Tailoring Supports to Youth in Schools: One Approach to Identifying Needs and Targeting Intervention

Youth experiences intersect along their race, gender, language and socioeconomic status, schools must consider the intersectionality in order to improve outcomes. The current study sought to understand if, and to what extent, different clusters of youth in one large urban high school perceive their psychosocial behaviors as well as social and interpersonal skills. Cross-sectional survey data from 1,164 high school youth were collected using four valid scales: Internalizing Behaviors, Externalizing Behaviors, Peer Relationships, and Social Skills. The analytic strategy was twofold. Cluster analysis was used to form homogeneous clusters of the 1,147 complete responses based on a combination of race/ethnicity, socioeconomic status, English language learner status, and gender. Next, a multivariate analysis of variance (MANCOVA) was used to compare cluster profiles on the aforementioned measures. The cluster analysis revealed a 5-factor solution. Findings of the MANCOVA showed that the clusters differed significantly on the set of psychosocial behaviors, social skills, and peer relationships variables (p<0.05, η2= 0.08), with significant univariate differences (p < 0.05) emerging on all four variables. Results point to cluster analysis as an emerging way to identify needs and tailor supports to youth in schools. Youth with identified needs could receive more specialized Tier II and III interventions aligned with their intersecting cultural and personal experiences, in addition to universal interventions to meet school wide priorities.

The needs of youth today are increasingly complex. Estimates suggest 10 to 20 percent of youth within the United States experience nonacademic barriers to learning such as mental health problems or severe emotional and behavioral challenges (World Health Organization, 2011). As a result, schools are called upon to individualize instruction, diversify learning supports, provide health and social services, and promote a positive, inclusive school culture and climate to ensure all youth succeed (Anderson-Butcher et al., 2008). The challenge for schools is youth have different backgrounds, experiences, skills, and competencies that influence their behaviors and outcomes in school contexts. There is a need to understand whether their social identities influence their needs in order to tailor individualized intervention services. The current study, informed by an intersectionality lens, examined whether youth attending a large urban high school presented with differential psychosocial behaviors and social and interpersonal skills based on their race, gender, language spoken at home, and socioeconomic status. The goal of this approach was to include the youth voice in their assessment of their skills, to mitigate the influence of subjective evaluation and biases in assessing youth behaviors, and to inform the development of tailored supports and targeted schoolwide interventions for at-risk subgroups in one large urban high school.

Theoretical Framework
The theory of intersectionality frames how multiple forms of inequality and identity are interrelated and perpetuate different outcomes in social contexts over time (Crenshaw, 1993). In schools, a breadth of research indicates race, gender, language spoken at home, and socioeconomic status perpetuate varying levels of advantage and disadvantage in educational contexts. These advantages and disadvantages are tied to systemic factors such as sexism, racism, classism, and a lack of English-speaking privilege reinforced across systems, yet influence individual youth. Studies show African American youth may perceive a more negative racial climate than White youth irrespective of socioeconomic status, but that African American female youth tend to perceive less racial fairness in schools than African American males (Watkins & Aber, 2009). Aside from race and gender, youth also can experience stressors associated with migration, immigration, and language differences place them at an increased risk for mental health symptoms (Potochnick & Perreria, 2010). Studies signify how demographic factors influence school experiences and overlapping systemic forces may influence the psychosocial needs of various subgroups of youth.
An intersectionality lens is interrelated to the profession of social work and education given researchers, educators, and practitioners are trained to approach social issues by recognizing the significance of the person-in-environment; considering risks in the context of individual, family, community, and system level factors; and, advocating that complex institutional and systemic forces perpetuate differential outcomes for social vulnerable and marginalized populations (National Association of Social Workers, 2009). School-based interventions that are culturally competent, tailored, and contextualized can minimize risks and increase protective factors for youth in their schools (Arora et al., 2019). Yet one challenge to examining school-wide data and tailoring interventions is approaching data using an intersectionality lens to accounts for youths' diverse cultural and racial backgrounds, capture broader systemic patterns of oppression, and assess their psychological and interpersonal needs.

Nonacademic Needs
By the time youth reach adulthood, approximately 20 percent will exhibit disruptive behavior or have an anxiety, mood, or substance-use disorder (Keyes, 2006). In addition, depression is one of the most common psychological disorders in adolescents which schools may address risk factors for via school-based mental health and social emotional learning (SEL) interventions (Domitrovich, Durlak, Staley, & Weissberg, 2017;Townsend et al., 2017). Scholars advocate a lack of psychosocial, social, and interpersonal skills are predictors of these poor health and well-being outcomes for youth. For instance, psychosocial behaviors such as internalizing coping behaviors, in conjuncture with bullying, may lead to long-term risks for engaging in internalizing behaviors and substance misuse (Klomek et al., 2011). These examples demonstrate how the differential nonacademic needs of youth in schools can influence their school experience.
Psychosocial behaviors. Examination of nonacademic needs ensure conditions in schools are appropriate for learning and positive outcomes. The psychosocial behaviors of high school youth include both internalizing and externalizing symptomologies that result in expressed behaviors indicative of emotional and behavioral disorders (Kramer et al., 2014). Although definitions of internalizing behaviors differ, such behaviors often include withdrawal, anxiety, and depression (Liu, Chen, & Lewis, 2011). Conversely, externalizing behaviors are outward behaviors that negatively affect others and the environment where high school youth interact. Externalizing behaviors also can include substance use, violence, theft, and physical aggression (Edwards, Mumford, & Serra-Roldan, 2007). Internalizing and externalizing behaviors increase the risk for low achievement and school dropout, and therefore are critical to examine among high school youth within context of school (Liu et al., 2011).
Social and interpersonal skills. Another critical need is the examination of the social and interpersonal skills. Social skills range beyond verbal responses and include non-verbal responses such as facial expressions and body posture (Spence, 2003). Interpersonal skills are related to helping others, persuasion, negotiation, perceptiveness, and problem solving, and are prosocial or positively reinforced behaviors (Morgeson, Reider, & Champion, 2005). Further, studies show the development of strong interpersonal skills can result in greater school engagement (DiPerna et al., 2018), and promote positive behavioral and academic outcomes (Furrer, Skinner, & Pitzer, 2014;Tan, Oe, & Hoang, 2018). Hence, social, and interpersonal skills serve as contributors to how high school youth experience and perform in school.
Current intervention approaches. In the United States, schools aim to address the needs of youth using a multi-tiered systems of support (MTSS) approach. MTSS involves three tiers of intervention: Tier 1, universal intervention; Tier II, intensive interventions for a select group; Tier III, intensive interventions for the highest need group (Arora et al., 2019). However, biases may exist and surface when school personnel look differently than the youth served. For this reason, youth voice, unique cultural experiences and interactions with school climate are crucial to understanding their needs. In order to identify high school youth at-risk and in need of Tier II and III intervention, schools are encouraged to conduct comprehensive screenings or needs assessments which include youth perspectives (see Anderson-Butcher et al., 2008;Hopson & Lawson, 2011;LeWallen et al., 2015). The current study uses an intersectionality lens to examine whether youth attending a large urban high school presented with differential psychosocial, social, and interpersonal skills based on their race, gender, language, and socioeconomic status.

Current Study
To support schools in meeting the diverse needs of high school youth, schools need strategies focused on intersectionality and identification of risks to tailor interventions to specific populations of high school youth. When identifying at-risk youth in schools and tailoring supports to them via MTSS model, studies show implicit biases and a high level of subjectivity exists in schools due to differential demographics of students and those teaching them or in administrative positions (Naser, Brown, & Verlenden, 2018). School-based professionals need unbiased approaches and methods inclusive of youth voice concerning their needs and assets to tailor interventions and supports to students. Using an exploratory approach and cluster analysis methodology, the current study sought to understand if, and to what extent, different populations of high school youth may experience individualized mental health needs related to their psychosocial behaviors, as well as their social and interpersonal skills.

Method
This study was approved by the lead author's university Institutional Review Board at the time of the study. The study used cross-sectional survey data from 1,164 high school youth attending a large urban high school in the Intermountain West region of the United States. However, 17 youth did not have complete data on the items of interest in this study making our total sample 1,147 youth. The total school enrollment was 2,281 at the time of the study, indicating a 50 percent completion rate. Data were collected as secondary data and part of a larger school-wide needs assessment using the Community and Youth Collaborative Institute School Experience Surveys (CAYCI-SES; Anderson-Butcher, et al., in press). All survey measures were collected using paper/pencil surveys or online surveys, based on the school administrator's choice. Participants ages ranged from 14 to 19 years of age (M = 15.86; SD = 1.38). In total, 32 percent of high school youth were in 9 th grade, 29 percent in 10 th grade, 20 percent in 11 th grade, and 18 percent in 12 th grade. Additional demographics are presented in Table  1.

Measures
Specific constructs assessing psychosocial behaviors and social and interpersonal skills were measured using the CAYCI-SES (Anderson-Butcher, Amorose, Iachini, & Ball, 2013), a needs assessment tool used to inform school planning and improvement efforts for educators, school officials, and community stakeholders. The CAYCI-SES is used in school districts across the country, and is listed on the National Center for Supportive Learning's Compendium of valid and reliable school climate measures: https://safesupportivelearning.ed.gov/survey/community-and-youth-collaborativeinstitute-cayci-school-experiences-surveys. Detailed psychometric information related to each measure can be found at the following website (see http://cayci.osu.edu/surveys/surveys-and-technical-reports/). Note subscales on the CAYCI-SES are scored by calculating one overall mean score for all items on the measure.
Psychosocial behaviors. Both internalizing and externalizing behaviors were examined using CAYCI-SES subscales. The Internalizing Behavior Scale (Anderson-Butcher et al., 2013) assesses the extent to which high school youth report behaviors and feelings dealt with internally, rather than by acting them out in the home or school. For example, an item on the scale is "In the past week, I felt sad." The Internalizing Behavior Scale is measured from 1 = Strongly disagree to 5 = Strongly agree, with higher mean scores indicating greater internalizing behaviors. The CAYCI-SES Externalizing Behaviors Scale (Anderson-Butcher, et al., 2013) measures the extent to which high school youth report behaviors directed outward, either toward the external environment, such as poor attention or impulsivity. An example of an item is "Have you ever gotten in trouble in class." The Externalizing Scale is measured using a 5-point Likert-type response format ranging from 1 = Never to 5 =Very Often, with higher mean scores indicating greater externalizing behaviors. The Internalizing Scale and Externalizing Scale demonstrated adequate reliability in this study α = 0.90; 0.82 respectively) and in past research (α = 0.90; 0.79; Anderson-Butcher et al., 2013).
Social and interpersonal skills. To measure social and interpersonal skills, high school youth completed instruments designed specifically to assess each construct. The CAYCI-SES Social Skills Scale (Anderson-Butcher et al., 2013) assesses high school youth perceptions of their ability to engage in positive social interactions with others. "I work well with others" is an example of an item on the scale. The scale (α = 0.79) is measured on a 5-point response format ranging from 1 = Strongly disagree to 5 = Strongly agree, with higher mean scores indicating greater social skills. The CAYCI-SES Peer Relationships Scale (Anderson-Butcher et al., 2013) assesses the extent to which middle/high school youth feel they are supported by and have positive relationships with their peers. An example item on the scale is "My friends support and care about me." All items on both scales are measured using a 5-point response format ranging from 1 = Strongly disagree to 5 = Strongly agree, with higher mean scores indicating better peer relationships. The Social Skills Scale (α = 0.79) and Peer Relationships Scale (α = 0.85) demonstrated adequate reliability in this sample, as well as in past research (α = 0.79; 0.85).

Analytic Strategy
The decision was made to use cluster analysis as this technique can perform a number of useful functions such as taking a heterogeneous sample of individuals and forming relatively homogenous groups and then organizing large quantities of multivariate information. Cluster analysis allows for labels to be assigned to subgroups, making data more manageable for researchers to identify groups that might best benefit from interventions (Clatworthy, Buick, Hankins, Weinman, & Horne, 2005;DiStefano & Kamphaus, 2006). In social work and other health professions, cluster analysis is commonly utilized to examine how individual characteristics, interpersonal and social traits, and lifestyle behaviors generate unique patterns of risk and protection (Anthony, 2008;Dumuid et al., 2018). Further, although cluster analysis introduces more subjectivity compared to other techniques such as latent class analysis, the procedures can be done using the Statistical Package for the Social Sciences (SPSS Version 23 used here). SPSS is a statistical package commonly utilized in social work, and allowed us to approach these data and our research question using an intersectionality lens.
The analytic strategy was twofold. First, an exploratory two-step cluster analysis was used to determine whether profiles of high school youth could be identified based on the following demographic indicators: race/ethnicity, socioeconomic status (measured by proxy indicator of free and reduced lunch; FRL), English language as second language status (ESL), and gender (male and female). Race/ethnicity was categorical and included the following: White, Hispanic/Latino, Asian, Multi-racial, Black/African American, Native Hawaiian, American Indian/Alaska Native, and an option to choose not to identify their race/ethnicity. Socioeconomic status (FRL), English language as a second language status, and gender were dichotomous yes/no indicators.
As a first step in SPSS, demographic indicators were input into the two-step cluster as categorical variables. The number of pre-determined clusters were set to the default in SPSS to allow up to fifteen homogenous groups to emerge. In addition, the default log-likelihood distance measure was used, along with the creation of a cluster membership variable to capture and code emergent homogenous groups. Following the input of demographic indicators, model quality was assessed. Good model quality indicates the four inputs (FRL, language, gender, and race/ethnicity) allowed for the discovery of some or all patterns based on the Schwarz Bayesian Criterion (Norušis, 2012). In the end, a solution with good model quality resulted in five clusters. As a second step, the cluster membership variable was used to examine descriptive statistics including means and standard deviations on the outcomes of interest (i.e., externalizing behaviors and internalizing behaviors, social skills, and interpersonal skills). Differences in means for each of the five emergent clusters were examined and compared using a multivariate analysis of covariance (MANCOVA) and Newman-Keuls post-hoc test. Since the measures were scored on five-point Likert scales, variables were not converted to z-scores prior to the MANCOVA analysis. Results of the MANCOVA were considered significant at the p < 0.05 level.

Results
To reiterate, the two-step cluster analysis revealed a five-factor solution. The demographic characteristics of each of the five clusters are described next. The name of the cluster and potentially overlapping concepts informed by intersectionality theory that appear to be influencing the majority of youth in the cluster also are reported. First, Cluster 1 named 'White Female Experience" (n =295) consisted of White, females, reporting no receipt of FRL, and who spoke English as their first language. Cluster 2 named "White Male Experience" (n = 251) consisted of White, males, reporting no receipt of FRL, and who spoke English as their first language. Cluster 3 named "Low-Income Experience" (n = 249) contained predominantly White (58.2%) high school youth among whom a majority identified as male (51.4%). Cluster 3 consisted of high school youth who all reported receipt of FRL who spoke English as their first language. Cluster 4 named "Mixed Race and Mixed Gender Experience" (n = 226) consisted of high school youth who did not report receipt of FRL. A majority of high school youth in Cluster 4 identified as female (54%) and spoke English as their first language (64.2%). High school youth in Cluster 4 also identified their race/ethnicity as Asian (32.3%). Finally, Cluster 5 named "Low-Income Mixed Race and Mixed Gender Experience" (n = 126) comprised high school youth who were predominantly female (59.5%), Hispanic/Latino (75.4%), and all of whom reported receipt of FRL and speaking Spanish as their first language. The ESL status of Cluster 5 differed from the other four clusters.

Descriptive Results
Descriptive statistics, including the mean for the overall school and by cluster profile on the key outcomes, are presented in Table 1. Notably, important information relevant to the school was captured via descriptive results. Figure 1 provides a visual of mean differences by cluster profile compared to the overall school mean on each outcome. MANCOVA results showed high school youth in Clusters 1 (White Female Experience) and 4 (Mixed Race and Mixed Gender Experience) showed significantly higher internalizing behaviors compared to not only other clusters, but also compared to the overall school mean. Furthermore, Cluster 5 (Low-Income Mixed Race and Mixed Gender Experience) showed significantly higher externalizing and lower peer relationships compared to all other clusters and to the overall school mean. In terms of peer and social skills, Cluster 1 (White Female Experience) showed significantly higher peer relationships and social skills compared to other clusters and the overall school mean.

Discussion
The aim of the current study was to use cluster analysis as an approach valuing intersectionality to examine to nonacademic needs among clusters of high school youth in one large urban high school. Findings include five clusters of high school youth that emerged with significantly diverse school experiences across measures of psychosocial behaviors and social and interpersonal skills. These findings contribute to our understanding of how this methodological approach can identify needs using an intersectionality lens and may serve as one way to target interventions to meet nonacademic needs. Results demonstrate different clusters of high school youth exist in schools and variability exists in key outcomes related to psychosocial behaviors as well as social and interpersonal skills among high school youth clusters. Overlapping systems of inequality linked to a young person's identity such as race, gender, socioeconomic status, and language spoken at home appear to influence their needs in schools, but also point toward their cultural strengths and assets. By identifying specific high school youth clusters using cluster analysis, schools can target high school youth by cluster which identify and align with specific needs, rather than expending resources or supports to high school youth who may not present with similar needs.
Apparent from the results are three identifiable differences in the reported behaviors and skills of high school youth in Low-Income Experience, Mixed Race and Mixed Gender Experience, and Low-Income Mixed Race and Mixed Gender Experience. Specifically, Low-Income Experience and Low-Income Mixed Race and Gender Experience were those who reported the highest internalizing behaviors, reported receipt of FRL, and identified as predominantly female. Notably, Low-Income Mixed Race and Mixed Gender Experience comprised those who were both receiving FRL and who did not speak English as their first language, which may in turn shed light on other barriers to learning these high school youth and their families face. In other words, symptomology of African American or Latinx females may be a sign of other challenges related to meeting basic needs, peer pressure, unaddressed trauma, or discrimination (Potochnick & Perreira, 2010;Keyes, 2006).
To address such needs, MTSS can be maximized to incorporate interventions and promote skill development in targeted groups. Low-Income Experience showed high internalizing behaviors significantly different than the Low-Income Mixed Gender Experience and Low-Income Mixed Race and Mixed Gender Experience. High school youth with greater internalizing behaviors may warrant further assessment of their symptomology for individual (Tier III) or small group (Tier II) interventions. A Tier I universal approach using an evidence-based program such as Strong Kids may be beneficial as the program is shown to increase prosocial behaviors while decreasing internalizing behaviors through the implementation of a social-emotional learning curriculum in classroom contexts (Kramer et al., 2014). Broader approaches to addressing the impact of poverty may further support individual or group interventions.
Moreover, Low-Income Mixed Race and Mixed Gender Experience reported strong social and interpersonal skills, yet those who were most at-risk for exhibiting externalizing behaviors compared to all other clusters. For these youth, individual interventions targeted at developing coping skills (Tier III) and skill development groups (Tier II) may be intentionally tailored to target the needs of these high school youth. For example, social emotional curriculums facilitated in group (Tier II) or individual sessions (Tier III) serve to increase protective factors, decrease problem behaviors, and provide high school youth skills and tools to support their overall healthy development (Domitrovich et al., 2017). School social workers may advocate that these culturally diverse high school youth have social and interpersonal skills to lead peer groups or serve as role-models in the school.

Implications and Limitations
Such important disaggregation of data through cluster analysis through an intersectionality lens may support data examination and drive development of tailored interventions that consider cultural strengths and systemic risks to support high school youth. To date, schools often utilize Office Discipline Referrals (ODRs) or academic indicators to identify high school youth who are at risk; however, researchers and practitioners must utilize different methods (Lane et al., 2015). Collecting self-report data on social skills and other individual-level risk and protective factors can support the development of school-wide and individual interventions to meet mental health, academic, and social needs (Anderson-Butcher et al., 2008). Notably, as identified in previous studies, differences in outcomes were associated with demographic factors such as socioeconomic status, gender, race, and language differences (Edwards, Mumford, & Serra-Roldan, 2007).
In future research, similar research methods may be used to identify needs and improve high school youth outcomes in school contexts. For example, identifying specific clusters of high school youth may point to the need to allocate specific services and supports to the district or school. Additionally, school social workers can more intentionally engage in case management approaches, working individually or in small groups to address specific behaviors. Although vulnerable schools may have limited resources to support individualized interventions, the implementation of SEL interventions outweigh the financial costs (Domitrovich et al., 2017). Further, schools can leverage resources to supporting culturally diverse students in building skills to develop social and emotional competencies have been used to counteract risks of disengagement in school (Durlak, Weissberg, Dymnicki, Taylor, & Schellinger, 2011).
Current limitations to the study include conducting the analysis in a single school context, lack of youth outcome data, using cross-sectional data, and only including one stakeholder perspective. Though the study utilized a large sample, perspectives from multiple schools with additional demographic indicators, schools, and multiple time-points are needed to support the study's findings and the applicability of identifying intersectional patterns of needs among subgroups of youth. It is also important to note demographics used in the cluster analysis are limited. For instance, gender should be viewed as non-binary, racial and ethnicity is far more often complex than our categories captured, and the measure of socioeconomic status via proxy indicator of free and reduced lunch. More research is needed on how schools can approach data using and intersectional lens, how effective this approach is in mitigating bias in identification, and how identification processes support the delivery of targeted interventions to subgroups of youth with unique and diverse needs.

Conclusion
As schools and school mental health professionals recognize the need for individualized supports, there is an imminent need to consider how systemic patterns of inequality and oppression play out in their schools and their effect on today's youth. Using an intersectionality framework, the identification of nonacademic needs and cultural strengths via the use of cluster analysis has the potential to help schools recognize systemic patterns of oppression and intervene to work toward equity by highlighting and celebrating their cultural differences, while also intervening to mitigate biases in identifying behaviors among subgroups experiencing systemic factors linked to socioeconomic, gender, racial, and language differences in schools. Schools can utilize this approach to further individualize and contextualize how to meet the needs of all youth and tailor interventions accordingly, minimizing personal and systematic biases toward minority groups of students.