This study evaluates the predictability of a new language screening procedure in 3-y-olds. It is used in several Child Health Centres (CHC) in Sweden and has the character of a field study involving more than 60 CHC nurses. The main questions concern the (i) development in 3-y-olds assessed as severely language delayed and (ii) whether there are any earlier unknown severely disabled children identified at 4 y of age. Ninety-six percent of the original study population participated in the follow-up. The calculations are based on results from 2237 children. A well-established screening routine, which has been shown capable of predicting the risk of not being able to follow expected schooling, and case records were used as an acceptable proxy outcome measure, pending a better gold standard. In the group of severely disabled 3-y-olds, sensitivity, specificity and positive predictive values were 86, 99 and 43%, respectively. Finally, three false-negatives were identified. In light of the present results, continued application of the 3-y screening is discussed.