Exploring the role of social networks in youth mental health

Abstract

Social network analysis (SNA) offers an innovative approach to understand youths’ complex social support systems. Growing work highlights links between youths’ social networks and mental health. We used SNA to examine associations between social network characteristics and psychological empowerment, depression, anxiety, and post-traumatic stress among youth living in areas with concentrated disadvantage. We administered egocentric social network surveys in which youth (i.e., egos) enumerated individuals in their social network (i.e., alters), answered questions about each alter, and described how alters were connected. We measured psychological empowerment and mental health symptoms with validated instruments. There were 111 youth in the analytic sample (mean age: 14.6 years [SD: 1.5]; female: n = 63 [57%]; Black: n = 79 [71%]; multiracial/other: n = 23 [21%]). Among 477 alters, most were parents/guardians, siblings, or extended family. Higher betweenness centrality (i.e., a measure of connectivity) was associated with greater anxiety and post-traumatic stress after adjusting for other variables. Findings illustrate that network connectivity may exert varying influences on mental health symptoms. Ongoing work is needed to describe the dynamic nature of youth support systems and how these relationships can be leveraged to promote well-being.

Introduction

Adolescent mental health and well-being represents a major public health priority. Worldwide, more than 10% of all adolescents experience mental health concerns, a figure which is compounded by long-term impacts on development, inequities in service access, and associated economic costs [67]. Thus growing work has recognized the importance of fostering individual, interpersonal, and environmental protective factors to support youths’ psychological development and prevent mental health morbidity [53, 66]. Following the COVID-19 pandemic, which has resulted in drastic increases in the rates of anxiety, depression, and substance use among adolescents [27, 43], characterizing such protective factors remains critical. For young people who face marginalization (e.g., discrimination, poverty, violence exposure), mental health challenges, including those introduced by the pandemic, may be exacerbated by structural forms of oppression [3, 68], necessitating novel theoretical approaches to understand their experiences and development of interventions which respond to their unique sociocultural context.

Interpersonal relationships represent one contextual factor that may differ for youth who are marginalized. Social supports (e.g., having someone to depend on) play an important role in promoting adolescent health and well-being and reducing involvement in health-affecting risk behaviors. For example, adolescents with greater social support are less likely to engage in problematic substance use, violence and delinquency, and other health risk behaviors [26, 34, 63]. In addition, stronger social ties foster greater academic achievement, adaptive coping skills, and resilience [10, 14, 60]. However, all social supports may not be protective. Having friends who engage in maladaptive health decision-making (e.g., substance use) may actually increase a young person’s risk of similar risk behaviors. Youth who are marginalized may also lack traditional forms of social support (e.g., from a parent or caregiver). Therefore, describing both the nature and quality of interpersonal relationships in marginalized young people is essential.

Social network analysis (SNA) is an innovative analytic approach in which connections between individuals, places, and/or events are mapped and visualized [49, 61]. SNA allows for more nuanced understanding of the ways in which dynamic interpersonal connections influence adolescent health. Beyond describing the quality of social connections, SNA offers insight into the composition and structure of social ties. SNA may be particularly relevant for characterizing social networks of youth experiencing marginalization, where relationships with extended family members, non-familial mentors, and other support figures (e.g., those with a shared identity) hold greater salience [19, 22, 48, 64]. In particular, applying SNA among youth exposed to violence and discrimination may illuminate patterns of information transfer or resource-sharing (i.e., instrumental support) that are not captured through traditional measures of social support.

Prior work applying SNA to adolescents has demonstrated the complex interactions between network characteristics, mental health, and risk behaviors. Certain network structural features are linked to better mental health. Network embeddedness (i.e., having close social ties with resource-sharing) positively impacts youths’ life satisfaction [40], and greater network degree (i.e., number of friends) decreases risk of depression [55], potentially due to increasing sense of belonging. Among sexual minority youth, decreased connectedness (i.e., how close an individual is to others in their network) and lower eigenvector centrality (i.e., having highly connected network members) predict more severe depressive symptoms [21], suggesting influences of isolation due to stigma and discrimination. Other qualitative aspects of youths’ networks, such as instrumental support, have shown benefit for psychological functioning, even buffering against the effects of minority stress [65].

Yet, as described above, social connections may not universally be protective. Youth with high betweenness centrality (i.e., those acting as liaisons between different network members) are more likely to engage in substance use [16]. These social “liaisons” may be exposed to a larger number of peer groups, putting them at higher risk for associated risk behaviors [23]. Network density (i.e., how interconnected individuals are in a network) can also be positive or negative, depending on whether these groups enable self-verification [18, 59]. However, many existing SNA studies in adolescents have focused on peer connections, usually within the school setting, which may neglect relationships with non-familial mentors and other community members. In addition, little work has examined how social networks contribute to other elements of youth well-being.

Well-being is a multifaceted construct that includes elements beyond mental health alone, such as self-esteem, connectedness, agency, civic participation, and resilience [46]. Psychological empowerment (PE), referring to an individual’s development of critical consciousness, self-efficacy, and perception of control, is a critical component of well-being, and may represent an important protective factor among youth experiencing marginalization [69, 70]. PE has previously been linked to greater self-esteem among youth and may foster goal-setting and hopes for the future [42, 54]. PE can be supported through participatory research activities and reciprocal youth-adult partnerships, making it an important factor in evaluating youth-led programs and interventions [1, 8]. Growing work suggests PE may be influenced by social networks. While existing frameworks have distinguished individual, organizational, and community forms of empowerment [50, 71], “relational empowerment” describes how interpersonal relationships may support PE through meaningful social interactions, engagement in collective action, transitive skills-building, and network mobilization [9, 31]. Drawing from feminist theory [58], this expansion of PE contrasts with a hegemonic focus on individuality and autonomy, instead recognizing the power that emerges from relationships between individuals. Resultant conceptual models of PE thus include four key domains: 1. emotional (e.g., motivation control), 2. cognitive (e.g., skills development; 3. behavioral (e.g., organizational participation); and 4. relational (e.g., facilitating others’ empowerment) [9]. Despite this theory, few studies have used SNA to explore relationships between specific social network characteristics and PE.

Given this context, we conducted a case study to illuminate relationships between social network characteristics, PE, and mental health among a community-based sample of adolescents living in areas with concentrated disadvantage. We accomplished this through the following exploratory research questions: 1) how are social network characteristics related to PE? 2) how are social network characteristics related to mental health (i.e., symptoms of depression, anxiety, and post-traumatic stress)?; and 3) does PE impact relationships between network features and emotional health? By focusing on a community-based sample of adolescents, we hoped to better characterize the role of social supports outside of the school environment. We also sought to support hypothesis generation related to social network characteristics and PE.

Methods

Design and participants

We administered cross-sectional egocentric social network surveys to 163 youth aged 13–19. Youth were recruited in-person and by telephone based on participation in one of two community-based violence prevention programs [36, 36, 37, 37]. We specifically recruited from this sample due to a high prevalence of discrimination and community violence in this cohort. An individual from the research team reviewed research procedures with each participant, and youth provided verbal assent to participate. Study data were collected and managed using REDCap, a secure, web-based electronic data capture tool [20]. This study was determined to be minimal risk by the Institutional Review Board, and a waiver of parental consent and a waiver to document written consent was obtained. Youth provided verbal informed assent to participate in the study. Youth received $40 renumeration for survey completion.

Demographics, psychological empowerment, and mental health

Demographic data included: age (reported in years), sex at birth (male/female), gender identity (girl, boy, trans girl, trans boy, genderqueer, non-binary, another identity), race, ethnicity, and parental education. Youth were considered gender diverse if their gender identity did not align with their sex at birth or they indicated any of the following identities: trans girl, trans boy, non-binary, genderqueer, or another identity. Race/ethnicity was operationalized to a three-level categorical variable (Black non-Hispanic, Black Hispanic, multiracial/other) and parental education was categorized into a binary measure (some high school or less versus high school graduate or above) for regression analyses. Psychological empowerment (PE) was measured with 14 Likert-type items (e.g., “I can usually figure out how to get an adult to see my point of view, even if they don’t agree with me,” 1 = strongly disagree to 4 = strongly agree) derived from a validated scale and summarized by averaging responses across items (range: 1–4; Cronbach’s α = 0.97) [42]. This 26-item measure was developed among a sample of urban youth and demonstrates good construct and criterion validity. We did not include items specific to the school setting given the community-based nature of our sample. Mental health measures included the 20-item Center for Epidemiologic Studies Depression Scale (CESD; Cronbach’s α = 0.87) [44], the seven-item Generalized Anxiety Disorder-7 (GAD-7, Cronbach’s α = 0.92) [51], and the 20-item Child Post-Traumatic Stress Disorder (PTSD) Symptom Scale (CPSS,Cronbach’s α = 0.96) [17]. Summary scores were obtained by calculating a sum across items, with reference cutoffs as follows: ≥ 16 on the CESD indicating high likelihood of depression; scores of 5–9, 10–14, and ≥ 15 corresponding to mild, moderate, and severe anxiety on the GAD-7; and ≥ 31 representing a probable PTSD diagnosis on the CPSS.

Social network measures

Social networks were generated by asking youth (i.e., egos) to enumerate up to 20 individuals (i.e., alters) that they turn to for support in their family, school, community, workplace, or online. Youth were instructed to list individuals who they know and who know them, who are important to them, and with whom they have had contact in the last year. Youth answered various questions about each alter, including how they are connected, how long they have known that person, and how often they interact. Youth also indicated whether alters in their network are connected to one another. In addition to social network questions, youth responded to items regarding substance use, violence exposure, and other aspects of physical and mental health. Youth also reported on alters’ substance use behaviors and violence experiences.

Social network data was imported into ORA-LITE (version 3.0.9; Netanomics), a network analysis and visualization software. Visualizations were constructed with nodes representing each individual in the network (i.e., ego and alters) and links showing connections between members of the network. For each youth, we calculated the following network structural characteristics, based on prior literature linking these structural characteristics with adolescent mental health: degree (i.e., number of alters listed), density (i.e., number of network links divided by total number possible), betweenness centrality (i.e., extent to which an ego connects alters to one another), and eigenvector centrality (i.e., a measure of connectivity determined by alter centrality). We selected structural variables that have demonstrated both positive and negative associations with adolescent mental health as part of our exploratory analysis. We also calculated the average overall strength of all ego-alter connections in a young person’s network, based on their perception of that individual as a source of support (e.g., “I have an adult in my life who really cares about me,” followed by, “When you answered the earlier questions about peer support, family support, school support, or community support, how much did they reflect your relationship with [NAME]?”; 4-point Likert, 1 = not at all to 4 = a lot). Additional social network measures included perceived emotional support (i.e., how likely the respondent is to turn to each alter for emotional support, encouragement, or advice), instrumental support (i.e., how likely the respondent is to turn to each alter for housing, transportation, or financial support), and regard (i.e., how much each alter’s opinion matters to the respondent) across all alters in the network (4-point Likert; scores operationalized to mean). We also calculated the mean percent of alters in each young person’s social network to whom youth are somewhat or very likely to turn to for emotional or instrumental support, as well as the mean percent of alters for whom his/her opinion matters somewhat or a lot to the youth.

Statistical analysis

Youth listing at least one alter on social network surveys and answering at least half of the items for PE and each psychological measure were eligible for analysis (n = 111; 68.1%). We selected these criteria due to poor data quality (i.e., > 50% data missing across all variables) among those who were excluded. Among the analytic sample, we performed multiple imputation by chained equations for missing survey data. We determined data to be missing at random based on correlation with other observed variables. The rate of missingness varied across dependent variables (PE: n = 11 [9.9%]; CESD: n = 17 [15.3%]; GAD-7: n = 6 [5.4%]; CPSS: n = 18 [16.2%]) and covariates (age: n = 15 [13.5%]; sex: n = 2 [1.8%]; race/ethnicity: n = 5 [4.5%]; parental education: n = 3 [2.7%]). Logistic regression, predictive mean matching, and proportional odds modeling were used to impute categorial, continuous, and ordinal variables, respectively, over 5 iterations. Following imputation, we performed multivariable linear regression to separately examine associations between each social network measure and PE, adjusting for age, sex, race/ethnicity, parental education, and network size (for network structural characteristics). Similar models were used to explore relationships between network measures and each psychological measure. Given that network features may be correlated with one another [57], we examined each variable separately in regression models to minimize multicollinearity. To explore whether PE moderated relationships between social network measures and psychological measures, we ran additional models with an interaction term (PE x network measure). Covariates were selected a priori based on prior studies by our team and utilizing best practices for statistical modeling using social network data [5]. Model fit was assessed by visualizing residual plots. Although mental health measures (i.e., CESD, GAD-7, and CPSS) demonstrated some right skewing, regression using log-transformed outcomes were consistent. Due to the exploratory nature of our analysis, we focus primarily on descriptive statistics but present results from regression as pooled parameter estimates (regression coefficient [95% CI]) according to Rubin’s rules for imputed data [47]. We selected our target sample size based on prior work in similar populations which has enabled preliminary characterization of network associations [13, 25, 35]. As a sensitivity analysis, we ran similar regression models for youth in the analytic sample without any missing data across variables of interest (complete cases analysis). We also compared characteristics of individuals in the analytic sample to those excluded due to missing data. All analyses were conducted in R version 4.2.1 (2022-06-23).

Results

Participant characteristics

Among 111 youth analyzed, mean age was 14.6 (SD: 1.5) years and about half were assigned female sex at birth (n = 63; 57%) (Table 1). A total of 8 youth (7%) identified as gender diverse. Most youth identified as Black (n = 79; 71%) or multiracial/other (n = 23; 21%), and the majority had parents/guardians that did not graduate high school (n = 77; 69%). Mean PE score was 2.2 (SD: 0.9) at baseline. On average, participants reported high depressive symptoms (CESD: mean = 18.7 [SD: 11.1]), whereas generalized anxiety and post-traumatic stress symptoms were mild-moderate (GAD-7: mean = 4.2 [SD: 5.5]; CPSS: mean = 14.2 [SD: 17.3]).

Table 1.

Demographic characteristics of participants (n = 111)

Characteristic	Mean (SD)/No. (%)a
Age (years)	14.6 (1.5)
Sex at birth (female)	63 (57%)
Gender diverse	8 (7%)
Race
American Indian or Alaska Native	4 (4%)
Black or African American	79 (71%)
Multiracial or Other	23 (21%)
Native Hawaiian or Other Pacific Islander	1 (1%)
White or Caucasian	1 (1%)
Ethnicity: Hispanic or Latino	13 (12%)
Parental education
Some high school or less	77 (69%)
High school graduate or above	31 (28%)
Psychological measures
Depressive symptoms (CESD)	18.7 (11.1)
Generalized anxiety disorder symptoms (GAD-7)	4.2 (5.5)
Post-traumatic stress disorder symptoms (CPSS)	14.2 (17.3)
Psychological empowerment	2.2 (0.9)

^aPercentages may not total 100% due to missing values

Social network characteristics

Example network visualizations for five participants are shown in Fig. 1, which were selected to illustrate differences in network composition, degree, and density. The mean network degree among the sample was 4.8 (SD: 3.3). The median was 4 (range: 2–20; skewness: + 2.4). Other social network structural characteristics are summarized in Table 2. Among 477 total alters listed, most were parents/guardians (n = 159; 33.3%), siblings (n = 94; 19.7%), or extended family members (n = 82; 17.2%) (Table 2). Alters also included friends (n = 72; 15.1%) and other individuals in the community, such as neighbors (n = 3; 0.6%) and friends’ parents (n = 2; 0.4%). A small proportion of alters (n = 11; 2.3%) were school personnel. Regarding alter gender, 251 (52.7%) were female, 174 (36.6%) were male, and 11 (2.3%) were another gender. About half of the alters in each youths’ network were individuals to whom youth were at least somewhat likely to turn for emotional or instrumental support. A similar proportion of alters in each network were individuals whose opinion mattered to the youth respondent. About a third of all alters (n = 162; 34.0%) were individuals who fell in all of these support categories.

Fig. 1.

Example social network visualizations for youth participants. Network visualizations were selected to illustrate differences in network composition, degree, and density. Alters from each setting (e.g., immediate family) are color-coded. The degree (i.e., number of connections) and density (i.e., number of links divided by total possible links) vary across participant networks. For example, Participant 69 exhibits a larger network compared to that of Participant 101, but Participant 101’s network is more densely interconnected due to exhibiting all possible nodal links

Table 2.

Social network characteristics of sample

Characteristic	Mean (SD)/No. (%)a
Social network structural measuresb
Number of alters (degree)	4.8 (3.3)
Betweenness centrality	0.02 (0.04)
Density	0.32 (0.23)
Eigenvector centrality	0.44 (0.11)
Alters immediate familyc
Mother	95 (20%)
Stepmother	7 (1.5%)
Father	45 (9.5%)
Stepfather	5 (1.1%)
Nonbiological guardian	7 (1.5%)
Brother/stepbrother	46 (9.7%)
Sister/stepsister	48 (10.1%)
Alters extended family
Maternal grandmother	14 (2.9%)
Paternal grandmother	5 (1.1%)
Maternal grandfather	9 (1.9%)
Paternal grandfather	1 (0.2%)
Aunt/uncle	27 (5.7%)
Cousin	26 (5.5%)
Alters from school
Teacher	9 (1.9%)
School guidance counselor	2 (0.4%)
Alters from community
Boyfriend/girlfriend/partner	4 (0.8%)
Therapist/social worker	1 (0.2%)
Employer	2 (0.4%)
Neighbor	3 (0.6%)
Friend	72 (15.1%)
Peer in neighborhood	1 (0.2%)
Friend's parent	2 (0.4%)
Other	10 (2.1%)
Alter gender
Male	174 (36.6%)
Female	251 (52.7%)
Non-binary or gender fluid	2 (0.4%)
Transgender male	1 (0.2%)
Transgender female	2 (0.4%)
Other	6 (1.3%)
Other alter characteristicsd
Percent somewhat likely to turn to for emotional support	12.0 (19.2)
Percent very likely to turn to for emotional support	28.7 (30.6)
Percent somewhat likely to turn to for instrumental support	15.5 (22.9)
Percent very likely to turn to for instrumental support	28.6 (31.9)
Percent for whom opinion matters somewhat	11.7 (21.8)
Percent for whom opinion matters a lot	34.4 (32.7)

^aPercentages may not total 100% due to missing responses

^bPresented as mean (standard deviation) across 111 youth respondents

^cPresented as frequency (percentage) of 477 total alters listed by all respondents

^dPresented as mean percent (standard deviation) of alters in each youth’s social network

Relationships between social network characteristics, psychological empowerment, and mental health

When examining relationships between social network characteristics, psychological empowerment, and mental health, higher betweenness centrality was significantly associated with greater anxiety (β: 47.1 [95% CI: 16.9, 77.2]) and post-traumatic stress (β: 162.2 [95% CI: 46.6, 277.8]) symptoms (Table 3). There were no significant relationships between other network structural measures and mental health constructs. No structural characteristics were associated with psychological empowerment. Alter features also showed no significant associations with mental health measures or psychological empowerment. The association of betweenness centrality and mental health was not significantly moderated by psychological empowerment. Results from the complete cases analysis were consistent with these findings. Compared to the 52 youth with missing social network and/or mental health data, youth in the analytic sample were more likely to be Black non-Hispanic, but otherwise did not differ across other demographic characteristics. Across model covariates, male sex and Black non-Hispanic identity were associated with lower scores on all mental health measures.

Table 3.

Associations between social network characteristics, mental health measures, and psychological empowerment

	Parameter estimates (β) (95% confidence intervals)
	Depression (CESD)		Anxiety (GAD-7)		Post-traumatic stress (CPSS)		Psychological Empowerment
	Unadjusted	Adjusteda	Unadjusted	Adjusteda	Unadjusted	Adjusteda	Unadjusted	Adjusteda
Alter featuresb
Overall support	−0.14 (−2.05, 1.77)	−0.10 (−2.00, 1.81)	−0.46 (−1.38, 0.47)	−0.60 (−1.48, 0.28)	1.56 (−1.78, 4.90)	1.75 (−1.75, 5.25)	0.04 (−0.12, 0.19)	0.02 (−0.15, 0.19)
Emotional support	0.19 (−1.99, 2.36)	0.91 (−1.29, 3.11)	−0.13 (−1.21, 0.95)	0.30 (−0.75, 1.35)	0.67 (−3.21, 4.54)	2.35 (−1.73, 6.42)	−0.02 (−0.20, 0.16)	−0.04 (−0.24, 0.16)
Instrumental support	−1.60 (−3.86, 0.66)	−0.50 (−2.96, 1.96)	−0.18 (−1.30, 0.94)	0.03 (−1.12, 1.19)	−1.05 (−5.14, 3.04)	0.70 (−3.82, 5.22)	0.03 (−0.16, 0.21)	−0.01 (−0.23, 0.21)
Regard	0.17 (−2.15, 2.48)	0.16 (−2.23, 2.54)	−0.60 (−1.73, 0.53)	−0.67 (−1.76, 0.42)	1.13 (−2.93, 5.19)	1.81 (−2.50, 6.12)	−0.01 (−0.20, 0.18)	−0.01 (−0.23, 0.20)
Structural measures
Network size	−0.02 (−0.66, 0.62)	−0.06 (−0.69, 0.57)	−0.12 (−0.44, 0.20)	−0.20 (−0.50, 0.10)	−0.11 (−1.25, 1.03)	−0.37 (−1.54, 0.79)	0.00 (−0.05, 0.05)	0.01 (−0.05, 0.06)
Density	−2.33 (−21.56, 16.89)	4.90 (−18.11, 27.92)	2.54 (−7.18, 12.26)	6.08 (−5.26, 17.43)	16.04 (−18.89, 50.97)	30.64 (−13.25, 74.53)	0.60 (−0.87, 2.07)	0.64 (−1.30, 2.58)
Betweenness centrality	55.62 (−11.91, 123.16)	60.73 (−4.07, 125.53)	43.76 (10.22, 77.32)*	47.08 (16.9, 77.22)**	146.86 (27.94, 265.78)*	162.22 (46.63, 277.81)**	−0.88 (−5.85, 4.08)	−1.59 (−6.79, 3.61)
Eigenvector centrality	1.06 (−10.67, 12.78)	−1.13 (−13.49, 11.22)	0.11 (−5.86, 6.08)	−1.56 (−7.67, 4.54)	−8.23 (−29.48, 13.02)	−7.84 (−30.99, 15.30)	−0.34 (−1.16, 0.48)	−0.31 (−1.26, 0.65)

^aMultivariable models adjusted for age, sex, race/ethnicity, parental education, and network size (for network structural characteristics). Models separately examined relationships between each network characteristic and each psychological measure

^bFor regression models, measures were calculated as the mean across alters in each young person’s social network (4-point Likert)

^* p < 0.05, ** p < 0.01

Discussion

In our case study examining social network features among predominantly Black youth living in neighborhoods with concentrated disadvantage, we found that betweenness centrality (i.e., the degree to which youth connect alters to one another) was positively correlated with both anxiety and post-traumatic stress symptoms. This finding contrasts somewhat from prior literature, which has shown increased network size, connectedness, and social support predict lower mental health symptomatology [21, 28, 40, 55]. Other network characteristics, including size and density, showed no relationships with mental health symptoms. In addition, characteristics of the alters in a young person’s network, including the magnitude of perceived support, did not demonstrate any association with psychological measures. [21, 28, 40, 55]

The majority of alters youth identified were immediate or extended family members, which aligns with findings in similar samples [12]. Many were viewed as sources of emotional and instrumental support, but these characteristics did not correlate with youths’ mental health. This may be partially related to our measurement of clinical mental health outcomes (i.e., anxiety, depression, and PTSD) requiring higher levels of support, rather than other aspects of well-being (e.g., self-esteem). In addition, providing only ego-syntonic support, or support which does not challenge disordered thought patterns, may not produce notable changes in mental health. Information regarding alters’ psychological functioning was not collected in this study, which may help further contextualize findings among youth. Prior work has shown that mental health may be influenced by that of peers and other social connections [45, 52]. In addition, one’s willingness to seek services or treatment may depend upon attitudes and experiences related to mental health care across an individual’s social network [2, 7, 62]. Thus, efforts to describe psychological well-being among alters, as well as the type of emotional support provided, may offer greater understanding of the ways in which social networks influence mental health and well-being.

Our observation regarding the association of betweenness centrality with higher scores on the GAD-7 and CPSS is somewhat surprising. As a measure of network position, betweenness centrality offers insight into information exchange between network members. Youth with higher betweenness centrality may be expected to experience better mental health due to their role as brokers of key resources or supports. However, such youth may also be at greater risk of exposure to psychosocial challenges (e.g., violence, substance use) among other network members, which could increase their susceptibility to anxiety or PTSD. This aligns with prior work which has shown links between centrality and other risk behaviors, including substance use [16, 23]. Notably, our study sample consisted primarily of minoritized youth, and risk and protective factors related to mental health for this population may differ than those among White youth in resourced areas. This is an important distinction, particularly given increasing rates of suicidality and other mental health concerns among minoritized youth [41].

Because our data is cross-sectional, associations between network connectivity and mental health symptoms may also indicate youths’ efforts to cultivate interpersonal connections in response to mental health challenges [39]. Prior work has illustrated the reciprocal interactions between social networks and mental health [30]. For example, middle school students with attention deficit hyperactivity disorder (ADHD) demonstrated smaller social networks (i.e., lower degree) and decreased closeness (i.e., breadth of peer connections), which was theorized to result from the negative impact of ADHD symptoms (e.g., impulsivity, hyperactivity) on peer perceptions [28]. Youth with internalizing symptoms, such as depression and anxiety, may be more likely to select peers exhibiting similar psychopathology, though this evidence is still evolving [39]. This type of homophily (i.e., selecting peers based on similarities in behavior, attitudes, or demographics) has shown important correlations with other adolescent risk behaviors, as well [38]. Our findings could reflect similar social processes, though longitudinal network studies are needed to better delineate these complex relationships.

We theorized that psychological empowerment would also show associations with social network variables given its conceptualization as a relational construct [9, 31, 71]. Our finding that social network measures were not significantly correlated with PE indicate that this relationship may be more nuanced. For example, PE may depend upon the overall level of empowerment in a young person’s social network. Other factors, such as the degree of community involvement or embeddedness among peers, mentors, and family members, may also play a role [11, 32, 33]. As noted, PE did not moderate relationships between social networks and psychological measures. While PE has been linked with other internalized protective factors, such as future orientation and resilience [54], it may exert less direct influence on mental health symptoms, like psychological distress, highlighting the need to intentionally address youths’ emotional well-being in empowerment interventions. It is also possible that other intrinsic skills (e.g., problem-solving, adaptive coping, goal-setting) may be more salient in fortifying PE. PE may also show a more reciprocal interaction with social networks that cannot be elucidated in the current study. Surveys of individuals engaged in participatory action research, an example of an empowering process, demonstrate that social networks may shift over time due to involvement in these activities [6]. Rather than measuring social network characteristics at a single time point as in our study, collecting longitudinal social network data along with mental health symptoms may offer greater insight into the directionality of these relationships.

Findings should be contextualized within several key limitations. First, the cross-sectional nature of our data prevents assumptions of causation, and the small size of our sample limits statistical power. Because this was an exploratory study, statistically significant findings should be interpreted with caution, and future work in larger samples is needed to conduct hypothesis-driven inferential analyses. Our study primarily surveyed youth from minoritized racial backgrounds who had participated in one of several community-based violence prevention programs; while inclusion of these young people is critical in advancing research equity, our sampling strategy may have reduced variability across social network variables or psychological outcomes, potentially masking larger sociostructural influences at play. Due to the convenience nature of our sample, results may also not generalize to youth across other identities. In addition, the length of surveys, particularly items involving egocentric social networks, may have caused response fatigue, impacting data quality. We also acknowledge that egocentric surveys only capture one perspective within a given network, and are prone to challenges from missing or incomplete data (e.g., an ego misclassifying inter-alter relationships); social cognitive mapping or whole network approaches can overcome this barrier by incorporating additional informants and characterizing social groups outside of an individual’s own network [4, 15, 29].

Conclusions

Our study provides preliminary insight into the ways in which social networks may exert more nuanced influences on mental health and well-being among young people facing other psychosocial challenges (e.g., violence exposure, discrimination, etc.). While youth identified many network members to whom they would turn for support, youth experiencing other forms of marginalization may require socioecological interventions that address systemic causes of inequity. In addition, further work is warranted to understand how dynamic changes in social networks may interact with mental health and well-being trajectories over time. Conducting large-scale social network studies with youth across multiple settings (e.g., schools, neighborhoods, online) may also provide a better understanding of these relationships. Programs that leverage SNA data to spark or strengthen social connections can be powerful [24, 56]. Indeed, by embedding social network interventions within larger frameworks of positive youth development, racial justice, and community vitality, youth may be better able to cultivate peer and adult support systems that augment parallel strategies to support their mental health and well-being.

Abbreviations

SNA

Social network analysis

PE

Psychological empowerment

Author contributions

All authors conceptualized the study. Dr. Szoko conducted data analysis, interpreted findings, drafted the manuscript, revised the manuscript, and prepared the manuscript for submission. Alexander Riley, Dr. Rankine, and Dr. Culyba provided administrative support, interpreted findings, edited the manuscript, and critically reviewed the manuscript for important intellectual content. Dr. Culyba also provided supervision of data analysis. All authors agree to be accountable for the presented work and approve of the final version of the manuscript.

Funding

Funding for the authors’ time and effort was supported in part by the following grants: NIH NCATS TL1TR001858-05 (Scholar: Szoko, PI: Kraemer). The funders had no role in the design and conduct of the study, collection, management, analysis, and interpretation of the data, and preparation, review, or approval of the manuscript, and decision to submit the manuscript for publication.

Data availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All study procedures were approved by the University of Pittsburgh Human Research Protections Office in accordance with the Belmont Report. This study was determined to be minimal risk by the University of Pittsburgh Institutional Review Board, and a waiver of parental consent and a waiver to document written consent was obtained. Youth provided verbal informed assent to participate in the study.

Consent to publish

Not applicable

Clinical trial number

Not applicable

Competing interest

The authors declare no competing interests.