This study investigated the predictive validity of universal screening in reading. Especially, this study focused on long-term predictive validity. The participants were 237 students from 3 elementary schools, and data were collected from school entry to the end of Year 2. Correlation coefficients, ROC curve analysis, and latent curve modeling were applied to answer the main research questions suggested in this study. The results showed that maze Korean CBM slope and score are strongly associated with reading achievement test and efficient in predicting students who are at risk for reading after 2 years of schooling. Implication and future directions for research are suggested.