCD study estimated the prevalence of NSSI and examined family-related correlates.¹⁴  The current study expands upon these findings by evaluating sociodemographic characteristics, diagnostic factors, and suicidal ideation (SI) in relation to lifetime NSSI history in preadolescents. It further aimed to examine temporally primary diagnoses and SI in relation to subsequent NSSI onset.

Under a nondisclosure agreement data use agreement, data were drawn from the ABCD study,¹⁵  a large-scale longitudinal investigation of child health and brain development. Between 2016 to 2018, 11 875 youth aged 9 to 10 years were enrolled from 22 sites across the United States. Participants were recruited to reflect the sociodemographic composition of the US population, as put forth by the US Census Bureau’s American Community Survey. The current study used data from the Curated Annual Release 5.0, which became available June 2023 and included complete baseline data.¹⁶  Additional information regarding study design and sampling recruitment is provided elsewhere.^17–19 

All sociodemographic information was drawn from the parent-completed demographic questionnaire except sexual orientation, which was drawn from child-report. Parents provided information about their child’s race (white, Black, American Indian/Alaskan Native, Asian American, multiracial, or other) and ethnicity (Hispanic or not Hispanic). Categories for race were collapsed into white, Black, multiracial, and other race. The small number of unweighted cases of NSSI for the American Indian/Alaskan Native and Asian American racial groups required collapsing these groups into an other category to permit analysis. Parents also reported their own education level, annual family income, and marital status. Parental income categories were collapsed into 5 categories ranging from “<$25 000” to “$100 000 and greater,” parental education was collapsed into 4 categories ranging from “less than high school” to “college graduate,” and parental marital status was collapsed into “married” and other. Sexual orientation was drawn only from the child-completed background information survey and was not based on parent-report. To assess sexual orientation, youth were asked if they were gay or bisexual. Potential responses were “yes,” “maybe,” “no,” and “I do not understand this question.” Participants who responded yes or maybe were combined to create a “gay, bisexual, or questioning” group and compared with participants who responded no, forming the heterosexual group, as well as to a third group consisting of participants who did not understand the question.

A computerized version of the Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version, using Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, (DSM-5²⁰ ) criteria was used to obtain parent- and child-reports of current (past 2 weeks) and past (before the past 2 weeks) NSSI, SI, and DSM-5 psychiatric disorders.¹⁹  On the basis of concepts drawn from the Columbia Classification Algorithm of Suicide Assessment,²¹  to assess the presence versus absence of NSSI, respondents are asked, “Sometimes when kids get upset or feel numb, they may do things to hurt themselves, like scratching, cutting, or burning themselves. In the past 2 weeks, how often have you [has your child] done any of these things or other things to try to hurt yourself [himself or herself]?” If respondents endorse yes, the following question is presented to ensure only self-injurious acts without suicidal intent are recorded: “Were you [was your child] trying to kill yourself [themselves] when you [they] did these things?” These questions are also asked about lifetime. To assess the presence versus absence of any SI, a range of SI subtypes was assessed (inclusive of passive SI, nonspecific active SI, active SI with method, active SI with intent, and active SI with plan). DSM-5 psychiatric disorders assessed can be found in Table 1. Following standard clinical practice,²²  a child was coded as having NSSI, SI, or a disorder if a positive response or diagnosis resulted from either parent- or child-report data.

Psychiatric treatment utilization was drawn solely from parent-report of their child ever having received mental health services.

We employed cross-tabulations to estimate the weighted lifetime, current (ie, past 2-week), and recent-onset (engagement in past 2 weeks only) prevalence of NSSI. We employed bivariate logistic regression to examine 5 sets of potential correlates of lifetime history of NSSI: Sociodemographic characteristics, any diagnosis, number of diagnoses (0, 1, 2, 3+ diagnoses), specific diagnoses, and SI. We subsequently conducted 5 separate multivariable models, each of which covaried the other variables within each set of correlates where applicable (eg, in the sociodemographic multivariable model, we included all sociodemographic variables); the latter 4 multivariable analyses also covaried all sociodemographic variables.

In examining temporal predictors of recent onset of NSSI, analyses were restricted to participants without an NSSI history before the past 2 weeks. We used bivariate logistic regression to examine sociodemographic characteristics, any past diagnosis (ie, before the past 2 weeks), number of past diagnoses, specific past diagnoses, and any past SI as predictors of past 2-week NSSI. We produced multivariable models predicting recent NSSI onset, which were conducted in a parallel manner to the multivariable models examining correlates of NSSI history. To maintain clean temporal separation between predictors and recent NSSI onset, sociodemographic covariates in these models were limited to those conceptualized as static (ie, sex, race, and ethnicity).

Finally, we used cross-tabulations to examine prevalence of psychiatric treatment utilization among youth with a lifetime NSSI history. We employed multivariable logistic regression to examine whether lifetime NSSI history was associated with psychiatric treatment utilization after adjusting for all sociodemographic variables.

All analyses were conducted using SPSS 27.0, and ABCD weighting¹⁸  (to the American Community Survey) was used to produce prevalence estimates.

In the sample of preadolescents with nonmissing NSSI history data (unweighted N = 11 835), the weighted lifetime prevalence of NSSI was 9.17% (SE = 0.30) and past 2-week prevalence of NSSI was 4.24% (SE = 0.21). The weighted prevalence of recent NSSI onset was 2.17% (SE = 0.15).

Sociodemographic correlates of lifetime NSSI are presented in Table 1. In multivariable analyses, female sex, identifying as Black and as either American Indian/Alaskan Native, Asian American, or other racial category were associated with lower odds of lifetime NSSI. Identifying as a sexual minority or not understanding the question, having a family income of <$25 000, and having unmarried parents were associated with higher odds of lifetime NSSI.

Diagnostic correlates of lifetime NSSI are presented in Table 2. In the multivariable model adjusting for all sociodemographic variables, major depressive disorder (MDD), social anxiety, generalized anxiety disorder, posttraumatic stress disorder, conduct disorder, and oppositional defiant disorder (ODD) were significantly associated with lifetime NSSI. In separate multivariable models, also adjusting for sociodemographic covariates, having any disorder increased the odds of lifetime NSSI, and the odds of lifetime NSSI increased in a dose-dependent manner as the number of disorders increased.

In a univariate analysis, SI (odds ratio [OR] 7.68; 95% confidence interval [CI] 6.60–8.93; P < .001) was significantly and robustly associated with NSSI history. In a multivariable model adjusting for sociodemographic factors, lifetime SI (OR 7.59; 95% CI 6.46–8.92; P < .001) remained significant. As sensitivity analyses, these analyses were repeated with passive and active SI, respectively (Supplemental Table 3). The results were largely unchanged.

Results of univariate and multivariable models examining static sociodemographic features (ie, sex, race, ethnicity) as predictors of recent NSSI onset are presented in Supplemental Table 4. In a multivariable model, only sex was a significant predictor of recent NSSI onset. Results of the multivariable model examining temporally primary diagnostic predictors of recent NSSI onset are presented in Supplemental Table 5. Psychosis and eating disorders were excluded from analyses predicting recent onset of NSSI because of their very low past prevalence. In multivariable analysis, only previous MDD, specific phobia, and ODD were significantly predictive of recent NSSI onset. In a separate multivariable model, having any previous disorder increased the odds of recent NSSI onset; having 2 and 3+ disorders increased odds further.

We also examined previous SI as a temporal predictor of recent NSSI onset. In a univariate model, previous SI (OR 4.26; 95% CI 3.13–5.79; P < .001) predicted recent NSSI onset. This finding held after controlling for child sex, race, and ethnicity (OR 4.19; 95% CI 3.04–5.79; P < .001). In sensitivity analyses, both passive and active SI were significant univariate predictors of NSSI onset. However, in the multivariable model, only passive SI remained significant (see Supplemental Table 3).

Of those with a history of NSSI, 36.71% (SE = 1.67) endorsed a history of any mental health treatment. Youth with an NSSI history were more likely to receive mental health treatment than youth without an NSSI history (OR 3.47; 95% CI 2.97–4.04; P < .001), even after adjusting for sociodemographic variables (OR 3.29; 95% CI 2.78–3.89; P < .001).

In the first national population-based US sample to assess NSSI in preadolescent youth to date, the lifetime and past 2-week prevalence of NSSI were 9.1% and 4.2%, respectively. These findings suggest critically high rates of NSSI for this early developmental period.

Several key sociodemographic variables were associated with increased odds of lifetime NSSI history. Male sex was significantly associated with NSSI history and onset. Although the previous literature is mixed with regard to sex differences in rates of NSSI, a recent meta-analysis found female sex prospectively predicted NSSI.²³  However, our finding that males may be at greater risk than females during the preadolescent period mirrors data on suicide and depression in prepubertal youth, suggesting males may be at greater risk for these outcomes at a younger age.^24,25  Given NSSI onset before age 12 years may be associated with worse course of NSSI and greater risk for suicidal behavior,^12,13  the present results highlight the importance of screening for NSSI in preadolescents, particularly among males, who may otherwise be overlooked given previous studies highlighting female risk in older age groups.

Findings from the current study extend previous research with sexual minority adolescents,²⁶  indicating preadolescent sexual minority youth are at similarly elevated risk for NSSI compared with heterosexual youth. One out of every 4 sexual minority preadolescents reported a lifetime history of NSSI; the effect size in this sample was comparable to that for adolescents in a recent meta-analytic review of NSSI and sexual orientation.²⁷  The reasons for elevated risk among sexual minority preadolescents require further exploration, although research suggests that, in sexual minority adolescents and adults, minority-specific stress may play an important role.²⁸  Further, there may be fewer resources available to support youth’s sexual identities in preadolescence, because many supportive resources target adolescents and young adults. Research identifying factors relevant to elevated risk in this population is needed.

Findings suggested racial minority preadolescents (with the exception of biracial youth) were less likely to engage in NSSI compared with white peers. However, because rates of suicide in Black youth are rising in the United States,²⁹  additional research is needed to examine whether rates of NSSI in Black youth may begin to follow a similar trajectory. Additionally, the current study did not examine experiences of racial discrimination, which may serve as a more sensitive indicator of elevated NSSI risk in minoritized youth.³⁰ 

Finally, 2 family-level factors were associated with elevated rates of lifetime NSSI. Odds of NSSI were higher for preadolescents whose primary caregiver was unmarried and whose household income was <$25 000 per year. These findings are consistent with previous research in smaller samples documenting associations between parent income and NSSI in preadolescents.³¹  Findings are also well aligned with previous analyses from the ABCD study, indicating increased financial stress and low parental monitoring were associated with greater risk for SI, NSSI, and suicide attempt (SA).¹⁴  Family income is a key indicator of a family’s access to resources, and understanding its role in risk for psychological outcomes is critical to provide evidence that can be acted upon. Similarly, marital status has direct implications for taxes and social resource eligibility. Thus, studying these variables in relation to NSSI can support policy-level changes to reduce population-level risk.

Consistent with the conceptualization of NSSI as a transdiagnostic phenomenon, multiple diagnostic factors were associated with lifetime history of NSSI and NSSI onset. Specifically, diagnostic comorbidity increased risk for lifetime history of NSSI in youth in a dose-response fashion. In line with previous research examining NSSI correlates in adolescent samples,^32–34  MDD emerged as the strongest single diagnostic predictor both cross-sectionally and temporally. ODD was the only other disorder to significantly predict NSSI in both cross-sectional and temporal analyses. This was a relatively novel finding in that, to our knowledge, only 1 previous study has examined self-reported ODD symptomatology in relation to NSSI, and this study was conducted with an adolescent sample.³⁵  Our results extend these findings to preteens and to the clinical diagnosis of ODD. Furthermore, our findings appear to suggest a robust pattern, given that ODD is more common among males,^36,37  and that ODD remained a significant predictor even after accounting for the higher prevalence of NSSI among preadolescent boys in both cross-sectional and temporal multivariable models. Collectively, these findings highlight the utility of considering general psychopathology severity, and MDD and ODD in particular, when screening for risk for NSSI in preadolescents. Given our robust findings for ODD and the contrasting paucity of previous research on this disorder in relation to NSSI, future prospective evaluations of their association are warranted.

To our knowledge, no studies have tested whether SI predicts onset of NSSI among youth. Our findings indicated that a history of SI significantly predicted >7 times elevated odds of lifetime NSSI and >4 times elevated odds of subsequent NSSI onset. Results demonstrate the importance of monitoring youth with histories of SI for NSSI, because SI onset may precede NSSI onset in a subset of preadolescents. This finding is consistent with evidence suggesting that a common function of NSSI is to cope with SI.^38,39 

There are a number of limitations of this work that warrant consideration. First, our assessment of NSSI was limited to a dichotomous indicator of presence versus absence. Assessment of NSSI methods, severity, frequency, and functions could provide more nuanced information regarding NSSI risk. That said, engaging in any NSSI is predictive of prospective risk for NSSI,⁴⁰  as well as suicidal thoughts and behavior^7,8 ; thus, it remains important to assess for any NSSI history. An additional limitation is this study’s reliance on retrospective report. Future waves of ABCD data will be needed to more rigorously test the temporal relations among sociodemographic factors, mental health diagnoses, SI, and NSSI. Relatedly, epidemiologic data suggest that rates of psychiatric symptoms and disorders are higher when assessed prospectively,⁴¹  and thus retrospective report of lifetime history may underestimate the true prevalence of NSSI in preadolescents; forthcoming ABCD longitudinal data will provide important insight on how these rates evolve.

Previous work examining NSSI among youth has tended to focus on adolescent samples, missing a critical period for detection and intervention during preadolescence. The current study has many strengths, most notably, the contribution of new information on correlates and predictors of NSSI in a national sample of preadolescents. Although NSSI is a low base rate behavior, this population-based sample allowed for meaningful conclusions to be drawn regarding predictors of lifetime history and onset of NSSI. Findings highlight key sociodemographic and diagnostic factors associated with NSSI risk that could be used to focus screening efforts to target high-risk preadolescents. Screening efforts followed by referral may be particularly well suited to be conducted by pediatric health care providers. Our findings suggest that, although youth with a history of NSSI are more likely to engage in psychiatric treatment than those without, >60% of youth with NSSI had no history of psychiatric care. Identifying ways to engage families in psychiatric services may be critical for improving long-term outcomes in these preadolescent youth. Further, there is limited evidence for the effectiveness of treatments directly targeting NSSI in youth⁴² ; results of this study suggest that a transdiagnostic approach may be indicated for at-risk preadolescents.

ABCD

Adolescent Brain and Cognitive Development

CI

confidence interval

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition

MDD

major depressive disorder

NSSI

nonsuicidal self-injury

ODD

oppositional defiant disorder

OR

odds ratio

SI

suicidal ideation