Development of a language screening instrument for Swedish 4-year-olds

Background: The Swedish Program for health surveillance of preschool children includes screening of language and communication abilities. One important language screening is carried out at age 4 years as part of a general screening conducted by health nurses at child health centres. The instruments presently in use for this screening mainly focus on expressive phonology. This may result in both over-referral of children with phonological difficulties and under-referral of children with language disorders (LDs), involving difficulties with vocabulary, grammar and/or language comprehension. Previous research has proposed non-word repetition as a clinical marker for LD. It has also been found that higher predictive power is achieved when non-word repetition is combined with the assessment of lexical/semantic skills. Taking these findings into account, the construction of a language screening instrument may yield more adequate referrals to speech-language therapists (SLTs). Aims: To construct a new standardized language screening instrument for 4-year-olds and to test its properties. Methods & Procedures: An instrument was developed and revised after piloting. A population of 352 children was screened at the regular 4-year check-up by 11 health nurses. The final sample consisted of 328 children aged 46-53 months (23% multilingual). Children performing below a preliminary cut-off were referred to an SLT (n = 52). Five SLTs carried out an assessment on average within 5 weeks using a gold standard language test battery. Children who screened negatively were followed up with a parent questionnaire at age 5;6. Outcomes & Results: Thirty-one true-positives and 11 false-positives were identified after SLT assessment. A further six children were identified as false-negatives (two through referral to an SLT and four through parent questionnaire at age 5;6). A receiver-operating characteristics curve with a C statistic of .94 was calculated. Based on optimal cut-off, the sensitivity of the screening instrument was found to be .84, and specificity was .96. Multilingual children performed similar to monolingual children; boys performed significantly lower than girls; and children with a family history of language-related problems performed lower than those without. Interrater reliability was high, as was Cronbach's alpha. Conclusions & Implications: The screening instrument seems sufficiently valid for its purpose to identify children who need further assessment by an SLT. A follow-up study including SLT assessment for all children to check for false-negatives would be interesting in future, as would studies comparing results from the 4-year screening with those from earlier screens