To assess the predictive value of a pediatric screening tool by linking 2 independent databases: an educational database that includes data from standardized academic assessments administered during kindergarten and a pediatric database that includes screening results. METHODS: A database that includes results of the Survey of Well-being of Young Children (SWYC) completed during pediatric visits were linked to an educational database that includes STAR Early Literacy examinations in kindergarten. Linear multilevel regression modeling was used to examine if screening results on the developmental and behavioral sections of the most recently completed SWYC form predicted trends in the percentile rank on the STAR exam over the school year, adjusting for potential confounders. (author abstract)