A Method to Analyze Initial CANS Assessments to Identify Profiles of High Need Youth Served Through Family-Based Interventions
Methods: A total of 2,729 initial CANS were extracted from the agency’s electronic health record since 2009. CANS items were coded as actionable (2 or 3) or not actionable (0, 1, null), and the TAI per youth was identified. A method of classification and regression tree (CART) using linear regression was utilized to identify individual CANS actionable items which most strongly predicted the TAI for youth. Information from the resulting tree was used to make logical combinations of three-four CANS items most predictive of TAI, and the average number of TAI for each of the resulting profiles was identified.
Results: Four main predictors of severity emerged from the analysis. The most significant predictor was issues with Judgment for Risk Behaviors (“p”<0.001); followed by Environmental Influences for Juvenile Justice involvement (“p”<0.001); Frustration Management for Violence (“p”<0.001); and the presence of any Neglect, Attachment or Adjustment to Trauma related to Abuse (“p”<0.001). The profile of youth who were not actionable in any of these four areas had average 9.5 (SD 7.3) TAI. On average, youth who were actionable in 1) any one of these areas had 18-20 TAI, 2) any two areas had 26-30 TAI, 3) any three of these areas had 37-42 TAI, and 4) all four areas had 50.8 (SD 11.9) total actionable CANS items.
Implications: CART provided a method to identify the most predictive factors associated with high need youth served by the agency. Areas related to behavioral (Judgment for Risk Behaviors and Emotional Frustration for Violence), relationship (Abuse) and Ecological (Environmental Influences for Juvenile Justice involvement) steams were all strongly predictive of severity of incoming youth, and a combination of issues in all three streams resulted in the highest need youth served. This exploration validated the importance of addressing all three steams for the population of youth served.