We used existing reading (n = 1,498) and math (n = 2,260) data to evaluate state test scores for screening middle school students. In Phase 1, state test data were used to create a research-derived cut score that was optimal for predicting state test performance the following year. In Phase 2, those cut scores were applied with future cohorts. Diagnostic accuracy of the research-derived cut scores was compared with the state’s proficiency benchmark from the previous year. Across grades and content areas, research-derived cut scores yielded higher sensitivity and lower specificity values relative to state-defined cut scores. Marked decreases in sensitivity and specificity were not observed in subsequent years. Results provide evidence for procedures in which previous state test data are repurposed for screening decisions.