Adolescent Predictors of Deliberate Self-Harm Thoughts and Behavior among Young Adults: A Longitudinal Cross-National Study

Abstract

Purpose:

This study builds upon and extends previous longitudinal research on deliberate self-harm (DSH) among youth by investigating which risk and protective factors during adolescence predict DSH thoughts and behavior in young adulthood`.

Methods:

Self-report data came from 1,945 participants recruited as state-representative cohorts from Washington State and Victoria, Australia. Participants completed surveys in 7th grade (average age 13), as they transitioned through 8th and 9th grades, and online at age 25. Retention of the original sample at age 25 was 88%. A range of risk and protective factors in adolescence for DSH thoughts and behavior in young adulthood were examined using multivariable analyses.

Results:

Across the sample, 9.55% (n=162) and 2.83% (n=48) of young adult participants reported DSH thoughts and behaviors, respectively. In the combined risk-protective factor multivariable model for young adulthood DSH thoughts, depressive symptoms in adolescence (AOR=1.05; CI=1.00–1.09) increased risk, while higher levels of adolescent adaptive coping strategies (AOR=.46; CI=.28-.74), higher levels of adolescent community rewards for prosocial behavior (AOR=.73; CI=.57-.93), and living in Washington State decreased risk. In the final multivariable model for DSH behavior in young adulthood, less positive family management strategies during adolescence remained the only significant predictor (AOR=1.90; CI=1.01–3.60).

Conclusions:

DSH prevention and intervention programs should not only focus on managing depression and building/enhancing family connections and support, but also promote resilience through efforts to promote adaptive coping and connections to adults within one’s community who recognize and reward prosocial behavior.

Implications and Contribution:

In addition to knowledge about long-term effects of risk factors (i.e., depressive symptoms, poor family management strategies), this study adds to the limited research on protective factors associated with reduced risk of DSH thoughts and behavior over time, specifically use of adaptive coping strategies and community rewards for prosocial behavior.

Deliberate self-harming behaviors represent a significant public health problem among young people worldwide.^1–6 Cross-nationally, debates remain regarding how to correctly define self-harming behavior. Researchers within European countries and in Australia predominantly use the term deliberate self-harm (DSH) as a more encompassing term for self-directed harmful behavior, regardless of suicidal intent.^5,6 In contrast, researchers within Canada and the U.S. use the term non-suicidal self-injury (NSSI), which only includes directly harmful behavior without suicidal intent.^5,6 International research found comparable rates of lifetime NSSI and DSH among adolescents (18% and 16%, respectively),⁵ and numerous predictors of NSSI and DSH overlap.⁶ Approximately 27% of adolescents report past-year thoughts of self-harm.⁷ Thus, we use the term DSH to refer to self-injurious thoughts and behavior, with or without suicidal intent. DSH behavior is associated with increased risk of suicide, including among young adults.^9,10 Epidemiological research has identified a variety of factors associated with DSH thoughts and behavior among adolescents such as psychiatric problems, lack of supportive interpersonal relationships (e.g., with parents or other non-parental adults), use of maladaptive coping strategies, and co-occurring negative health behaviors.^2,7,11–13 However, most research regarding DSH among young people analyzes cross-sectional data. Longitudinal studies are critical for informing prevention and intervention programming targeting DSH.

In 2015, Plenar, Schumacher, Munz, and Groschwitz⁶ performed a systematic review of longitudinal research examining DSH. Since then, a few additional longitudinal studies with adolescents have been published.^13–18 Although these prospective studies have targeted important gaps in the literature, several weaknesses still remain in the extant longitudinal research on DSH. First, the follow-up periods for most of these studies were short,^6,15,18 with the majority of research designs including only a 12-month time-interval between baseline and follow-up assessments (range: 6 months – 14.5 years). Second, given the short follow-up periods, most researchers only examined factors associated with DSH solely during adolescence or solely during young adulthood (i.e., among college students).^16,17 To our knowledge, only one study¹⁹ examined factors in adolescence that predicted DSH during young adulthood. However, consistent with most longitudinal research on DSH, the investigators only examined risk factors, neglecting to examine protective factors important in prevention research and intervention efforts. Thus, a third weakness of the extant literature involves a lack of research on malleable protective factors that may prevent DSH thoughts and behavior and/or reduce the likelihood DSH behavior will continue beyond adolescence. Finally, most previous longitudinal research focused on intrapersonal/individual factors associated with DSH.^6,18,19 No studies have examined risk and protective factors across levels of adolescents’ social ecologies (i.e., individual, interpersonal, school, and community).

The current study builds on previous research that examined DSH over a 12-month time-period during adolescence in school-based, state-representative samples from Washington State and Victoria, Australia.¹³ This prior study investigated predictive associations between a range of individual, family, and substance use antecedents measured in grades 7 and 9 (age 13 or 15 years) and DSH behavior 12-months later. Depressive symptoms, antisocial behavior and alcohol use increased risk for incident DSH behavior, but only depressive symptoms were predictive of persistent DSH behavior (i.e., DSH behavior at time 1 and time 2). Cross-national differences in DSH behavior were not completely explained by the antecedents included in multivariable models.

The current study builds upon and extends these previous findings by analyzing data from the same state-representative samples to investigate which risk and protective factors during adolescence predict DSH thoughts and behavior in young adulthood at age 25. Testing whether these relationships are the same across countries is an important validity check in cross-national studies.²⁰ Data are from the International Youth Development Study (IYDS), a unique cross-national study designed to overcome methodological inconsistencies in data collection that commonly bias cross-national studies.^21,22 Three research questions guided the analyses. First, what adolescent risk and protective factors predict DSH thoughts at age 25? Second, what adolescent risk and protective factors predict DHS behavior at age 25? Third, do adolescent factors that predict DSH thoughts and behavior at age 25 differ by state?

Methods

The International Youth Development Study (IYDS) follows two cohorts of participants: one from Washington State, USA and one from Victoria, Australia. Relevant to the current analysis in 2002, 7th grade schools were randomly selected using a probability proportionate to grade-level size sampling procedure. A class within each school was then randomly selected. Study protocols remained equivalent across locations and time points. The University of Melbourne Human Ethics in Research Committee and the Royal Children’s Hospital Ethics in Human Research Committee in Victoria and the University of Washington Human Subjects Institutional Review Board in Washington State approved the study. Additional information on response rates and similarities between the two states is available elsewhere.²³

Sample

Self-report data came from 1,945 participants who were in 7th grade (average age 13) at baseline, surveyed again as they transitioned through 8th and 9th grades, and then at age 25 in 2014. Retention rates of the original sample, in both states, were 83% or higher at all follow-up surveys, including 88% at age 25. An analysis of attrition found that nonparticipation was not significantly related to DSH thoughts or behavior.

Measures

The IYDS survey, administered at school during adolescence and online in young adulthood, was adapted from the Communities That Care Youth Survey.²⁴ The survey draws on a range of established scales, including the Short Mood and Feelings Questionnaire.²⁵ Similar versions of the survey have been published elsewhere.^26,27 Items from the adolescent surveys deemed too young for young adults were adjusted and additional questions were added to the young adult survey, as appropriate. All measures were pilot-tested in both states to ensure cross-state comparability.²⁸ The survey measures have demonstrated longitudinal validity and reliability among adolescents and young adults in both states.^29,30

Young adult (age 25) DSH thoughts and behavior.

Two items measured young adult DSH thoughts and behavior. First, the item, “Have you ever had thoughts about deliberately harming or injuring yourself or attempting suicide?” was used to assess DSH thoughts. Second, the question, “Have you ever deliberately hurt yourself or done anything that you knew might have harmed you or even killed you?” was used to assess DSH behavior. Response options for both questions were ‘No,’ ‘Yes, but not in the past 12 months,’ ‘Yes, in the past 12 months and also before this time,’ and ‘Yes, only in the past 12 months.’ Responses were combined to create two dichotomized variables reflecting DSH thoughts in the past year (1) versus no DHS thoughts in the past year (0, reference group), and DSH behavior in the past year (1) versus no DSH behavior in the past year (0, reference group).

Risk and protective factors.

The analyses examined 12 risk and protective factors measured in adolescence. Risk and protective factors were chosen based on previous studies implicating them as risk or protective factors for DSH thoughts and/or behavior.^13–17,31 In early bivariate analyses conducted to inform the current study, we tested for effects of adolescent antisocial behavior and alcohol use on risk for DSH thoughts and behavior in adulthood and found no significant results. Based on these results, and for parsimony, we chose not to include these variables in the current multivariable analysis (full details available upon request). Risk factors included depressive symptoms, bullying victimization, poor family management practices, family conflict, academic failure, and low commitment to school. Protective factors included adaptive coping, emotional control, attachment to parents, interaction with prosocial peers, peer rewards for prosocial behavior, and community rewards for prosocial behavior. Participant responses for each individual risk and protective factor were averaged to obtain single scale scores across grades 7–9. Detailed descriptions and summary statistics for all risk and protective factors are presented in Table 1.

Table 1.

Summary statistics and tests of state differences for adolescent risk and protective factors

Risk (R) and protective (P) factors during adolescence	No. of scale items	Response options	Combined sample (N = 1,954)		Washington State sample (n = 961)		Victorian sample (n = 984)
			Mean (SD)/%	α	Mean (SD)/%	α	Mean (SD)/%	α	t	d	p
Individual-level variables
Depressive symptoms (R)a	13	0–2 (not true to true) 0–1 (no victimization	7.24 (4.96)	.76	7.04 (4.97)	.75	7.43 (4.95)	.76	−1.77	−.08	.078
Bullying victimization (R)b	1	to yes, victimized most days)	62.91%	n/a	48.81%	n/a	51.19%	n/a	.43	−.03	.514
Adaptive coping (P)c	4	1–4 (definitely no to definitely yes)	2.79 (.44)	.69	2.82 (.43)	.68	2.76 (.44)	.71	2.94	.13	.003
Emotional control (P)d	4	1–4 (definitely no to definitely yes)	2.69 (.52)	.68	2.73 (.53)	.68	2.65 (.50)	.67	3.83	.17	<.001
Family-level variables
Poor family management (R)e	9	1–4 (definitely no to definitely yes)	1.72 (.45)	.77	1.65 (.45)	.78	1.78 (.44)	.76	−6.58	−.30	<.001
Family conflict (R)f	4	1–4 (definitely no to definitely yes)	2.22 (.71)	.68	2.24 (.71)	.66	2.20 (.71)	.70	1.24	.06	.215
Attachment to parent(s) (P)g	4	1–4 (definitely no to definitely yes)	2.95 (.61)	.79	2.93 (.62)	.79	2.97 (.60)	.79	−1.28	−.11	.201
School-level variables
Academic failure (R)h	2	1–4 (definitely no to definitely yes)	2.09 (.60)	.85	2.12 (.63)	.87	2.07 (.58)	.82	1.82	.08	.069
Low commitment to school (R)i	7	1–5 (never to almost always)	2.29 (.54)	.80	2.26 (.52)	.80	2.32 (.55)	.80	2.52	−.11	.012
Peer group-level variables
Interaction with prosocial peers (P)j	2	0–4 (none of my friends to 4 of my friends)	3.10 (.74)	.61	3.16 (.75)	.63	3.04 (.73)	.59	3.43	.16	<.001
Peer rewards for prosocial behavior (P)k	2	1–5 (no or very little chance to very good chance)	3.40 (.75)	.71	3.66 (.72)	.69	3.15 (.69)	.66	15.74	.71	<.001
Community-level variable
Community rewards for prosocial behavior (P)l	3	1–4 (definitely no to definitely yes)	2.35 (.75)	.74	2.36 (.78)	.75	2.34 (.71)	.73	.75	.03	.454

Note. α=Cronbach’s alpha.. t = t-statistic. SD = standard deviation. d = Cohen’s d. Statistically significant differences between states for continuous variables calculated using independent t-tests, and indicated by bolded p-values.

^aExample item: “I felt miserable or unhappy”

^bExample item: “Have you been bullied recently (teased or called names, had rumors spread about you, been deliberately left out of things, threatened physically or hurt)?”

^cExample item: “I think about the best ways to handle it.”

^dExample item: “I know how to relax when I feel tense.”

^eExample item: “Would your parents know if you did not come home on time?”

^fExample item: “We argue about the same things in my family over and over”

^gExample item: “Do you feel very close to your mother?”

^hExample item: “Are your school grades better than the grades/marks of most students in your class?”

ⁱExample item: “How often do you feel that the schoolwork you are assigned is meaningful and important?”

^jExample item: “In the past year (12 months), how many of your best friends have: “Tried to do well in school?”

^kExample item: “What are the chances you would be seen as cool if you: “Worked hard at school?”

^lExample item: “There are people in my neighborhood who are proud of me when I do something well.”

Control variables.

To reduce confounding due to background characteristics, five control variables, measured in adolescence, were included in the multivariable models. These control variables included the state in which adolescents resided, self-reported age and gender, and parent-reported family socioeconomic status (SES; from reports of highest level of maternal or paternal education and family income).³² Models also controlled for any report of DSH behavior during adolescence (DSH thoughts were not assessed during adolescence), when students were asked, “In the past year, have you every deliberately hurt yourself or done anything that you knew might have harmed you or even killed you?”.

Survey response accuracy

The accuracy of participants’ responses was measured during adolescence in each of the Grade 7–9 surveys. Responses were coded as questionable if participants reported (1) “I was not honest all of the time,” when asked how honest they were when completing the survey; (2) use of a fictitious drug in the past month (included in the survey for accuracy checking); and (3) drug use on > 120 occasions in the past month. Fifteen participants in 7^th, 35 in 8^th, and 27 in 9^th grade met the criteria for questionable responses and were excluded from the analyses. The final analytic sample size was N=1,945.

Data analysis

Stata SE software for Windows version 15.1³³ was used to (a) examine the prevalence of adolescent DSH behavior, and young adult DSH thoughts and behavior, and (b) conduct tests of state differences (t-tests) in adolescent risk and protective factors. Consistent with the probability proportionate to grade-level size sampling procedure used in the IYDS, sampling weights were calculated separately for each class as the inverse probability of selection in a particular class within the school grade. These sampling weights were used to calculate prevalence estimates and 95% confidence intervals [CIs] for adolescent DSH, and young adult DSH thoughts and behavior using design-based estimation of proportions. Next, differences in outcome prevalence estimates by gender and state were calculated. All prevalence estimates used robust “information sandwich” estimates of standard errors, with adjustment for clustering of students within schools. Model adjusted proportions were projected for state-by-gender groups with age considered as a covariate fixed at the mean age for each grade level. Pooled standard deviations³⁴ were used to calculate effect sizes in tests of state differences (t-tests). Zero-order correlation analyses were performed to identify highly correlated pairs or sets of variables that might result in collinearity in the subsequent multivariable analyses.

Multivariable models were estimated in Mplus version 8.2.³⁵ Full information maximum likelihood estimation was used in all analyses to minimize potential bias due to missing data.³⁵ To conduct the most conservative test possible of adolescent risk and protective factors for DSH thoughts and behavior among young adults, we tested three multivariable logistic regression models: 1) a risk factor-specific model; 2) a protective factor-specific model; and 3) a combined risk and protective factor model. Each model controlled for gender, state, age, family SES, and adolescent DSH behavior. A final step in the analyses examined the moderating role of state, using state-predictor interaction terms. State-predictor interaction terms were calculated by multiplying the state variable (coded 0 and 1) by risk and/or protective factors reaching statistical significance in the fully adjusted combined risk-protective factor model. To present the most parsimonious model with the least assumptions, the final logistic regression multivariable model included the addition of statistically significant state-predictor interaction terms.

Results

Rates of DSH thoughts and behavior

The analytic sample included 1,945 young adults (51% female; n=984 in Victoria), aged 13 to 17 years old (mean=14.01 years, standard deviation=0.43) at study outset. Rates of any DSH behavior reported during 7^th, 8^th, or 9^th grades were significantly higher for Victorian students (n=231, 23.55%; 95% confidence interval (CI)=20.66–26.71), compared to Washington State (n=124, 13.03%; CI=10.82–15.58) students (Figure 1). This finding likely reflects the higher rates of any adolescent DSH behavior among both Victorian males (n=110, 22.5%; CI=18.99–25.90) and females (n=121, 24.86%; CI=21.06–29.08), compared to Washington State males (n=54, 12.21%; CI=9.84, 15.04) and females (n=70, 13.84%; CI=11.19–17.00). For the combined Victorian-Washington State sample, there was no statistically significant difference in rates of any adolescent DSH behavior by gender.

Figure 1.

Prevalence of adolescent engagement in DSH, and young adult thoughts of DSH and engagement in DSH.

Notes: DSH deliberate self-harm. WASH = Washington State. VIC = Victoria. * p < .05; ** p < .001. Bolded estimates reflect significant differences between states for early adolescent engagement in DSH and young adult thoughts of DSH, and between sexes for young adult engagement in DSH.

Victorian participants (n=98, 11.30%; CI=9.27–13.70) reported significantly higher rates of DSH thoughts in young adulthood at age 25, compared to those in Washington State (n=64, 7.72%; CI=5.83–10.14). No statistically significant differences were noted in rates of young adult DSH thoughts by gender, although females reported slightly higher rates than males (10.69% vs. 8.07%, respectively). Rates of DSH behavior in young adulthood were low (2%−3%) and showed no statistically significant state difference. Female participants reported significantly higher rates of DSH behavior in young adulthood (n=32, 3.45%; CI=2.46–4.82), compared to their male counterparts (n=16, 1.89%; CI=1.10–3.25). No significant differences were observed for males and females in Washington State compared to Victoria for either young adult DSH thoughts or behavior. Finally, we noted that only 4.07% (95% CI=2.23%−7.31%) of adolescents who reported DSH behavior continued this behavior in young adulthood.

State differences in levels of early adolescent predictors

The percentage of missing data on the analyzed variables ranged from 0 to 12.8% (M=1.74%). State differences were evident for several adolescent predictors for young adult DSH thoughts and behavior. Washington State participants reported significantly higher rates of adaptive coping (p=.003) and emotional control (p<.001), compared to those in Victoria (Table 1). Results also showed significantly higher levels of prosocial peer interactions and peer rewards for prosocial behavior (p<.001) among Washington State, compared to Victorian participants. Less positive family management practices (p<.00) and lower commitment to school (p=.012) were found for participants in Victoria, compared to those in Washington State.

Adolescent predictors of young adult DSH thoughts and behavior

Intercorrelations between young adult DSH thoughts and behavior and all demographic and predictor variables were low (r=.01 to .20) and in the expected direction. The correlation between young adult DSH thoughts and behavior was high (.76). Since these two variables were used in separate analyses, both variables were retained for analysis. Young adult DSH thoughts were most strongly correlated with adolescent adaptive coping (r=−.17) and bullying victimization (r=.20). Young adult DSH behavior was most strongly correlated with female gender (r=.15). Intercorrelations between the analyzed predictors did not show evidence of multicollinearity, with no correlations >.80.

Young adult thoughts of DSH.

Risk factor-specific multivariable models demonstrated that young adult DSH thoughts were significantly predicted by adolescent depressive symptoms (adjusted odds ratio [AOR]=1.09; 95% CI=1.00–2.25; Table 2, model 1), living in Washington State (AOR=.64; CI=.45-.91), family SES (AOR=1.50; CI:1.00–2.25), and bullying victimization (AOR=1.53; CI=1.01, 2.31). For the protective factor-specific multivariable model, living in Washington State, higher levels of adolescent adaptive coping (AOR=.41; CI=.26-.65; Table 2, model 2), and community rewards for prosocial behavior (AOR=.76; CI=.59-.99) predicted lower odds of young adult DSH thoughts. For the combined risk-protective factor multivariable model (Table 2, model 3), these relationships remained robust. Depressive symptoms in adolescence (AOR=1.05; CI=1.00–1.09) increased risk, while higher levels of adolescent adaptive coping (AOR=.46; CI=.28-.74), higher levels of adolescent community rewards for prosocial behavior (AOR=.73; CI=.57-.93), and living in Washington State decreased risk for young adult DSH thoughts, after accounting for all other predictors in the model.

Table 2.

Multivariable logistic regression models investigating adolescent predictors of young adult thoughts of deliberate self-harm.

	Risk factors, Model 1 (N = 1,631)			Protective factors, Model 2 (N= 1,617)			Fully adjusted, Model 3 (N = 1,637)
Predictors	AOR (SE)	[95% CI]	p	AOR (SE)	[95% CI]	p	AOR (SE)	[95% CI]	p
Demographic factors
Age (years)	.93 (.21)	[.60, 1.45]	.762	.97 (.23)	[.61, 1.53]	.886	.97 (.22)	[.62, 1.52]	.909
Female	1.24 (.23)	[.86, 1.80]	.247	1.21 (.22)	[.84, 1.74]	.300	1.16 (.21)	[.80, 1.66]	.432
Washington State	.64 (.12)	[.45, .91]	.013	.65 (.12)	[.45, .95]	.025	.62 (.11)	[.43, .89]	.009
Family SES	1.50 (.31)	[1.00, 2.25]	.048	1.45 (.29)	[.98, 2.15]	.065	1.46 (.29)	[.98, 2.16]	.060
Adolescent DSH behavior	.90 (.21)	[.58, 1.41]	.660	.99 (.22)	[.65, 1.52]	.065	.90 (.20)	[.58, 1.39]	.626
Risk factors
Depressive symptoms	1.09 (.02)	[1.05, 1.13]	<.001				1.05 (.02)	[1.00, 1.09]	.046
Poor family management	1.26 (.29)	[.81, 1.97]	.303
Family conflict	.88 (.12)	[.67, 1.15]	.346
Academic failure	1.11 (.19)	[.79, 1.55]	.553
Low commitment to school	1.14 (.24)	[.76, 1.71]	.526
Bullying victimization	1.53 (.32)	[1.01, 2.31]	.042				1.49 (.31)	[.99, 2.24]	.057
Protective factors
Adaptive coping				.41 (.10)	[.26, .65]	<.001	.46 (.11)	[.28, .74]	.001
Emotion control				.82 (.16)	[.55, 1.20]	.307
Attachment to parents				.78 (.12)	[.57, 1.06]	.114
Prosocial peers				1.02 (.14)	[.79, 1.33]	.867
Peer rewards for prosocial behavior				.91 (.13)	[.69, 1.19]	.483
Community rewards for prosocial behavior				.76 (.10)	[.59, .99]	.040	.73 (.09)	[.57, .93]	.012
Pseudo R2	.057			.065			.071
Adjusted R2	.033			.041			.052

Note. The final hierarchical multivariable logistic regression models for risk and protective factors are presented here. SES = socioeconomic status; AOR = adjusted odds ratio; SE = standard error; CI = confidence interval. Female gender (coded 0 = male, 1 = female); Washington State (coded 1 = Washington State, 0 = Victoria). Pseudo R² = McFadden’s R²; Adjusted R² = McFadden’s Adjusted R². Bolded effects are significant at p < 0.05.

No state-predictor interactions were statistically significant in the risk factor-specific or the combined risk-protective factor multivariable models. For the protective factor-specific multivariable model, adolescent lower levels of community rewards for prosocial behavior were more strongly associated with young adult DSH thoughts in Washington State, compared to Victoria (AOR=1.69, CI=1.03–2.76).

Young adult DSH behavior.

In the risk factor-specific multivariable model, adolescent depressive symptoms (AOR=1.10; CI=1.02–1.17; Table 3, model 1) and poor family management strategies (AOR=3.01; CI=1.44–6.31) increased risk for young adult DSH behavior. No statistically significant protective factors for young adult DSH behavior were found in the protective factor-specific multivariable model. In the final multivariable model, less positive family management strategies demonstrated the only significant association with young adulthood DSH behavior (AOR=1.90; CI=1.01–3.60; Table 3, model 3). Tests of state differences revealed no significant cross-state differences in the associations between adolescent predictors and young adult DSH behavior.

Table 3.

Multivariable logistic regression models investigating adolescent predictors of young adult deliberate self-harm behavior.

	Risk factors, Model 1 (N = 1,631)			Protective factors, Model 2 (N = 1,617)			Fully adjusted, Model 3 (N = 1,637)
Predictors	AOR (SE)	[95% CI]	p	AOR (SE)	[95% CI]	p	AOR (SE)	[95% CI]	p
Demographic factors
Age (years)	1.43 (.55)	[.68, 3.09]	.346	1.37 (.54)	[.63, 2.98]	.421	1.39 (.52)	[.66, 2.90]	.382
Female	1.58 (.54)	[.80, 3.09]	.184	1.78 (.60)	[.91, 3.46]	.092	1.70 (.57)	[.88, 3.20]	.115
Washington State	.99 (.31)	[.53, 1.84]	.963	.98 (.33)	[.50, 1.91]	.955	.91 (.28)	[.49, 1.69]	.775
Family SES	.72 (.26)	[.35, 1.47]	.369	.77 (.28)	[.38, 1.59]	.485	.78 (.28)	[.39, 1.57]	.493
Adolescent DSH behavior	1.13 (.44)	[.53, 2.42]	.745	1.41 (.52)	[.68, 2.92]	.355	.99 (.38)	[.47, 2.11]	.983
Risk Factors
Depressive symptoms	1.10 (.04)	[1.02, 1.17]	.007				1.06 (.03)	[1.00, 1.12]	.053
Poor family management	3.01 (1.14)	[1.44, 6.31]	.003				1.90 (.62)	[1.01, 3.60]	.048
Family conflict	.71 (.17)	[.44, 1.15]	.167
Academic failure	.85 (.26)	[.47, 1.54]	.585
Low commitment to school	.56 (.21)	[.27, 1.16]	.118
Bullying victimization	.96 (.32)	[.49, 1.85]	.894
Protective Factors
Adaptive coping				.70 (.28)	[.32, 1.53]	.369
Emotion control				.98 (.34)	[.49, 1.92]	.942
Attachment to parents				.98 (.27)	[.57, 1.69]	.954
Prosocial peers				1.25 (.30)	[.77, 2.01]	.367
Peer rewards for prosocial behavior				.66 (.16)	[.41, 1.07]	.094
Community rewards for prosocial behavior				.82 (.19)	[.53, 1.28]	.387
Pseudo R2	.055			.037			.040
Adjusted R2	−.002			−.022			.018

Note. The final hierarchical multivariable logistic regression models for risk and protective factors are presented here. SES = socioeconomic status; AOR = adjusted odds ratio; SE = standard error; CI = confidence interval. Female gender (coded 0 = male, 1 = female); Washington State (coded 1 = Washington State, 0 = Victoria). Pseudo R² = McFadden’s R²; Adjusted R² = McFadden’s Adjusted R². Bolded effects are significant at p < 0.05.

Discussion

We sought to address gaps in the literature on longitudinal antecedents of young adult DSH by prospectively examining adolescent risk and protective factors associated with DSH thoughts and behavior during young adulthood among a cross-national sample. Findings showed that DSH thoughts during young adulthood were most strongly associated with living in Victoria (versus Washington State) and depressive symptoms during adolescence, while adaptive coping and community rewards for prosocial behavior during adolescence decreased risk of DSH thoughts in young adulthood. Young adult DSH behavior was most strongly associated with poor family management during adolescence. In the final models, only one significant difference emerged by state, showing adolescent lower levels of community rewards for prosocial behavior were more strongly associated with young adult DSH thoughts in Victoria, compared to Washington State. These findings extend previous research conducted by the authors regarding factors associated with DSH behavior over a 12-month period during adolescence.¹³

Consistent with the limited prospective research examining longitudinal relationships between risk factors and DSH thoughts and behavior, we identified adolescent depressive symptoms as a strong predictor of young adult DSH thoughts.^6,19 The enduring effects of depressive symptoms during adolescence on future DSH thoughts supports the need for prevention and early identification and management of depression among adolescents. Implementing universal depression screening in primary care offices and schools, where follow-up procedures, such as referrals to mental health treatment are available, may facilitate early identification of mental health problems and connection to needed care.²³ Early intervention for depression during adolescence is imperative and may help reduce mental health problems in the future.³⁶

Depressive symptoms during adolescence did not emerge as a strong predictor of DSH behavior during young adulthood. Instead, poor family management represented a more important risk factor associated with DSH behavior over time than depressive symptoms during adolescence. The construct of poor family management examined in the current study is consistent with a permissive parenting style, which is associated with poorer developmental outcomes.³⁷ In contrast, an authoritative parenting style, characterized by nurturing, responsive, and supportive behavior that includes, for example, setting firm limits for children, explaining rules, and reasoning through problems, while listening to a child’s viewpoint, is associated with improved outcomes among youth.³⁷ Researchers have identified weak and inconsistent (permissive) parenting styles as an important risk factor associated with DSH behavior among youth;³⁸ however, this represents the first study to examine this relationship prospectively. The association between poor family management during adolescence and DSH behavior in young adulthood demonstrates the enduring effects of family dynamics on youth development. Findings from this study suggest that educating parents about parenting styles associated with healthy youth development may represent a way to reduce young adult DSH behavior. Multiple tested-effective interventions have been shown to improve parents’ family management skills.³⁹ Implementation of these programs may help to reduce young adult DSH behavior.

In addition to knowledge about lasting effects of risk factors, this study adds to the limited research on protective factors associated with reduced risk of DSH thoughts over time. Our findings highlight the importance of teaching youth adaptive coping strategies to manage life stress. Teaching adolescents to use problem-oriented coping strategies, such as thinking about the best ways to handle a problem and not blaming or criticizing oneself, possibly within school-based social-emotional learning or mental health promotion programs,^40,41 could facilitate resilience to DSH thoughts and behavior during adolescence and in the future.^31,42 Further, this study extends previous research on community/neighborhood protective factors, as most research on DSH behavior has examined neighborhood connection and safety.^2,43 Our finding that recognition and reinforcement of good behavior by adults in one’s community/neighborhood represented a significant protective factor for DSH thoughts over time has implications for building adolescents’ adult support networks.

Study strengths and limitations

The longitudinal data collection from adolescence to young adulthood represents a significant strength of this study, allowing for temporal ordering of predictor and outcome variables. Another strength involves the cross-national sample that allowed us to compare findings among youth in two different countries. Although the results of this study are generalizable only to the states and ages examined, previous analyses suggest the cohorts are similar to national samples in the two countries.^23,44 In addition, the comprehensive measures assessed during adolescence enabled us to examine a wider range of risk and protective factors associated with DSH thoughts and behavior during young adulthood than previous studies. The survey measures have demonstrated good reliability and validity,^24,28,29 including longitudinal validity,^45,46 in large samples. Still, the measures were all self-report. Also, the DSH items combined “hurting” oneself, which would suggest NSSI behavior, with “killing” oneself, which suggests suicidal behavior. Though these behaviors may be associated,^9,10 they likely differ.⁴⁷ A limitation of our analysis involves the low base-rate of DSH behavior and potential lack of power to detect significant interactions in logistic regression models. In addition, we only examined direct effects of protective factors in the current study. However, as protective factors are theorized to have modifying influences on risk factors, models also should examine interactive and moderating effects once risk pathways are understood.

Study implications

Findings showed depressive symptoms and poor family management during adolescence were most strongly associated with DSH thoughts and behavior, respectively, during young adulthood. The enduring effects of these factors during adolescence on future DSH thoughts and behavior suggest the need for prevention and early identification and management of depression,^23,36 as well as parental education on effective parenting styles to promote healthy youth development.³⁹ In addition to knowledge about lasting effects of risk factors, this study adds to the limited research on protective factors associated with reduced risk of DSH thoughts and behavior over time, specifically use of adaptive coping strategies and community rewards for prosocial behavior. Findings regarding the importance of using a problem-focused coping style, rather than an emotion-focused coping style,^31,42 and receiving recognition and reinforcement for good behavior from adults in one’s community/neighborhood support and extend previous cross-sectional research on protective factors associated with reduced risk of DSH thoughts and behavior among adolescents.⁴³ Prevention and intervention programs should not only focus on managing depression and building/enhancing family connections and support, but also promote resilience through efforts to build/enhance adaptive, problem-focused coping strategies and connections to adults within one’s community who recognize and reward prosocial behavior.