Tailoring Services to Youth in Schools: Use of Cluster Analysis to Identify Those at-Risk

Background and Purpose: Schools must consider the diverse perspectives of all youth in order to understand and improve youth outcomes in different contexts. To date, research shows positive social skills and peer relationships are critical for promoting positive behavioral and academic outcomes in schools (Furrer, Skinner, Pitzer, 2014; Tan, Oe, & Hoang, 2018). Studies also confirm that race, sex, and other demographic indicators are predictive of how youth behave and experience school (Fan, Williams, & Corkin, 2011). Using an exploratory approach, the current study sought to understand if and to what extent different subgroups of youth in one high school perceive their social skills, peer relationships, and psychosocial behaviors. 

Methods: This study used cross-sectional survey data from 1,247 youth attending a large urban high school. Data were collected as part of larger battery of tools using the Community and Youth Collaborative Institute School Experience Surveys (CAYCI-SES). The following CAYCI-SES scales were used: Internalizing Behaviors, Externalizing Behaviors, Social Skills, and Peer Relationships (Anderson-Butcher, Amorose, Iachini, & Ball, 2013). The analytic strategy was twofold. Cluster analysis was used to form homogeneous subgroups based on a combination of race/ethnicity, socioeconomic status (measured by proxy indicator of free and reduced lunch; FRL), English language learner status (ESL), and gender. Next, a multivariate analysis of variance (MANOVA) was used to compare cluster profiles on measures of youths' psychosocial behaviors, social skills, and peer relationships.

Results: The cluster analysis revealed a 5-factor solution: 1) White, females, no report of FRL and no ESL; 2) White, males, no report of FRL, and no ESL; 3) mixed race (58% White), mixed gender (51% male), report of FRL, and ESL; 4) mixed race (32% Asian), mixed gender (54% female), no report of FRL, and mixed ESL (62%); and, 5) mixed race (75% Hispanic/Latino), mixed gender (60% female), report of FRL, and ESL. Findings of the MANOVA showed that the clusters differed significantly on the set of youth 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. To demonstrate several findings, Cluster 4 reported significantly higher internalizing behaviors, lower externalizing behaviors, yet reported higher social skills and better peer relationships compared to youth in the other clusters (p < 0.05). Further, Cluster 1 had significantly higher internalizing behaviors compared to Clusters 3 and 5 (p < 0.05).

Conclusions and Implications: The aim of the current study was to support the school in identifying needs and tailoring services to youth in the school. Results point to cluster analysis as a potential way to identify needs and tailor nonacademic supports. For example, youth identified in cluster groupings could receive more specialized mental health or nonacademic supports, in addition to universal interventions to meet school wide priorities. Further, findings point to the need for interventions aimed toward social skill development and improved peer relationships among specific subpopulations of youth. Schools and districts can use this approach to identify and address youth behaviors and in turn improve their school performance.