Predicting Academic Performance on High-Stakes Tests: Development of the Brief Behavior, Academic, and Social Screener (B-BASS)

In order to develop the Brief Behavior, Academic, and Social Screener (B-BASS), an experimental measure for identifying students at risk for academic underachievement and for predicting performance on high-stakes tests, data were obtained from elementary teachers for 198 third- through fifth-grade rural elementary students in the Southeastern United States. Teachers provided ratings on items within the following global domains: cognitive ability, social/emotional traits, socioeconomic and family characteristics, and executive functions using a brief, practical, contextually appropriate format, one sensitive to local norms. Results reveal strong test-retest correlations (r indices > .95) and moderate to high inter-correlations among the items. Based on a multiple linear regression analysis, B-BASS domain scores account for 81.4% and 66.2% of the variance in North Carolina End-of-Grade Tests of English Language Arts/Reading and Mathematics, respectively. These values are impressive given the brief administration time of the B-BASS, approximately 1 minute per student. Based on a stepwise analysis, two items significantly predict English Language Arts/Reading; specifically, Reading Ability (p < .001) and Effort (p < .005). Predictors of Mathematics in order of magnitude are Math Ability (p < .001) and Hyperactivity (p < .001). According to discriminant function and ROC curve analyses, B-BASS items also significantly predict proficient vs non-proficient status from the ELA (p < .001) and Math (p < .001) EOGs, with true positive rates of 93.9% and 85.4%, respectively. B-BASS composites predict 78.5% of variance in ELA EOG scores, with the strongest predictors being Academic Ability (p < .001), Social Adaptability (p < .05), and Executive Function (p < .005). For the EOG Mathematics scores, the strongest composite predictors, in order of magnitude, are: Academic Ability (p < .001), Overall Ability (p < .05), Home Support (p < .05), and Executive Function (p < .05); these domain scores predict 66.2% of the variance. Composites also predict significantly categorical status (proficient vs non-proficient) on the ELA (p < .001) and Math (p < .001) EOGs; classification accuracy was determined to be 92.9% and 91.4%, respectively. Implications of using B-BASS are discussed.