Test-Based Versus Teacher-Based Predictions of Academic Achievement: A Three-Year Longitudinal Follow-Up

The predictive validity and utility of two different kindergarten screening approaches were compared in a large sample of white males followed longitudinally from kindergarten to Grade 2. Predictions of reading achievement to the end of Grade 2 revealed comparable overall hit rates for the test-based and teacher-based approaches. However, the two approaches yielded different conditional (posterior) probabilities for predictions of high vs. low risk status. Teacher-based predictions were characterized by low false positive rates and high false negative rates. While predictions of high risk were highly accurate (p = .86), they were seldom made, which resulted in teachers missing 87% of the severely disabled readers. In contrast, test-based predictions were characterized by higher false positive rates and lower false negative rates. While predictions of high risk were less accurate (p = . 71), the screening battery identified a much greater proportion of the high-risk children, missing only 32% of the severely disabled readers. Combining teacher-based and test-based predictions yielded results that were not very different from test-based prediction alone. These results highlight some of the issues concerning validity and utility that must be considered in evaluating different screening approaches in early identification.