Predictor Importance and Classification Accuracy of Trauma Measures to the Identification of Early Literacy Skills in Kindergarteners

Background: Trauma exposure impacts children’s ability to function in their school environment. Childhood trauma is associated with lower IQ scores, delayed language skills, and lower academic achievement (Perfect et al., 2016). The lack of childhood trauma measurement in school-based screening models (Ecklund et al., 2018) perpetuates the under-identification of at-risk students particularly during their development of crucial literacy skills. Few studies offer measurement guidance for the identification of trauma-exposed students in kindergarten (Santiago et al., 2018) or examine the association of trauma exposure to literacy skills (Pears et al., 2011). The purpose of this study was to determine the utility of direct-child and inter-generational trauma measures for use in pre-existing screening models for kindergartners’ literacy skills.

Methods: Data was collected from the 4^th-year Kindergarten cohort of a multi-state longitudinal study examining contextual factors of literacy skills development. Parents completed the Child and Adolescent Trauma Screen (Sachser et al., 2017) and the UCLA PTSD Brief-Screen Index (Rolon-Arroyo et al., 2020) for their children while they completed the Adverse Childhood Experiences (ACEs; Felitti et al., 1998) respective of their childhood. Parent-rated child behavior measures and literacy screening results were included in the model. A dominance analysis was used to test predictor importance of trauma measures to kindergartners’ end of the year literacy scores while a ROC Curve analysis tested the additive effects of a multi-screener model for classification of kindergartners at-risk for poor reading and language outcomes.

Results: The sample included 145 kindergartners. Based on parent reports, 32% of students experienced an adverse event with 17% of students exhibiting trauma symptoms. All of parents (n=64) reported at least one ACE. The inclusion of trauma measures in the pre-existing screening models explained 44% of the variance (R²=0.44; F_{(9, 135)}=11.99, p<.001) for word recognition skills and explained 34% of the variance and was statistically significant (R²=0.34; F_{(9, 135)}=7.6, p<.001) for nonsense word decoding skills. When controlling early literacy skills, Parent ACEs was the only statistically significant trauma predictor (β=-2.29, t=-3.70, p<.001) of word recognition skills while the UCLA PTSD Index (β=-1.21, t=-2.02, p<.05) and the CATS Events (β=4.90, t=2.27, p<.05) were statistically significant trauma predictors of word decoding skills. Dominance Analysis demonstrated that Parent ACEs was the completely dominant unique predictor of word recognition skills of kindergartners. While controlling for the beginning of the year screening measures, the addition of trauma measures and child behaviors improved the classification accuracy of the traditional screening model for kindergartners at-risk for poor reading skills (AUC=0.83; p<.001).

Conclusions and Implications: Results produced the first empirical link between trauma and early literacy skills with Parents ACEs being a unique dominant predictor of kindergartners’ end of the year literacy scores. ROC Curve results indicated that a holistic model of identifying at-risk readers, to include both trauma measures and parent-rated child behaviors, may classify at-risk readers better than entry-level literacy skills screeners alone. Results can inform social workers’ screening activities for Trauma-Informed Care and Multi-Tiered Systems of Support models. Results provide preliminary empirical support for Whole-Child Learning Models.