The Presence of Externalizing Diagnoses Increases the Risk of Suicidal Thoughts and Behaviors in Youth Enriched for Early-Onset Depression

Abstract

Objective

To understand the additional impact of externalizing diagnoses and specific externalizing symptoms (eg, impulsivity, irritability) on risk for suicidal thoughts and behaviors (STBs) in a sample of youth enriched for depression.

Method

Participants included 219 children from the Pediatric Suicidality Study, which has followed a sample of preschoolers enriched for depression into early adolescence (ages 10-14 years). Semistructured interviews to assess STBs and clinical diagnoses including major depressive disorder (MDD), attention-deficit/hyperactivity disorder (ADHD), and oppositional defiant disorder (ODD) were conducted at preschool age (3-6 years), preadolescence (8-12 years), and 2 years later (10-14 years).

Results

Meeting diagnostic criteria for an externalizing disorder at any study time point was significantly associated with experiencing STBs after the preschool period, even when controlling for comorbid depression (66.7% vs 27.6%; χ² = 5.70, p = .017). Findings were similar when limiting the analyses to suicidal behaviors only (77.3% vs 44.6%; χ² = 4.00, p = .045). Analyses examining specific externalizing symptoms found that impulsivity and irritability, but not inattention or defiance, were significantly associated with experiencing STBs across the 2 years between follow-up visits leading into early adolescence (impulsivity: estimate = 0.68, SE = 0.29, χ² = 5.47, p = .019; irritability: estimate = 0.94, SE = 0.35, χ² = 7.39, p = .007).

Conclusion

Youth with comorbid depression and an externalizing disorder or externalizing symptoms of irritability or impulsivity may be at an elevated risk of STBs into early adolescence. These findings highlight the importance of early identification and treatment of externalizing disorders and symptoms in addition to treating depression when managing STBs in this population and provide additional potential treatment targets.

Plain language summary

A sample of preschoolers enriched for depression was followed into early adolescence and assessed for suicidal thoughts and behaviors (STBs), major depressive disorder (MDD), attention-deficit/hyperactivity disorder (ADHD), and oppositional defiant disorder (ODD). Having an externalizing disorder (eg, ADHD or ODD) at any time was significantly associated with STBs after the preschool period. Further, externalizing symptoms of impulsivity and irritability, but not inattention or defiance, were significantly associated with STBs across the two years between follow-up visits leading into early adolescence. These findings highlight the importance of identifying and treating externalizing disorders and symptoms in addition to treating depression when managing STBs in this population.

Suicidal thoughts and behaviors (STBs) are a growing problem among children and adolescents. Among youth ages 10 to 14 years, suicide is currently the second leading cause of death in the United States, accounting for more deaths than any single disease.¹ Suicidal ideation and suicide attempts are also increasing, reaching lifetime rates of 15.1% and 2.6%, respectively, in youth younger than 13 years of age.² With the rising prevalence of STBs among children and adolescents, the need for effective identification and treatment of contributing disorders is a pressing concern. One of the most well-established predictors of STBs in youth is major depressive disorder (MDD).³ A recent meta-analysis revealed that in adolescents younger than age 18 with MDD, the prevalence of attempting suicide at least once is 12.5%.⁴ Our group has also recently reported a strong association between early-onset depression and the expression of STBs.⁵ Specifically, in a cohort of children with preschool-onset MDD (PO-MDD), the rate of STBs at ages 8 to 12 years was 67.9% compared with 22.6% in nondepressed peers.⁵ However, even among children and adolescents with MDD, there is variability in who experiences STBs, suggesting a need to further characterize risk factors within the depressive phenotype.

Another potential predictor of STBs that has gained increasing attention in recent years is externalizing disorders such as attention-deficit/hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD).^2,6 One large cross-sectional study of youth ages 6 to 12 across Europe found that externalizing disorders alone were strongly correlated with increased suicidal ideation and thoughts of death.⁷ Another study by Sheftall et al.⁸ found some evidence for a link between ADHD symptoms and suicidal ideation in a group of children ages 6 to 9 enriched for a parental history of suicide attempt, although this finding was present only in univariate analyses. Additional studies have shown that children diagnosed with ADHD before the age of 6 were more likely to endorse STBs in middle childhood and adolescence,⁹ with similar findings observed for concurrent ADHD and STBs.^6,10 Although ODD has been less frequently investigated, it has also been associated with increased STBs in both children^6,11 and adolescents.^6,12 Despite these findings, there is limited research on the role of externalizing disorders when accounting for comorbid mood disorders such as MDD, even though the high rates of comorbidity across these disorders is well established in children and adolescents.^13,14 Indeed, rates of MDD with comorbid ADHD and comorbid ODD have been documented as high as 31.1% to 42% and 50.0% to 62%, respectively, in preschool-age children.^15,16 Other work has found that girls ages 6 to 18 with ADHD had a 2.5 times higher risk for MDD at 5-year follow-up compared with girls without an ADHD diagnosis at baseline.¹⁷ Related to this risk trajectory, one study found an increased likelihood of developing emotion regulation difficulties and depressive symptoms in children with ADHD.¹⁸ However, although these studies demonstrate a reliable association between externalizing diagnoses and MDD, there is still little known about the potentially compounding or interactive effects of comorbid externalizing disorders and MDD on the expression of STBs.

Also, data examining how specific externalizing symptoms such as irritability, impulsiveness, defiance, and inattention might contribute to expression of STBs are lacking. Studies investigating these specific symptoms are limited and have yielded mixed findings. For example, some researchers have shown that inattentive symptoms in youth are associated with an increased risk of STBs^19,20 and nonsuicidal self-injury,²¹ although others have reported the opposite.²² Impulsivity has also been linked to STBs in some studies,^21,23 yet this association has not been consistently observed.¹⁹ Fewer studies have investigated potential links between irritability and defiance and STBs, with irritability being associated with suicidal ideation in one study²³ and defiance showing no association in another.¹² The mixed findings in this area, potentially due to differences in ages of the youth assessed and/or the varied measures used to determine externalizing symptoms, highlight a need for further investigation into how externalizing symptoms might contribute to the expression of STBs. To our knowledge, no studies have specifically assessed the relations between distinct externalizing symptom domains and STBs in youth at elevated risk for experiencing comorbid depression or how these domains may interact with depression in children and adolescents.

The present study was conducted to examine how externalizing diagnoses and symptoms contribute to the expression of STBs in children and adolescents (ages 8-14) at an elevated risk of comorbid depression. We aimed to address this question in a group of youth highly enriched for PO-MDD who exhibited high rates of STBs into early adolescence. Given the still relatively low rates of STBs in early adolescence in nonclinical populations, examining the present questions in a sample enriched for PO-MDD allowed us to maximize our ability to detect additional effects of externalizing diagnoses and symptoms, especially considering the high comorbidity between depressive and externalizing disorders in childhood.^13,14 Our aims were to determine if, overall, externalizing diagnoses or symptoms impact the presence of STBs after the preschool period and into early adolescence (approximately 6-14 years of age) and if, specifically, they impact the persistence of STBs across a 2-year period leading into early adolescence (approximately 10-14 years of age). We hypothesized that externalizing diagnoses and symptoms would be associated with increased rates of STBs in this highly enriched group and would predict persistence of STBs into early adolescence. Additionally, we explored whether we could detect a relation between externalizing disorders/symptoms and suicide behaviors only as a question of importance to identify youth at highest risk.

Method

Participants

Participants were 219 youths from the Pediatric Suicidality Study (PED-SI), a follow-up study of a randomized controlled trial of parent–child interaction therapy with emotion development module (PCIT-ED) for PO-MDD that also included a group of nonclinical participants. Figure S1, available online, details the flow of eligible and assessed participants in the PCIT-ED and PED-SI studies. To summarize, 385 children (139 girls) ages 3 to 6 years completed a baseline PCIT-ED assessment, and 368 of these children were eligible to participate in PED-SI (ie, did not have suspected autism spectrum disorder or developmental delays). Of the 368 eligible children, 229 met criteria for PO-MDD on the baseline standardized clinical interview and were randomized to either waitlist or intervention, though all children with PO-MDD were provided the opportunity to complete the PCIT-ED intervention before the end of the study. Children who either did not meet criteria for PO-MDD at the baseline assessment or were recruited as a non-depressed comparison group (n = 82, 33 girls) were not provided with PCIT-ED, but were invited to participate in the PED-SI follow-up. The PED-SI follow-up included 2 assessments: (FUV1), which occurred approximately 4 years after the PCIT-ED study, when most participants were 8 to 12 years old (mean [SD] age = 10.08 [1.0] years, range 7.06-12.27), and follow-up visit 2 (FUV2), which occurred approximately 2 years after FUV1, when most participants were 10 to 14 years old (mean age = 12.27 [1.2] years, range 8.87-15.91). Across these follow-up visits, 210 participants completed FUV1, and 155 completed FUV2; 219 participants completed at least 1 of the 2 follow-up visits. Comparisons of characteristics between participants who completed at least 1 follow-up visit vs those who did not are shown in Table S1, available online. Of note, participants who completed at least 1 follow-up visit were significantly older at the time of their baseline assessment and more often White than participants who did not complete any follow-up visit. In addition, although we conducted a brief phone screen with families from the PCIT-ED study who did not complete a follow-up PED-SI assessment to obtain basic information (n = 21 at FUV1, n = 46 at FUV2), those participants are not included in the analyses presented here.

Measures

The Schedule for Affective Disorders and Schizophrenia for School-Age Children–Early Childhood (K-SADS-EC)²⁴ and K-SADS–Present and Lifetime (K-SADS-PL)²⁵ were administered in the PCIT-ED and PED-SI studies, respectively, to obtain information on psychiatric symptoms and diagnoses. The K-SADS-EC and K-SADS-PL are semistructured, age-appropriate diagnostic interviews for DSM-5 disorders. Either the K-SADS-EC (preschool assessments) or the K-SADS-PL (PED-SI follow-up assessments) was administered to a primary caregiver at each visit. For both versions, the caregiver completed the MDD module in full, regardless of presence of depressive symptoms on screening items. For the remainder of the diagnostic modules, the caregivers reported on screening items, and the supplemental items were administered only if meeting the screen criteria.

During the PED-SI study, not only were caregivers administered the K-SADS-PL modules as above, but also children were administered the screen and supplement items from the MDD module in full as well as the screen items (and when meeting criteria the corresponding supplement) for social phobia, generalized anxiety, posttraumatic stress, obsessive-compulsive, and mania modules. Children were not asked to report on their own ADHD, ODD, or conduct disorder (CD) symptoms, as prior studies have demonstrated lower reliability and validity for child-reported externalizing symptoms relative to caregiver reports.²⁶ For each item, caregivers and, when applicable, children were asked about presence of the symptoms in the prior 30 days (present) and since last visit (past). For the 9 participants who completed an FUV2 assessment without having participated at FUV1, the time period covered at FUV2 was the past 2 years. For the K-SADS-PL modules that were delivered to both caregiver and child, symptoms were considered present if endorsed by the caregiver and/or child at a threshold level, and diagnoses were based on information provided by both informants.

K-SADS interviewers were trained by an expert clinician and rated with an expert clinician until reliability was established. For all symptoms and diagnoses, interviewers participated in a weekly case conference to maintain adherence to the interview and standardization of symptom ratings and to prevent drift from these standards. To assess interrater reliability, an experienced clinician independently coded the K-SADS-PL by watching recorded assessments, and their ratings were compared with the consensus diagnoses obtained during weekly case conference meetings. In total, 20% of FUV1 assessments (42 parent, 40 child) and 20% of FUV2 assessments conducted by new raters (14 parent, 16 child) were coded for reliability. Depression was defined as a consensus diagnosis of MDD or MDD–not otherwise specified (MDD-NOS) between trained raters and master clinicians based on interview responses. The κ values for MDD/MDD-NOS were 0.78 and 0.77 for parent and child, respectively. STBs were assessed in the MDD section of the K-SADS-EC and K-SADS-PL and were defined as endorsement at the subthreshold or threshold level of any of the following:

• 1.Passive suicidal ideation (eg, a wish to be dead)
• 2.Active suicidal ideation (eg, wanting to kill oneself)
• 3.Suicidal behaviors (eg, preparations or planned attempt, such as acquiring a means)
• 4.Suicide attempt (intentional actions taken with the goal of killing oneself).

The κ values for STBs were 0.93 and 0.94 for parent and child reports, respectively. When examining types of STBs, items 1 and 2 comprised the suicidal ideation variable, and items 3 and 4 comprised the suicide behavior variable.

Externalizing diagnoses included ADHD, ODD, and CD. Externalizing symptom severity scores were calculated as a sum of each ADHD, ODD, and CD screen item that was endorsed, as these items were administered systematically to all parents. For the externalizing symptom domains, inattention and impulsivity were assessed via the K-SADS ADHD screen questions, and irritability and defiance were assessed via the K-SADS ODD screen questions. Specific items that comprise each symptom score are detailed in Table S2, available online.

Analysis

Using longitudinal data collected in the PED-SI study, we conducted the following cross-sectional analyses. Externalizing diagnoses and symptoms were investigated as potential predictors of STBs using logistic regression models. First, externalizing diagnoses (any lifetime externalizing disorder and then lifetime ADHD and ODD separately; CD was not analyzed separately due to low rates of endorsement) were entered as the independent variable (IV) predicting presence of STBs through FUV2, the last time point assessed, when participants were 10 to 14 years old (mean age = 12.27 years). The model was run without covariates and covarying for sex at birth and lifetime MDD/MDD-NOS through FUV2, to assess the independent effects of having an externalizing diagnosis over and above a diagnosis of MDD. Second, a separate regression model was run with externalizing symptom severity (sum of the 12 ADHD, ODD, and CD screen items) as the IV predicting STBs at FUV2. Again, the model was run without covariates and covarying for sex at birth and current MDD symptoms at FUV2 to separate the contributions of current externalizing symptoms from depressive symptoms on STBs.

Given the prior associations of externalizing diagnoses with poverty,²⁷ we also added income-to-needs ratio (INR) as a covariate. The INR was assessed as the caregiver-reported family income divided by the number of people in the household, where 1.0 indicates the nationally defined poverty line. However, as our sample was largely composed of middle- to high-income families (mean [SD] INR = 3.0 [1.27]) and INR did not significantly contribute to any outcome, these are included as supplementary analyses in Table S3, available online. Additionally, as our study sample was in part composed of a group that received a specific clinical intervention, we conducted supplementary analyses covarying for the number of PCIT-ED sessions completed (Table S3, available online).

Given the previously high rates of STBs in this study sample, we then limited the analyses to participants who reported STBs at any point from study initiation to FUV1 to assess whether the impact of externalizing diagnoses or symptom severity on STBs was related to the presence of STBs at the final time point or also impacted their persistence across time. We repeated the prior logistic regression models in this subset of participants with prior STBs. Specifically, we first ran a model with externalizing diagnoses as the IV predicting STBs at FUV2 and including sex at birth and lifetime diagnosis of MDD as covariates, followed by a second set of models with externalizing symptom severity as the IV predicting STBs at FUV2 and including sex at birth and current MDD symptoms as covariates. For both we also conducted supplementary analyses covarying for INR and the number of PCIT-ED sessions completed (Table S3, available online).

To assess relations between individual externalizing symptoms and the presence of STBs in early adolescence, we ran logistic regression models for each of the 4 proposed externalizing symptoms domains: inattention, impulsivity, irritability, and defiance. In each model, one of the symptoms domains was entered as the IV predicting FUV2 STBs, with sex at birth and concurrent MDD symptoms included as covariates as well as supplementary analyses covarying for INR and PCIT-ED sessions completed (Table S4, available online).

Finally, to probe more directly the relation between externalizing disorders and externalizing symptom severity and suicide-related behaviors, we repeated the prior analyses comparing presence/absence of suicide behaviors vs no suicide behaviors and presence/absence of suicide behaviors vs suicidal ideation, as an even stronger test of potential links between externalizing behaviors and suicide behaviors.

Results

A total of 219 participants completed a FUV1 and/or FUV2 assessment, with a mean (SD) age at last assessment of 11.6 (1.6.) years. Demographic and diagnostic information for the sample is presented in Table 1. Rates of each externalizing diagnosis through the FUV2 were ADHD 36.1% (n = 79), ODD 39.3% (n = 86), and CD 3.2% (n = 7), with 98.2% of diagnoses captured before/at FUV1. As detailed in Table 2, meeting diagnostic criteria for an externalizing disorder at any time point during the study was significantly associated with endorsing STBs after the preschool period even when covarying for sex at birth and lifetime MDD/MDD-NOS (66.7% vs 27.6%; χ² = 5.70, p = .017). When ADHD and ODD diagnoses were assessed individually, both were initially associated with endorsing STBs after the preschool period, but these findings did not remain significant after including covariates in the model. Similarly, the severity of externalizing symptoms at FUV2 was significantly associated with endorsing concurrent STBs, but this association did not remain significant when concurrent MDD symptoms were entered into the model. Of note, lifetime MDD diagnosis and concurrent MDD symptoms were both significant predictors of STBs in the first and second models.

Table 1.

Participant Characteristics (N = 219)

	No.	Mean	(SD)
Child age, y
Baseline	219	5.26	(1.02)
FUV1	210	10.08	(1.00)
FUV2	155	12.27	(1.20)
Last assessment (either FUV1 or FUV2)	219	11.59	(1.56)
Years between preschool period and FUV1	210	4.35	(0.91)
Years between FUV1 and FUV2	146	2.12	(0.39)
Baseline income-to-needs ratio	219	3.00	(1.27)
	No.	n	(%)
Sex at birth	219
Female		82	(37.4)
Male		137	(62.6)
Hispanic ethnicity	219	16	(7.3)
Race	219
Asian		5	(2.3)
Black		20	(9.1)
Multiracial		21	(9.6)
White		173	(79.0)
MDD/MDD-NOS
Lifetime through FUV2	219	160	(73.1)
Current at FUV2	155	41	(26.5)
Suicidal thoughts and behaviors
After preschool period through FUV2	219	132	(60.3)
After FUV1 through FUV2	155	62	(40.0)
Externalizing disorder
Lifetime through FUV2	219	112	(51.1)
	No.	Mean	(SD)
FUV2 severity scores
Externalizing	153	0.81	(1.38)
Impulsivity	154	0.36	(0.66)
Irritability	153	0.31	(0.63)
Defiance	153	0.53	(1.03)
Inattention	154	1.11	(1.43)

Note: FUV1 = follow-up visit 1; FUV2 = follow-up visit 1; MDD = major depressive disorder; NOS = not otherwise specified.

Table 2.

Externalizing Disorders and Externalizing Symptom Severity as Predictors of Suicidal Thoughts and Behaviors (STBs)

	All participants (N = 219)
	No STBs after preschool (n = 87)			STBs after preschool (n = 132)			No covariates		Covarying for sex and MDDa
	No.	n	(%)	No.	n	(%)	χ2	p	χ2	p
Externalizing disorders
FUV2 lifetime externalizing disorder	87	24	(27.6)	132	88	(66.7)	30.01	<.001	5.70	.017
FUV2 lifetime ADHD	87	17	(19.5)	132	62	(47.0)	16.17	<.001	1.88	.170
FUV2 lifetime ODD	87		(23.0)	132	66	(50.0)	15.35	<.001	1.99	.158
	No.	Mean	(SD)	No.	Mean	(SD)	χ2	p	χ2	p
Externalizing severity
FUV2 externalizing severity	54	0.43	(0.96)	99	1.02	(1.52)	6.03	.014	1.58	.209

	Participants with lifetime STBs at FUV1 (n = 91)
	No STBs after FUV1 (n = 45)			STBs after FUV1 (n = 46)			No covariates		Covarying for sex and MDDa
	No.	n	(%)	No.	n	(%)	χ2	p	χ2	p
Externalizing disorders
FUV2 lifetime externalizing disorder	45	24	(53.3)	46	33	(71.7)	3.24	.072	2.46	.117
	No.	Mean	(SD)	No.	Mean	(SD)	χ2	p	χ2	p
Externalizing severity
FUV2 externalizing severity	45	0.47	(1.06)	46	1.52	(1.77)	8.95	.003	6.65	.010

Note: ADHD = attention-deficit/hyperactivity disorder; FUV1 = follow-up visit 1; FUV2 = follow-up visit 2; MDD = major depressive disorder; NOS = not otherwise specified; ODD = oppositional defiant disorder.

^aSex was sex at birth. MDD was lifetime MDD/MDD-NOS through FUV2 for the models with lifetime disorders as predictors and current MDD/MDD-NOS at FUV2 for the models with current FUV2 severity as predictors.

In the subsample of participants who endorsed STBs at or before FUV1 (mean age = 10.08 years), participants who reported continued STBs in the 2 years between follow-up visits had a higher rate of lifetime externalizing disorders than those without STBs in this interval (71.7% vs 53.3%), although this association was not statistically significant (Table 2). However, externalizing symptom severity reported to occur between follow-up visits significantly differed between these groups, with greater externalizing symptom severity in early adolescence with STBs during this period even when covarying for sex and current MDD symptoms (mean [SD] = 1.52 [1.77] vs 0.47 [1.06]; χ² = 6.65, p = .010).

We assessed the relations between specific externalizing symptoms domains of inattention, impulsivity, irritability, and defiance and STBs as reported in Table 3. This table details the associations between the FUV2 externalizing symptom domain scores and STBs in the 2 years between FUV1 and FUV2 for all participants. Impulsivity and irritability severity were significantly associated with endorsement of STBs across this 2-year period, even when covarying for sex at birth and concurrent MDD/MDD-NOS measured at FUV2 (impulsivity: estimate = 0.68, SE = 0.29, χ² = 5.47, p = .019; irritability: estimate = 0.94, SE = 0.35, χ² = 7.39, p = .007). Defiance and inattention severity were not significantly associated with STBs in this period.

Table 3.

Follow-Up Visit 2 (FUV2) Externalizing Symptom Domain Scores as Predictors of Suicidal Thoughts and Behaviors (STBs) Between Follow-Up V1 (FUV1) and FUV2

	Estimate	SE	χ2	p
DV = STBs after FUV1 (n = 154)
Intercept	−0.20	0.24	0.71	.398
Female sex	0.01	0.20	0.00	.946
FUV2 current MDD/MDD-NOS	1.04	0.22	23.32	<.001
FUV2 impulsivity severity	0.68	0.29	5.47	.019
DV = STBs after FUV1 (n = 153)
Intercept	−0.31	0.25	1.59	.207
Female sex	−0.10	0.20	0.27	.602
FUV2 current MDD/MDD-NOS	0.92	0.22	17.15	<.001
FUV2 irritability severity	0.94	0.35	7.39	.007
DV = STBs after FUV1 (n = 153)
Intercept	−0.19	0.25	0.61	.434
Female sex	−0.07	0.20	0.14	.708
FUV2 current MDD/MDD-NOS	0.97	0.22	19.38	<.001
FUV2 defiance severity	0.36	0.20	3.43	.064
DV = STBs after FUV1 (n = 154)
Intercept	−0.16	0.27	0.37	.542
Female sex	−0.04	0.20	0.05	.831
FUV2 current MDD/MDD-NOS	1.03	0.22	22.73	<.001
FUV2 inattention severity	0.16	0.13	1.58	.208

Note: DV = dependent variable; MDD = major depressive disorder; NOS = not otherwise specified.

When limiting the outcome variable to suicide behaviors only (vs no suicide behaviors), we found that having a lifetime externalizing disorder was significantly related to endorsing suicidal behaviors after the preschool period, even when covarying for sex at birth and lifetime MDD/MDD-NOS (77.3% vs 44.6%; χ² = 3.91, p = .048) (Table 4). Moreover, when examining ADHD and ODD diagnoses separately, although both lifetime ADHD and ODD were significantly related to endorsing suicidal behaviors in the first models, only lifetime ADHD remained significant after including the covariates (χ² = 6.72, p = .010). Externalizing symptom severity at FUV2 was also significantly associated with suicide-related behaviors specifically, but did not remain significant when covariates were included in the model. Finally, in the models that directly compared suicide behaviors with suicidal ideation, lifetime ADHD and externalizing symptom severity were both related to endorsing suicide behaviors in the first model, with only lifetime ADHD remaining significant after including covariates (χ² = 5.34, p = .021) (Table 5).

Table 4.

Externalizing Disorders and Externalizing Symptom Severity as Predictors of Suicidal Behaviors

	All participants (N = 219)
	No suicidal behaviors after preschool (n = 175)			Suicidal behaviors after preschool (n = 44)			No covariates		Covarying for sex and MDDa
	No.	n	(%)	No.	n	(%)	χ2	p	χ2	p
Externalizing disorders
FUV2 lifetime externalizing disorder	175	78	(44.6)	44	34	(77.3)	13.63	<.001	4.00	.045
FUV2 lifetime ADHD	175	51	(29.1)	44	28	(63.6)	16.66	<.001	6.93	.009
FUV2 lifetime ODD	175	63	(36.0)	44	23	(52.3)	3.83	.050	0.21	.647
	No.	Mean	(SD)	No.	Mean	(SD)	χ2	p	χ2	p
Externalizing severity
FUV2 externalizing severity	122	0.63	(1.13)	31	1.52	(1.95)	8.62	.003	2.22	.136

Note: ADHD = attention-deficit/hyperactivity disorder; FUV2 = follow-up visit 2; MDD = major depressive disorder; NOS = not otherwise specified; ODD = oppositional defiant disorder.

^aSex was sex at birth. MDD was lifetime MDD/MDD-NOS through FUV2 for the models with lifetime disorders as predictors and current MDD/MDD-NOS at FUV2 for the models with current FUV2 severity as predictors.

Table 5.

Externalizing Disorders and Externalizing Symptom Severity as Predictors of Suicidal Behaviors Compared With Suicidal Ideation Only

	All subjects with suicidal thoughts and/or behaviors after preschool (n = 132)
	Suicidal ideation only after preschool (n = 88)			Suicidal behaviors after preschool (n = 44)			No covariates		Covarying for sex and MDDa
	No.	n	(%)	No.	n	(%)	χ2	p	χ2	p
Externalizing disorders
FUV2 lifetime externalizing disorder	88	54	(61.4)	44	34	(77.3)	3.27	.071	1.66	.198
FUV2 lifetime ADHD	88	34	(38.6)	44	28	(63.6)	7.15	.008	5.34	.021
FUV2 lifetime ODD	88	43	(48.9)	44	23	(52.3)	0.14	.712	0.00	.979
	No.	Mean	(SD)	No.	Mean	(SD)	χ2	p	χ2	p
Externalizing severity
FUV2 externalizing severity	68	0.79	(1.23)	31	1.52	(1.95)	4.38	.036	1.31	.253

Note: ADHD = attention-deficit/hyperactivity disorder; FUV2 = follow-up visit 2; MDD = major depressive disorder; NOS = not otherwise specified; ODD = oppositional defiant disorder.

^aSex was sex at birth. MDD was lifetime MDD/MDD-NOS through FUV2 for the models with lifetime disorders as predictors and current MDD/MDD-NOS at FUV2 for the models with current FUV2 severity as predictors.

Discussion

The current study aimed to investigate the contributions of externalizing disorders and symptom domains on the expression of STBs in a group of children enriched for PO-MDD and followed into early adolescence (approximately 10-14 years of age). Within this sample, youth who experienced STBs after the preschool period were more likely to have a lifetime diagnosis of an externalizing disorder (eg, ADHD, ODD, CD). This association remained significant when focusing solely on suicidal behaviors, underscoring the elevated risk due to the comorbidity of externalizing disorders and depression in youth. Furthermore, in youth with a history of STBs, externalizing symptoms predicted the persistence of STBs across a 2-year period in early adolescence, even when accounting for potential concurrent depression diagnoses. After breaking externalizing symptoms down into subcomponents, irritability and impulsivity, but not defiance or inattention, were specifically associated with the persistence of STBs into early adolescence. While both impulsivity and irritability are transdiagnostic symptoms, with irritability even being a potential symptom of depression, our findings held even when controlling for concurrent depressive symptoms. This is noteworthy given that 26.5% of the sample met clinical diagnostic criteria for MDD/MDD-NOS at their most recent follow-up visit. These findings demonstrate the unique contribution of childhood externalizing disorders and symptoms in the expression and persistence of STBs in youth. Our results also highlight the importance of connecting early diagnosis with effective treatment of externalizing conditions to improve the management of STBs, with particular attention paid to irritability and impulsivity, even in the presence of comorbid MDD.

The results of this study align with previous research showing an association between STBs and externalizing diagnoses in children and adolescents without comorbid mood disorders^6,7,12,28 and extends these findings in novel ways. Given the high comorbidity rates between externalizing diagnoses and MDD^13,14,29 and the link between both disorders and STBs, it is important to determine whether these comorbidities confer additional risk for STBs. By demonstrating a link between externalizing diagnoses and STBs in a sample of youth enriched for early-onset depression, this study highlights an important additional target for intervention. Moreover, our findings suggest this is particularly true for youth who have already experienced STBs. In youth who had STBs at or before the first follow-up visit in preadolescence (mean age 10.08 years), having higher externalizing symptoms was associated with the persistence of STBs in the subsequent 2 years. Additionally, our findings suggest that in this population, lifetime ADHD has an even greater association with suicide behaviors compared with ideation. This emphasizes that there may be risk specifically for suicide behaviors, including attempts, in youth with ADHD. Moreover, there were significant differences in externalizing symptoms assessed dimensionally between youth who continued to have STBs into adolescence and those who did not, consistent with prior work demonstrating that externalizing symptoms are associated with increased STBs even when externalizing diagnoses are controlled for or absent.^19,23,30 Clinicians should be aware of the potential additional risk of STBs conferred by the presence of externalizing symptoms. In youth with STBs, though treatment of MDD when present is essential, it should not preclude treatment of comorbid externalizing disorders, as these may also play a significant role in maintaining suicidal ideation and in future suicide attempts.

Given the limited literature on which specific externalizing symptom domains most impact STBs, we assessed this directly. Here we found a significant relation between impulsivity and irritability with STBs in early adolescence. Impulsivity is one of the most commonly assessed symptoms in prior literature, given the impulsive nature of many suicide attempts.³¹ A study by Ojala et al.²¹ found that youth with increased severity of impulsivity at ages 9 to 12 years were more likely to endorse suicidal behaviors at age 15, findings supported by the results of a cross-sectional study showing that adolescents (mean age 14.6 years) with more severe self-reported impulsive behaviors were significantly more likely to endorse suicidal ideation compared with matched controls.¹⁹ Our findings are consistent with this prior work by linking impulsivity to STBs in early adolescence, over and above symptoms of depression.

Although irritability has been less frequently investigated, our findings demonstrate a positive association between irritability and STBs in early adolescence. This finding is consistent with research showing a significant association between concurrent irritability and suicidal ideation in male youth 15 to 20 years old.²³ Although the link between irritability is perhaps less obviously direct than impulsivity, there are multiple potential mechanisms. In one large study, rising irritability across childhood was related to both suicidality and depressive symptoms,³² consistent with a pathway from irritability to STBs via internalizing diagnoses.³³ However, we demonstrate that irritability is associated with STBs even when controlling for depression, suggesting that these effects are at least partially separable. Evidence from a study of youth ages 6 to 16 years with MDD, persistent depressive disorder, or both found that irritability predicted parent-reported impulsive behaviors,³⁰ suggesting that irritability could contribute to the expression of impulsive tendencies, which might ultimately lead to expression of STBs, consistent with an externalizing pathway from irritability to STBs.³³ This may be one mechanism at play here, as we did not include irritability and impulsivity in the same model. However, irritability may also contribute to STBs via negative effects on family or peer relationships, as irritability has a well-demonstrated negative effect on these relationships,^34,35 and family and peer conflicts are known risk factors for suicidality in youth.³⁶ However, we also acknowledge that our measure of irritability is relatively limited in that we were restricted to the temper loss item present in the K-SADS screening sections rather than a full measure of both tonic and phasic irritability.^37,38

In contrast to impulsivity and irritability, we found no relation between defiance or inattention and STBs. Research on the role of defiance in the context of youth STBs has been limited, although our findings are consistent with one prior study that found male adolescents with a defiant ODD subtype did not endorse higher levels of STBs.¹² In contrast, most studies investigating inattention have documented an association between increased inattention and an elevated risk of STBs.19, 20, 21 However, these studies focused predominantly on community-based samples of adolescents 13 to 17 years old without an ADHD diagnosis^19,20 or on adolescents at risk for neurodevelopmental disorders based on positive twin screens.²¹ One study by Doyle and Fite²² that focused on younger children (ages 6-12 years), who were seeking psychiatric treatment and were thus more similar to our participant group, found that inattention was associated with a decreased level of suicidal behaviors. These mixed findings raise the possibility that increased symptoms of inattention may indicate risk for STBs in youth without significant mental health needs (ie, no specific diagnoses or treatment-seeking behaviors), but may be less relevant or even potentially protective in youth with more severe psychiatric issues. These findings also suggest that inattention may play a greater role in the expression of STBs in adolescents than in younger children.

One key avenue for future research is to investigate how the development, expression, and persistence of STBs in youth with PO-MDD and comorbid externalizing diagnoses compare with STBs youth with externalizing diagnoses alone. Prior literature has established that externalizing diagnoses alone are associated with a higher risk of experiencing STBs.^6,7 However, to our knowledge, no studies have compared youth with externalizing diagnoses alone with youth with externalizing diagnoses and high rates of comorbid MDD. Future research should also focus on understanding how MDD and externalizing symptoms may interact to increase risk for either maintenance or new onset of STBs later into adolescence and adulthood and how specific externalizing symptoms in nonclinical samples of youth increase risk. Additional important questions for future research include examining differences in STB frequency or severity as a function of comorbid irritability and/or impulsivity, alongside potential risk for suicide attempt and/or death by suicide.

This study has several limitations. Given the unique study sample, generalizability of our findings may be limited. However, to our knowledge, this is the first study to specifically investigate the association between externalizing diagnoses and the expression of STBs in a sample of youth enriched for MDD. Whereas the findings may be limited in their generalizability, the ability to deeply characterize the relation between externalizing psychopathology and STBs in a group of youth enriched for such early-onset psychopathology contributes important clues to early elements of risk for STBs. In addition to our study being highly enriched for early-onset depression, the predominately White composition of the sample limits generalizability to minoritized youth. This is a particularly important topic for future research to address given increases in Black youth suicide, differing factors contributing to suicide between racial groups,³⁹ and disparities in the diagnosis and treatment of externalizing disorders such as ADHD in minority populations.⁴⁰ Moreover, our study population also reported relatively low rates of trauma and adverse childhood experiences, limiting our ability to assess interactive effects of adverse childhood experiences and psychopathology on STBs. This study also did not control for any treatment or interventions that participants may have obtained outside of the study itself, which in theory could have altered (likely reduced) the rate of STBs experienced by youth in this sample. However, with 40% of the sample reporting experiencing STBs in the 2-year period between follow-up visits, we suggest that even with a history of treatment many youth continue to struggle with STBs. The study results are also limited in that the association found between externalizing diagnoses and STBs cannot be defined as causal based on the statistical analyses conducted. Whereas externalizing diagnoses such as ADHD are neurodevelopmental disorders, and greater than 90% of the lifetime diagnoses were present before the STBs reported here, externalizing symptoms and STBs were measured concurrently, meaning it cannot be said with certainty that one preceded the other. Additionally, the sample size in this study was relatively small (albeit relatively large given its unique characteristics) and thus underpowered to look at more complex relations between factors of potential interest.

This study highlights the importance of investigating how both childhood internalizing and externalizing disorders relate to the expression and persistence of STBs across childhood. Our findings indicate that in youth who are at an elevated risk of early-onset depression, having a diagnosis of an externalizing disorder such as ADHD or ODD may increase risk for experiencing both suicidal ideation and suicidal behaviors into early adolescence. In youth with depression or elevated depressive symptoms, assessing irritability and impulsivity may also provide additional information about who is most at risk for persistent or escalating STBs. These findings not only offer insight into the identification of youth at highest risk for expression of STBs, but also highlight important potential treatment targets.

CRediT authorship contribution statement

Maida Duncan: Writing – review & editing, Writing – original draft. Alecia C. Vogel: Writing – review & editing, Writing – original draft, Supervision, Funding acquisition, Conceptualization. Ashna Ramiah: Writing – review & editing, Writing – original draft. Rebecca Tillman: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Deanna M. Barch: Writing – review & editing, Methodology, Investigation, Funding acquisition, Conceptualization. Joan Luby: Writing – review & editing, Supervision, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization. Laura Hennefield: Writing – review & editing, Validation, Supervision, Funding acquisition, Conceptualization.