Social Connectedness and Adolescent Suicide Risk

Abstract

Background:

Despite evidence of the importance of interpersonal connectedness to our understanding of suicide risk, relatively little research has examined the protective and buffering effects of connectedness among adolescents. Aims were to determine: 1) whether overall connectedness (composite of family, peer, school) and specific domains of connectedness were related to a lower likelihood of suicide attempts, and 2) whether these factors buffer the prospective risk of suicide attempt for high-risk subgroups (i.e., recent suicidal ideation and/or lifetime history of suicide attempt, peer victimization, or sexual and gender minority status).

Methods:

Participants were 2,897 adolescents (64.7% biological female), ages 12 to 17 (M=14.6, SD=1.6), recruited in collaboration with the Pediatric Emergency Care Applied Research Network (PECARN) from 14 emergency departments for the Emergency Department Screen for Teens at Risk for Suicide Study (ED-STARS). Suicide risk and protective factors were assessed at baseline; three- and six-month follow-ups were completed (79.5% retention). Multivariable logistic regressions were conducted, adjusting for established suicide risk factors.

Results:

Higher overall connectedness and, specifically, school connectedness were associated with decreased likelihood of a suicide attempt across 6 months. Overall connectedness and connectedness domains did not function as buffers for future suicide attempts among certain high-risk subgroups. The protective effect of overall connectedness was lower for youth with recent suicidal ideation or a suicide attempt history than for those without this history. Similarly, overall connectedness was protective for youth without peer victimization but not those with this history. Regarding specific domains, family connectedness was protective for youth without recent suicidal ideation or a suicide attempt history and peer connectedness was not protective for youth with peer victimization.

Conclusions:

In this large and geographically diverse sample, overall and school connectedness were related prospectively to lower likelihood of suicide attempts, and connectedness was more protective for youth not in certain high-risk subgroups. Results inform preventive efforts aimed at improving youth connectedness and reducing suicide risk.

Suicide is a leading cause of death among adolescents ages 12–17 in the United States (US) (Centers for Disease Control and Prevention [CDC], 2019). The suicide rate for this age group has increased over the past two decades, although a modest overall decrease was observed in 2019 and 2020 (Garnett, Curtin, & Stone, 2022). Moreover, 22% of US high school students report having considered making a suicide attempt and 10% report having made at least one suicide attempt within the past year (CDC, 2023). Notably, an increase in suicidal behavior (Cousien, Acquaviva, Kernéis, Yazdanpanah, & Delorme, 2021) and emergency department (ED) visits due to suicide attempts, especially among girls (Yard et al., 2021), were observed during the COVID-19 pandemic.

Suicide risk factors among adolescents are well established (Cha et al., 2018; Turecki et al., 2019). History of suicidal ideation and self-harm behaviors (King, Grupp-Phelan, et al., 2019; Nock et al., 2013), and severity of suicidal ideation (King, Grupp-Phelan, et al., 2019) are associated with elevated risk for subsequent suicidal behavior. Psychopathology (e.g., substance use, depressive disorders) (Burdzovic Andreas & Brunborg, 2017; Esposito-Smythers & Spirito, 2004), childhood trauma (Zatti et al., 2017), as well as interpersonal difficulties, such as victimization and lack of social connectedness (Arango, Opperman, Gipson, & King, 2016; Whitlock, Wyman, & Moore, 2014), have also been associated with risk for suicidal behavior. Youth who identify as sexual and gender minorities (SGM) report elevated suicidal ideation and behaviors and are at particularly high risk for suicide (CDC, 2023; Marchi et al., 2022).

Given the tragedy of suicide, as well as the breadth of factors that increase risk, it is critical to identify what may be protective for youth. As we look to address the alarming mental health impact of the COVID-19 pandemic (Office of the Surgeon General, 2021), and given the effect of the pandemic on isolation and disconnection, an improved understanding of connectedness and adolescent suicide risk may be particularly important.

Interpersonal constructs, such as social integration, belongingness, and connectedness, are integral to theoretical frameworks of suicide. In Durkheim’s social theory of suicide, lack of social connectedness is thought to be a primary suicide risk factor (Durkheim, 1897). In the Interpersonal Theory of Suicide (IPT), thwarted belongingness (i.e., unmet connectedness need) along with a sense of perceived burdensomeness are believed to predict suicidal ideation (Van Orden et al., 2010). Theoretical conceptualizations and, more generally, the importance of interpersonal connectedness to our understanding of suicide risk, are supported by research (Foster et al., 2017; King & Merchant, 2008; Whitlock et al., 2014). Notably, connectedness has been observed to serve both protective (i.e., decreasing the likelihood of an adverse outcome), and buffering (i.e., attenuating or lessening the impact of a known risk factor on an adverse outcome) functions (Kleiman, Riskind, Schaefer, & Weingarden, 2012).

Family and school connectedness were found to be protective against a range of health risk behaviors (e.g., suicidal thoughts and behaviors, substance use, sexual behaviors) in a cross-sectional study of high school students (Resnick et al., 1997). Moreover, school connectedness has been cross-sectionally associated with less suicidal behavior in a general school population as well as in specified subgroups of students with risk factors for negative outcomes (e.g., sexual minority, involvement in bullying) (Marraccini & Brier, 2017). Among a sample of peer victimized youth, family, school and community connectedness were negatively related to suicidal ideation longitudinally (Arango et al., 2019). In another study, increased peer connectedness following discharge from a psychiatric hospitalization was prospectively related to decreased likelihood of a suicide attempt, and increased family connectedness was related to decreased suicidal ideation (Czyz, Liu, & King, 2012).

Interpersonal connectedness across domains has also been shown to buffer the associations between vulnerabilities and suicidal thoughts and behaviors. In a cross-sectional sample of sexual minority youth, connectedness to an adult in the school setting attenuated the impact of electronic and peer victimization on suicidal behaviors (Duong & Bradshaw, 2014). Among youth involved in peer victimization, connectedness to a non-parent adult was protective of suicidal thoughts and behaviors cross-sectionally (Borowsky, Taliaferro, & McMorris, 2013). In a racially and ethnically diverse sample, connectedness in the school setting cross-sectionally buffered the impact of adverse childhood experiences on suicidal ideation for non-Hispanic White adolescents (Areba et al., 2021).

Few prospective studies have examined the relationship between connectedness and suicide attempts (Gunn, Goldstein, & Gager, 2018; Kim, Walsh, Pike, & Thompson, 2020; King, Grupp-Phelan, et al., 2019). However, one study found that, for youth who experienced electronic victimization, higher school connectedness was prospectively related to less suicidal behavior (Kim et al., 2020). In analysis of the National Longitudinal Study of Adolescent to Adult Health (Add Health) data collected between 1994 and 1996, Gunn et al. (2018) found that increases in school connectedness were prospectively related to decreased suicide attempts.

Notably, however, the data for this report were collected over two decades ago and information about suicidal behavior was only assessed for youth who reported suicidal ideation, despite the fact that a subset of the youth who engage in suicidal behavior have no known history of suicidal ideation (Rodway, Tham, Turnbull, Kapur, & Appleby, 2020). In the recent large-scale Emergency Department Screening for Teens at Risk for Suicide (ED-STARS) study, a multivariable prediction model indicated that past-week suicidal ideation, lifetime suicidal ideation severity, lifetime history of suicidal behavior, and low school connectedness predicted suicide attempts with a relatively high degree of accuracy within three months of the ED visit (King, Grupp-Phelan, et al., 2019).

The present study was designed to extend previous findings by examining relationships between connectedness and suicide attempts across a longer, six-month period. A longer-term examination of these relationships, even the addition of several months of data, is meaningful given the rapid developmental changes (cognitive, social, biological) observed in adolescence (Steinberg, 2005). Moreover, risk of suicidal behaviors notably increases in the transition into adolescence (Nock et al., 2013), a timeframe captured in our study. We also aimed to use an additional assessment timepoint to test for more enduring effects of connectedness. Using a large, geographically diverse US sample, we examined overall connectedness, as well as connectedness in three separate domains (family, peer, school) to better understand the context in which connectedness may be beneficial. Given that much of previous research has emphasized the examination of connectedness in a single domain (e.g., school or family), this study expands the extant literature by evaluating the role of connectedness in separate domains facilitating comparison of their potential differential impact. Adolescents were recruited in EDs, allowing us to capture a subset of youth (e.g., high-risk youth) who may not have been accessed in previous studies that have recruited youth in schools (e.g., Gunn et al., 2018). Our aims were to determine: 1) whether overall connectedness (composite of family, peer, school) and specific domains of connectedness have a protective effect on the likelihood of future suicide attempts across a six-months period, and 2) whether overall connectedness, and specific connectedness domains buffer the prospective risk of suicide attempt for subgroups of youth characterized by current suicidal ideation and/or lifetime suicide attempt histories, peer victimization, or SGM status.

Methods

Participants

Participants were 2,897 adolescents (64.7% biological female), ages 12 to 17 (M=14.6, SD=1.6), recruited as part of the ED-STARS Study, which was implemented in collaboration with the Pediatric Emergency Care Applied Research Network (PECARN), with the primary aim of developing the Computerized Adaptive Screen for Suicidal Youth (CASSY) (King et al., 2021). The current sample is comprised of the subsample of participants who completed 80% or more of the baseline survey (including at least one connectedness measure), and were part of the subgroup (oversampled for high-risk youth due to principal study aims) who were randomized to follow-up (King et al., 2021). Recruitment occurred in 13 EDs and via an ED-linked approach at an Indian Health Service (IHS) hospital. Adolescents identified race as White (51.6%), Black or African American (23.1%), American Indian or Alaska Native (3.2%), Asian (1.0%), Native Hawaiian or Other Pacific Islander (0.8%), and multi-racial (5.7%). Approximately 23% identified as Hispanic or of Latin American descent. Race and ethnicity information was collected in line with the US Bureau of Census categories. One third (32.3%) of participants identified as SGM; 30.4% identified as a sexual minority; and 6.1% identified as a gender minority. Approximately 23% (n=675) of participants presented with a primary psychiatric complaint (e.g., non-suicidal self-injury, suicidal ideation and behavior, depression, aggression) while the remainder presented with a primary medical complaint. Approximately 30% of participants’ mothers/stepmothers completed high school or less (n=863), while 63.2% completed at least some college (n=1,831). Approximately 40% of participants’ fathers/stepfathers completed high school or less (n=1,165), while 45.5% completed at least some college (n=1,319).

Retention at follow-up was significantly lower for biological females, Hispanic or Latin American adolescents, and adolescents whose parents had lower education levels (Table S1).

Procedure

Eligible adolescents were approached during randomly selected screening windows at PECARN ED sites. Adolescents at the IHS hospital were identified by daily review of ED admissions records and contacted at home. Adolescents were excluded if they participated in the study previously, were a ward of the state, did not speak English, were medically unstable or presented with severe cognitive impairment. Participants and parents/guardians completed the baseline survey using a tablet in the ED (at home for IHS participants). Follow-up assessments were conducted with adolescents and parents/guardians (if still a minor) at three- and six-months following the ED visit via computer-assisted phone interviews (with option of in-person interviews for IHS participants). Approximately 80% of participants had data available on post-baseline suicide attempt outcome (see Measures). Data were available at the 3- and 6-month follow-up for 1,809 participants (62.4%). Data were available at only the 3-month or 6-month follow-up for 295 (10.2%) and 200 (6.9%) participants, respectively. Follow-up data were obtained from parents and youth (n=1973, 68.1%), only parents (n=111, 3.8%), or only youth (n=220, 7.6%). Adolescent compensation was $15 and $25 for the baseline and each follow-up assessment, respectively. Institutional Review Board approval was obtained at all sites. Adolescents and parents provided written informed assent and consent. Additional information on study procedures can be found in King et al. (2021).

Measures

Baseline Assessment of Primary Study Variables.

Recent suicidal ideation was measured with one item (‘In the past week, have you been having thoughts about killing yourself?’) from the Ask Suicide-Screening Questionnaire (ASQ) (Horowitz et al., 2012), completed by adolescent self-report. Lifetime suicidal ideation severity was measured with the adolescent-completed Suicidal Ideation Severity Scale of the Columbia-Suicide Severity Rating Scale (C-SSRS) (Posner et al., 2011), and lifetime history of suicide attempts was assessed with adapted C-SSRS questions (‘Have you ever in your life made a suicide attempt?’ or ‘Have you ever in your life tried to harm yourself because you were at least partly trying to end your life?) in addition to the ASQ item (‘Have you ever tried to kill yourself?’). The use of multiple questions enabled us to capture a larger subset of adolescents with suicide-related experiences (Hatkevich et al., 2020).

Family connectedness was assessed using two items adapted from the Parent-Family Connectedness scale: ‘How much do people in your family understand you?’ ‘How much does your family pay attention to you?’ (Resnick et al., 1997). Peer connectedness was measured using two adapted items from Hemingway’s Adolescent Connectedness Scale: ‘I have friends I’m really close to and trust completely.’ ‘Spending time with my friends is a big part of my life.’ (Karcher & Sass, 2010). School connectedness was measured using two items from the School Connectedness Scale: ‘You feel close to people at your school.’ ‘You feel like you are part of your school’ (Resnick et al., 1997). Items were responded to on a 5-point Likert scale (e.g., “strongly disagree” to “strongly agree”; 1–5). Family, peer, and school connectedness scale scores were derived by calculating the sum of the two respective connectedness items (range 2–10). If one (or both) of the two items was missing a response, the corresponding scale was coded as missing.

A measure of overall connectedness was derived by calculating the mean of the six items measuring connectedness across the three domains (family, peer, school). At least five connectedness items were required. The Cronbach coefficient alpha among the six connectedness items was 0.78. As an exploratory, post hoc analysis, we considered an alternative to calculating the overall connectedness scale that would allow unequal weighting across the six items. Principal components analysis suggested that the weights for the first principal component would be sufficiently similar and, thus, using this approach would not be worth the loss of interpretability.

Peer victimization was assessed using 2 items from the Peer Victimization and Perpetration Questionnaire (Klomek, Marrocco, Kleinman, Schonfeld, & Gould, 2007; Nansel et al., 2001). Severity of victimization (i.e., not bullied, bullied once or twice, bullied sometimes or more frequently) was coded as the highest severity level endorsed for the two items (‘How often have you been bullied in school this term?’ ‘How often have you been bullied away from school property this term?’). Adolescents’ reports of peer victimization were coded as follows: no victimization, low victimization (occurring once or twice) and high victimization (occurring sometimes or more).

Gender identity was assessed via the self-reported item “What is your current gender identity (please select all that apply)” with the following response options: Male; Female; Trans male/Trans boy; Trans female/Trans girl; Genderqueer or Gender non-conforming; Other/not listed above. Sexual identity was assessed via the item “Do you see yourself as (please select all that apply)” with the following response options: Straight; Mostly Straight; Bisexual; Mostly gay/lesbian; Gay/lesbian; Queer; Unlabeled; Not Sure. SGM youth indicated current gender identity that was not cisgender and/or any sexual identity other than straight; otherwise, youth were assumed to not be SGM.

Primary Study Outcome.

Our primary outcome was a suicide attempt between baseline and 6-month follow-up. A suicide attempt was indicated by a known suicide (e.g., per medical chart review), adolescent-reported C-SSRS items (timeframes adapted to assess post-baseline period) or parent/adolescent report of the adolescent’s hospitalization or ED visit for a suicide attempt or suicide. If the suicide attempt status was only known through 3-month follow-up, the same status was assumed through month 6.

Statistical Analysis

Descriptive statistics for primary study variables were computed. Univariable linear regression models were used to examine associations between overall connectedness, as well as specific connectedness domains (family, peer, school), with suicide risk factors and demographic variables. Multivariable logistic regression models were conducted to examine the relationship between overall connectedness, and specific connectedness domains, with post-baseline suicide attempts, adjusting for lifetime suicide attempt history, lifetime suicidal ideation severity, and recent suicidal ideation.

Multivariable logistic regression models were conducted to evaluate the buffering effect of connectedness for three suicide risk factors (recent suicidal ideation and/or suicide attempt history, peer victimization, SGM) in the prediction of post-baseline suicide attempts. Each of these sets had four corresponding logistic regression models, one for overall connectedness, and one for each connectedness domain. The model effects consist of the main effect of connectedness, the main effect for membership in the particular group (e.g., SGM), and connectedness*group interaction effect, while also adjusting for lifetime suicidal ideation severity (adjusting for recent suicidal ideation and lifetime suicide attempt history in the SGM and peer victimization models). Initially, we analyzed an at-risk group which included only youth with recent suicidal ideation. Those analyses were not reported, as the group including youth with recent suicidal ideation and/or a lifetime suicide attempt history captures a broader and, likely more representative, group of youth at elevated risk for suicide. Participants with a missing value for the primary outcome, group membership, connectedness domain, or other model covariates were excluded from the corresponding rows/columns of Tables 3 and 4. Tests were considered statistically significant if the two-sided p-value was ≤ 0.05. Although post-hoc, we conducted power analyses to determine detectable effect sizes with 80% power. For the multivariable models conducted to examine the connectedness interaction (e.g., connectedness*SGM status), odds ratios ranged from 5.2 to 40.1 times the maximal-vs.-minimal-connectedness odds ratio in the more severe risk group than the corresponding connectedness odds ratio for those not in the risk group.

Table 3.

Multivariable logistic regressions assessing the protective effect of connectedness on post-baseline suicide attempt

	Overall Connectedness		Family Connectedness		Peer Connectedness		School Connectedness
Effect	SA Odds Ratio (95% CI)	p	SA Odds Ratio (95% CI)	p	SA Odds Ratio (95% CI)	p	SA Odds Ratio (95% CI)	p
Connectedness	0.74 (0.59, 0.93)	0.010	0.96 (0.87, 1.06)	0.443	0.93 (0.87, 1.00)	0.054	0.89 (0.82, 0.96)	0.004
Recent SI		<.001		<.001		<.001		<.001
Yes vs No	2.96 (1.94, 4.58)		3.24 (2.13, 4.98)		3.14 (2.07, 4.84)		2.92 (1.92, 4.52)
No response vs No	1.99 (1.06, 3.58)		2.15 (1.15, 3.86)		2.12 (1.14, 3.81)		2.01 (1.08, 3.62)
Lifetime SI Severity	1.27 (1.10, 1.47)	0.001	1.29 (1.11, 1.49)	<.001	1.29 (1.11, 1.49)	<.001	1.27 (1.10, 1.47)	0.001
Lifetime SA (Yes vs No)	3.75 (2.30, 6.31)	<.001	3.85 (2.36, 6.47)	<.001	3.86 (2.36, 6.49)	<.001	3.77 (2.30, 6.34)	<.001

Note: Based on multivariable logistic regressions. Four models depicted: one for each of the four connectedness measures. SI=Suicidal Ideation, SA=Suicide Attempt.

Ns for multivariable models using all listed predictors=2,286–2,292.

Table 4.

Multivariable logistic regressions assessing the buffering effects of connectedness on post-baseline suicide attempt

		Overall Connectedness		Family Connectedness		Peer Connectedness		School Connectedness
Risk Model	Effect	Odds Ratio (95% CI)	p	Odds Ratio (95% CI)	p	Odds Ratio (95% CI)	p	Odds Ratio (95% CI)	p
Recent SI OR Lifetime SA	Recent SI or Lifetime SA (Yes vs No)	0.49 (0.06, 4.84)	0.528	0.45 (0.08, 3.17)	0.403	1.94 (0.39, 12.32)	0.433	2.58 (0.57, 14.50)	0.226
	Connectedness	0.35 (0.19, 0.64)	0.001	0.66 (0.51, 0.85)	0.002	0.78 (0.62, 0.98)	0.031	0.76 (0.60, 0.96)	0.023
	Connectedness * Recent SI or Lifetime SA	2.14 (1.11, 4.09)	0.023	1.48 (1.13, 1.93)	0.005	1.20 (0.94, 1.51)	0.141	1.16 (0.90, 1.48)	0.239
	Lifetime SI Severity	1.33 (1.15, 1.54)	<.001	1.34 (1.17, 1.55)	<.001	1.36 (1.19, 1.58)	<.001	1.33 (1.16, 1.54)	<.001
SGM	SGM (Yes vs No)	0.55 (0.13, 2.30)	0.412	0.52 (0.14, 1.88)	0.314	1.20 (0.42, 3.52)	0.731	0.88 (0.34, 2.31)	0.798
	Connectedness	0.63 (0.45, 0.87)	0.006	0.89 (0.77, 1.02)	0.095	0.91 (0.81, 1.01)	0.087	0.84 (0.75, 0.95)	0.006
	Connectedness * SGM	1.40 (0.90, 2.16)	0.132	1.19 (0.99, 1.44)	0.069	1.04 (0.90, 1.21)	0.551	1.11 (0.94, 1.30)	0.210
	Recent SI		<.001		<.001		<.001		<.001
	Yes vs No	2.77 (1.81, 4.29)		3.03 (1.99, 4.68)		2.92 (1.91, 4.51)		2.75 (1.80, 4.26)
	No Response vs No	1.83 (0.98, 3.31)		1.97 (1.06, 3.56)		1.95 (1.04, 3.52)		1.87 (1.00, 3.37)
	Lifetime SI Severity	1.24 (1.07, 1.44)	0.004	1.26 (1.09, 1.46)	0.002	1.26 (1.09, 1.46)	0.002	1.24 (1.07, 1.44)	0.004
	Lifetime SA (Yes vs No)	3.67 (2.26, 6.16)	<.001	3.74 (2.30, 6.27)	<.001	3.80 (2.33, 6.38)	<.001	3.67 (2.25, 6.17)	<.001
Peer Vic.	Peer Victimization		0.022		0.809		0.016		0.123
	High vs None	0.10 (0.02, 0.52)		0.62 (0.15, 2.63)		0.18 (0.05, 0.58)		0.34 (0.12, 1.02)
	Low vs None	0.40 (0.06, 2.74)		0.80 (0.14, 4.57)		0.49 (0.11, 2.05)		0.84 (0.22, 3.13)
	Connectedness	0.52 (0.36, 0.74)	<.001	0.91 (0.77, 1.07)	0.244	0.81 (0.72, 0.91)	<.001	0.81 (0.71, 0.92)	0.001
	Connectedness * Peer Victimization		0.010		0.604		0.003		0.058
	Connectedness * (High vs None)	2.16 (1.31, 3.57)		1.12 (0.90, 1.38)		1.32 (1.13, 1.56)		1.25 (1.04, 1.50)
	Connectedness * (Low vs None)	1.41 (0.79, 2.52)		1.07 (0.83, 1.38)		1.15 (0.94, 1.40)		1.07 (0.86, 1.32)
	Recent SI		<.001		<.001		<.001		<.001
	Yes vs No	2.95 (1.93, 4.56)		3.13 (2.06, 4.83)		3.06 (2.01, 4.72)		2.91 (1.90, 4.51)
	No Response vs No	1.91 (1.02, 3.46)		2.09 (1.12, 3.75)		2.02 (1.08, 3.64)		1.99 (1.06, 3.58)
	Lifetime SI Severity	1.25 (1.08, 1.45)	0.003	1.28 (1.11, 1.48)	<.001	1.28 (1.10, 1.48)	<.001	1.25 (1.08, 1.46)	0.002
	Lifetime SA (Yes vs No)	3.72 (2.28, 6.24)	<.001	3.76 (2.31, 6.31)	<.001	3.76 (2.30, 6.34)	<.001	3.80 (2.32, 6.39)	<.001

Note: Based on multivariable logistic regressions. Twelve models depicted: three risk model types for each of four connectedness measures. SI=Suicidal Ideation, SA=Suicide Attempt, SGM=sexual and/or gender minority, Vic.=Victimization.

Ns for multivariable models =2,280–2,294.

Results

Descriptive statistics

At the baseline assessment, 33.4% of adolescents (n=967) reported a lifetime suicide attempt history, and approximately 20% (n=568) reported recent (past week) suicidal ideation. Approximately 39% (n=1,137) reported recent suicidal ideation and/or a suicide attempt history at baseline. Among the 2,304 youth with prospective suicide attempt status known, 7.1% (n=164) had a reported suicide attempt between baseline and 6-month follow-up.

Participants’ mean overall, family, peer, and school connectedness scores were 3.6 (SD=0.78), 7.5 (SD=1.87), 7.6 (SD=2.11), and 6.7 (SD=2.11), respectively, at baseline. Pairwise Spearman correlations between connectedness scales were as follows: family with peer connectedness was 0.21 (n=2,885), family with school connectedness was 0.41 (n=2,879), and peer with school connectedness was 0.46 (n=2,877). Approximately 54% of adolescents (n=1,577) denied peer victimization; 21.2% (n=614) endorsed victimization occurring once or twice (low peer victimization); 24.1% (n=698) endorsed that victimization occurred sometimes or more frequently (high peer victimization).

Relationships between demographics, connectedness, and suicide risk factors

The bivariate associations between overall connectedness and connectedness domains with demographic characteristics, and suicide risk factors are displayed in Tables 1 and 2. Notable associations are described below. Biological female adolescents reported significantly lower overall, family and school connectedness than biological males. Black adolescents reported significantly lower peer connectedness than White adolescents, and Hispanic and Latin American adolescents reported lower peer connectedness than non-Hispanic and Latin American adolescents. Participant age was negatively related to connectedness, with older youth reporting lower overall, as well as family, peer, and school connectedness. Adolescents who were not attending school reported less overall, as well as family and school connectedness than middle school students, and less overall and school connectedness than high school students. Maternal and paternal education were positively related to overall, as well as family, peer, and school connectedness. As an example, adolescents who had mothers who were college graduates or completed professional training reported higher overall, family, peer, and school connectedness than adolescents who had mothers who had a high school degree or less. Moreover, receipt of public assistance was negatively related to overall, family, peer, and school connectedness. Table S2 depicts post-hoc results examining pairwise comparisons for significant bivariate associations (e.g., parent education and school connectedness) between connectedness domains and demographic characteristics or suicide risk factors.

Table 1.

Bivariable associations between demographic characteristics and connectedness

	Overall Connectedness		Family Connectedness		Peer Connectedness		School Connectedness
	Estimate (95% CI)	p	Estimate (95% CI)	p	Estimate (95% CI)	p	Estimate (95% CI)	p
Sex		<.001		<.001		0.072		<.001
Male	Reference		Reference		Reference		Reference
Female	−0.24 (−0.31, −0.18)		−0.66 (−0.81, −0.50)		−0.17 (−0.35, 0.02)		−0.63 (−0.81, −0.46)
Age	−0.05 (−0.07, −0.03)	<.001	−0.10 (−0.15, −0.05)	<.001	−0.06 (−0.11, −0.00)	0.048	−0.14 (−0.19, −0.08)	<.001
Race		0.173		0.559		<.001		0.177
White	Reference		Reference		Reference		Reference
Black or African American	−0.05 (−0.13, 0.03)		−0.06 (−0.25, 0.13)		−0.40 (−0.61, −0.18)		0.13 (−0.09, 0.35)
Multi-racial	−0.13 (−0.27, 0.00)		−0.24 (−0.56, 0.09)		−0.24 (−0.62, 0.13)		−0.31 (−0.69, 0.06)
Other	0.07 (−0.08, 0.22)		0.01 (−0.35, 0.37)		0.13 (−0.28, 0.53)		0.25 (−0.16, 0.66)
Unknown/Missing	−0.04 (−0.14, 0.05)		0.07 (−0.16, 0.30)		−0.41 (−0.67, −0.14)		0.07 (−0.20, 0.33)
Ethnicity		0.002		<.001		0.028		0.064
Hispanic or Latino	−0.01 (−0.09, 0.07)		0.17 (−0.02, 0.35)		−0.23 (−0.45, −0.02)		0.02 (−0.19, 0.24)
Not Hispanic or Latino	Reference		Reference		Reference		Reference
Unknown	−0.18 (−0.28, −0.08)		−0.49 (−0.72, −0.25)		−0.28 (−0.55, −0.01)		−0.31 (−0.58, −0.04)
Child’s grade in school		<.001		<.001		0.134		<.001
5th - 8th grade	Reference		Reference		Reference		Reference
9th - HS graduate	−0.15 (−0.22, −0.08)		−0.36 (−0.52, −0.20)		−0.09 (−0.28, 0.10)		−0.46 (−0.64, −0.27)
Not in school	−0.80 (−1.27, −0.34)		−1.22 (−2.32, −0.12)		−1.18 (−2.45, 0.09)		−2.42 (−3.67, −1.16)
Unknown	−0.23 (−0.40, −0.06)		−0.54 (−0.95, −0.13)		−0.36 (−0.82, 0.11)		−0.53 (−0.99, −0.07)
Mother/Stepmother Education		<.001		<.001		<.001		<.001
HS graduate or less	Reference		Reference		Reference		Reference
Some college/technical training	0.07 (−0.01, 0.16)		0.13 (−0.08, 0.33)		0.26 (0.02, 0.49)		0.03 (−0.20, 0.26)
College/professional training	0.19 (0.11, 0.27)		0.29 (0.11, 0.48)		0.45 (0.24, 0.67)		0.39 (0.17, 0.60)
Don’t know/NA//Missing	−0.14 (−0.28, 0.00)		−0.38 (−0.71, −0.04)		−0.30 (−0.67, 0.08)		−0.16 (−0.54, 0.22)
Father/Stepfather Education		<.001		0.004		<.001		<.001
HS graduate or less	Reference		Reference		Reference		Reference
Some college/technical training	0.09 (−0.00, 0.18)		0.11 (−0.10, 0.32)		0.31 (0.07, 0.56)		0.09 (−0.15, 0.34)
College/professional training	0.17 (0.10, 0.25)		0.22 (0.03, 0.41)		0.40 (0.19, 0.61)		0.41 (0.20, 0.63)
Don’t know/NA//Missing	−0.12 (−0.22, −0.02)		−0.22 (−0.46, 0.02)		−0.26 (−0.53, 0.01)		−0.23 (−0.51, 0.04)
Receipt of Public Assistance		<.001		0.003		<.001		<.001
No	Reference		Reference		Reference		Reference
Yes	−0.16 (−0.23, −0.09)		−0.20 (−0.36, −0.05)		−0.41 (−0.59, −0.23)		−0.34 (−0.51, −0.16)
Unknown	−0.24 (−0.39, −0.08)		−0.52 (−0.89, −0.14)		−0.54 (−0.97, −0.12)		−0.38 (−0.80, 0.04)

Note: Based on unweighted univariable linear regressions. P-values are based on joint F-tests, while effects and confidence intervals are either for the slope or difference in means of the indicated level from the reference level. HS=High School.

Ns for bivariable associations =2,293–2,300.

Table 2.

Bivariable associations between baseline suicide risk factors and connectedness

	Overall Connectedness		Family Connectedness		Peer Connectedness		School Connectedness
	Estimate (95% CI)	p	Estimate (95% CI)	p	Estimate (95% CI)	p	Estimate (95% CI)	p
Lifetime SI Severity	−0.15 (−0.17, −0.14)	<.001	−0.39 (−0.42, −0.35)	<.001	−0.17 (−0.21, −0.13)	<.001	−0.34 (−0.38, −0.30)	<.001
Lifetime SA		<.001		<.001		<.001		<.001
No	Reference		Reference		Reference		Reference
Yes	−0.52 (−0.59, −0.46)		−1.38 (−1.53, −1.23)		−0.58 (−0.76, −0.39)		−1.15 (−1.33, −0.97)
Recent SI		<.001		<.001		<.001		<.001
Yes	−0.65 (−0.73, −0.58)		−1.55 (−1.73, −1.37)		−0.82 (−1.04, −0.60)		−1.56 (−1.77, −1.35)
No	Reference		Reference		Reference		Reference
No response	−0.57 (−0.69, −0.46)		−1.44 (−1.72, −1.17)		−0.75 (−1.08, −0.42)		−1.25 (−1.56, −0.93)
Recent SI or Lifetime SA		<.001		<.001		<.001		<.001
No	Reference		Reference		Reference		Reference
Yes	−0.56 (−0.63, −0.50)		−1.46 (−1.61, −1.32)		−0.62 (−0.79, −0.44)		−1.28 (−1.45, −1.11)
SGM		<.001		<.001		<.001		<.001
No	Reference		Reference		Reference		Reference
Yes	−0.38 (−0.45, −0.32)		−0.95 (−1.11, −0.79)		−0.37 (−0.56, −0.19)		−0.96 (−1.14, −0.78)
Frequency of Peer Vic.		<.001		<.001		<.001		<.001
None	Reference		Reference		Reference		Reference
Low	−0.12 (−0.19, −0.04)		−0.23 (−0.42, −0.03)		−0.09 (−0.31, 0.13)		−0.39 (−0.60, −0.17)
High	−0.53 (−0.60, −0.45)		−0.92 (−1.10, −0.74)		−0.79 (−1.00, −0.58)		−1.45 (−1.65, −1.24)

Note: Based on unweighted univariable linear regressions. P-values are based on joint F-tests, while effects and confidence intervals are either for the slope or difference in means of the indicated level from the reference level. SI=Suicidal Ideation, SA=Suicide Attempt, SGM=sexual and/or gender minority, Vic.=Victimization.

Ns for bivariable associations =2,289–2,300.

Relationships between connectedness and suicide risk factors were all in the expected directions. Notably, overall connectedness, and connectedness domains (family, peer, school) were negatively related to lifetime suicidal ideation severity, lifetime suicide attempt history and recent suicidal ideation. Overall connectedness and each connectedness domain were negatively related to peer victimization. SGM adolescents reported significantly lower overall, family, peer, and school connectedness than non-SGM adolescents.

Protective role of connectedness

In the multivariable models predicting post-baseline suicide attempts, overall connectedness and school connectedness were related to fewer post-baseline suicide attempts (Table 3). In analogous multivariable models, family and peer connectedness were not related to post-baseline suicide attempts.

Buffering role of connectedness

Separate models examining the buffering impact of overall connectedness or connectedness domains on the prospective risk of suicide attempts for youth in defined risk groups (recent suicidal ideation and/or lifetime suicide attempt, peer victimization, SGM) are detailed in Table 4. Figure 1 displays the significant interactions between risk group and connectedness. The protective effect of overall connectedness was lower among youth with, as compared to those without, recent suicidal ideation and/or a suicide attempt history. Similarly, family connectedness was only significantly protective of post-baseline suicide attempts among youth without recent suicidal ideation or a suicide attempt history. Though there were significant interactions between overall connectedness and peer connectedness with peer victimization in relation to post-baseline suicide attempts, overall connectedness and peer connectedness did not buffer the impact of low and high peer victimization on post-baseline suicide attempts.

Figure 1.

Examination of the buffering impact of connectedness on post-baseline suicide attempt

Note: Figure depicts select alternate parametrizations (each connectedness main effect combined with the associated interaction effect) from the four Table 4 multivariable models that have significant interaction effects to show risk group-specific effects for connectedness. The overall connectedness measure is a composite of family, peer, and school connectedness.

ORs are from multivariable models for which Ns=2,282–2,294.

Discussion

Overall connectedness and school connectedness were protective for youth evaluated in an ED as they were related to a reduced likelihood of a suicide attempt during a six-month period, even when controlling for other established suicide risk factors (recent suicidal ideation, lifetime suicide ideation severity, lifetime suicide attempts). However, overall connectedness and connectedness domains did not generally function as a buffer, defined as lessening the impact of a known risk factor on an adverse outcome. Specifically, the positive impact of overall connectedness was lower among youth with recent suicidal thinking or a suicide attempt history and family connectedness was not protective for youth with this history. Overall and peer connectedness did not buffer risk for future suicide attempts among youth reporting peer victimization. Results highlight the key protective role of connectedness and are in line with theoretical conceptualizations of suicide risk (Durkheim, 1897; Van Orden et al., 2010), as well as previous research (King & Merchant, 2008; Whitlock et al., 2014). Our findings underscore the value that overall and school connectedness can serve for adolescents and extend previous findings in this sample (King, Grupp-Phelan, et al., 2019). Gunn et al. (2018), in their longitudinal investigation of the Add Health data collected from 1994 to 1996, also found that increases in school connectedness were related to fewer suicide attempts at a one-year follow-up. We extended these findings by sampling a more contemporary cohort (2015 – 2018) of adolescents and examining relationships between connectedness and adolescent suicide attempts among subgroups of youth at elevated risk for suicide. Additionally, we had access to suicide attempt outcome data on all participants (not just those reporting suicidal ideation), and examined the protective impact of a connectedness composite, which averages connectedness across family, peer, and school domains.

School connectedness was the only specific connectedness domain related to fewer suicide attempts. School connectedness is critical to healthy adolescent development and has been linked to health risk and health-promoting behaviors, including reduced substance use, increased physical activity (Weatherson et al., 2018), and more productive coping (Frydenberg, Care, Chan, & Freeman, 2009). Schools are well positioned to serve a protective role through the identification of those at-risk and through the provision of support. Our measure of school connectedness likely reflects relationships between youth and a variety of individuals within the school setting (e.g., teachers, peers, administrators, support staff, etc.). Results highlight the importance of bolstering a sense of connectedness at the one-on-one level, such as between a youth and a specific teacher. A more nuanced understanding of the potentially varied protective impact of quantity and types of relationships is warranted, specifically aimed at disentangling the construct of school connectedness into components including feelings towards teachers, peers, and school (García-Moya, Bunn, Jiménez-Iglesias, Paniagua, & Brooks, 2018). Additionally, school settings may positively impact a youth’s development broadly (e.g., by building self-confidence, self-efficacy, problem-solving skills). It may be that these effects cascade in ways that decrease suicide risk. Notably, connectedness may serve as a proxy for overall well-being (i.e., youth with stronger psychological adjustment may be better equipped to cultivate connectedness).

School-based programs can positively impact a sense of belonging in the school setting (Allen et al., 2021). For example, supportive relationships with teachers and peers, connectedness, and school engagement mediated the impact of a school-based health promotion intervention, and these factors were related to fewer depression symptoms, as well as lower peer victimization and bullying perpetration (Singla, Shinde, Patton, & Patel, 2021). Future studies should examine the mechanisms (e.g., willingness to seek support, access to resources) through which school connectedness may lower suicide risk and identify which youth may benefit most from connectedness-focused interventions. In addition, improving school environments (e.g., school safety) is likely fundamental to the development of school connectedness. The CDC’s “Whole School, Whole Community, Whole Child” guidelines include recommendations at the school and community level, as well as at the individual services level, with the goal of supporting youth’s cognitive, socio-emotional, and physical development (CDC, 2014).

We found that the protective effect of overall connectedness was lower for youth with a lifetime suicide attempt history or recent suicidal ideation than those without this history. These results highlight that the positive effect of connectedness appears to be less pronounced as the level of risk increases. It is likely that youth in the defined high-risk subgroup had additional symptomatology contributing to their risk status. Nonetheless, results point to the importance of universal prevention strategies aimed at promoting connectedness and may also indicate which youth are most likely to benefit from interventions aimed at strengthening social support (King, Arango, et al., 2019).

It is notable that the more severe the report of peer victimization, the less protection was conferred by overall connectedness and peer connectedness. This finding suggests that it may be helpful to gauge the severity of peer victimization, which has been related to increased suicide risk (Arango et al., 2016; Czyz et al., 2012; Klomek et al., 2007). It may be that at lower levels of peer victimization, having a friend to sit with or who will accompany a youth between classes may be buffering, while at higher levels of peer victimization this may not be enough. There may also be common factors (e.g., social deficits, anxiety) that place youth at risk of being peer victimized and that impact youth’s ability to build or maintain social connections. Alternatively, some youth may be connected to a peer group who experience greater victimization, and in those instances, a sense of closeness to that group could also confer risk. Moreover, when co-rumination or co-brooding are present peer interactions may increase risk (Bastin, Mezulis, Ahles, Raes, & Bijttebier, 2015), and this may be pronounced among youth drawn to peers who are also managing mental health concerns. Additional work examining the interplay between subtypes of peer victimization, domains of connectedness, and suicidal behavior is warranted. For example, a cross-sectional study found a moderating role for parental involvement in the relationship between victimization and suicidal behavior (Peprah et al., 2023). Similarly, a multinational, cross-sectional study indicated that parent support moderated the impact of verbal victimization on youth suicide attempts, but not the impact of physical or relational victimization (Barzilay et al., 2017).

Finally, adolescent demographic factors were significantly related to their perceived levels of overall connectedness, as well as connectedness in specific domains. Biological females, SGM, and Black or African American youth reported lower connectedness than biological males, non-SGM, and White youth, respectively. Moreover, parental education was positively related, while receipt of public assistance was negatively related, to connectedness. These findings are in line with previous research indicating that biological females, youth from lower socio-economic backgrounds, and sexual minority youth report lower connectedness (Alivernini et al., 2019; Joyce, 2015). Lower connection across domains is likely compounding (or may be an outcome of) the substantial challenges and injustices that these minoritized populations face. Community-based collaborative approaches with stakeholder input may help us to better understand connectedness challenges and how connectedness can be bolstered within subgroups of youth.

Findings are particularly relevant given the substantial effects of COVID-19 on youth’s interpersonal connections and well-being. COVID-19-related safety precautions fundamentally altered the social worlds of youth, including the format of school, the ways youth connect with peers, and increases in family-level stress. Though our data were collected prior to the COVID-19 pandemic, our findings related to the protective role of school connectedness are relevant. Given challenges inherent in the transition to virtual learning, future work should consider how connectedness in the school setting can be cultivated through different learning contexts (e.g., in-person, virtual, hybrid models).

Findings should be considered in the context of study limitations. Despite a large sample, as well as oversampling for high-risk youth, our study could have benefited from additional power to detect potential significant interaction effects. Additional longitudinal studies focused on SGM and high-risk youth (e.g., history of suicidal thinking and behavior, history of peer victimization), with sufficient power to detect modest but clinically significant protective effects, are warranted. Notably, brief scales were used to assess constructs of interest, although each of these has demonstrated predictive validity for suicide attempts (King, Grupp-Phelan, et al., 2019). Nevertheless, more comprehensive measures of key constructs, such as to examine subtypes of peer victimization (e.g., relational, verbal, physical, electronic), and their relationships with suicide attempts, may be informative. It is important to note that we did not specifically measure connectedness in the context of social media. It may be that youth victimized in the school setting struggle to feel connected to peers or adults in that context and may instead cultivate connectedness via social media. Though items assessing connectedness were phrased broadly (e.g., “I have friends I’m really close to and trust completely”) and likely capture virtual relationships, research evaluating the impact of connectedness facilitated through social media is warranted. We did not examine how connectedness changed following receipt of ED services. Services and supports mobilized following an ED visit, especially for youth presenting for psychiatric concerns, may impact self-reported connectedness. Moreover, questions used to assess connectedness across domains are self-reported perceptions of connectedness and we did not measure instrumental support or gather the perspective of other informants (e.g., teachers, peers). Additionally, self-reported perceptions of connectedness may be impacted by state level factors (e.g., current mood) that were not examined. Finally, though we had a relatively diverse sample, we did not have nationally representative data, which may limit the generalizability of study findings.

Conclusions

In conclusion, findings underscore the importance of overall connectedness and connectedness in the school setting to youth suicide prevention. This study highlights that at-risk groups, including as SGM youth, as well as Black or African American youth, experience lower levels of connectedness. However, the suggested positive impact of overall and family connectedness was lower and not significant, respectively, for youth with recent suicidal thinking or a suicide attempt history, when compared to those without this history. For youth with more severe peer victimization, less protection was conferred by overall and peer connectedness. Findings are important given disruptions in connectedness as a result of the COVID-19 pandemic and may inform suicide prevention strategies.

Supplementary Material

Supinfo

Table S1. Demographic characteristics by retention status.

Table S2. Post-hoc analyses of pairwise comparisons for significant bivariate associations between connectedness and either demographic characteristics or suicide risk factors.

Key points.

• Suicide is a leading cause of death among youth. Social integration, belonginess and connectedness are integral to theoretical frameworks of suicide, and research supports the importance of interpersonal connectedness to our understanding of suicide risk.

• This study extends upon previous literature by examining the prospective protective and buffering relationships between overall connectedness, and connectedness in specific domains, and suicide attempts.

• Among a geographically diverse sample of youth, we found that overall connectedness, and specifically connectedness in the school setting, were related to a decreased likelihood of a suicide attempt across 6 months. Overall connectedness and connectedness domains did not function as buffers for future suicide attempts among high-risk subgroups of adolescents.

• Findings have the potential to inform prevention and intervention approaches. Particularly approaches aimed at improving youth connectedness.