The implementation of multi-tiered systems in schools necessitates the use of screening assessments which produce valid and reliable data to identify students in need of tiered supports. Data derived from these screening assessments may be evaluated according to their classification accuracy, or the degree to which cut scores correctly identify individuals as
at-risk'' ornot-at-risk.” The current study examined the performance of mean scores derived from over 1700 students in Grades 1, 2, 4, 5, 7, and 8 using Direct Behavior Rating-Single Item Scales. Students were rated across three time points (Fall, Winter, Spring) by their teachers in three areas: (a) academically engaged behavior, (b) disruptive behavior, and © respectful behavior. Classification accuracy indices and comparisons among behaviors were derived using Receiver Operating Characteristic (ROC) curve analyses, partial area under the curve (pAUC) tests, and bootstrapping methods to evaluate the degree to which mean behavior ratings accurately identified students who demonstrated elevated behavioral symptomology on the Behavioral and Emotional Screening System. Results indicated that optimal cut-scores for mean behavior ratings and a composite rating demonstrated high levels of specificity, sensitivity, and negative predictive value, with sensitivity point estimates for optimal cut-scores exceeding .70 for individual behaviors and .75 for composite scores across grade groups and time points.