Validation of a Questionnaire Based Checklist Identifying 3 Year Olds with Delayed Language In a Prospective Birth Cohort

Background: Language delay is an early developmental marker for a subgroup of children with autism spectrum disorders (ASD). Assessment tools that address language delay in questionnaire based epidemiological studies may enable early identification of children with milder forms of ASD and provide insights into the trajectories of language development in ASD and other neurodevelopmental disorders. The Norwegian Mother and Child Study (MoBa) is a pregnancy cohort comprising more than 100,000 participating children. It includes questionnaires throughout pregnancy and when the child attains the ages of 6, 18 and 36 months and 5, 7 and 8 years that focus on child health and development as well as on mothers’ and fathers’ health and well-being. The quality of epidemiological studies based on questionnaires depends on the validity of the selected instrument. Ideally, an instrument should be validated both in each study population and in other populations.

Objectives: To validate child language measures included in the 36-month maternal questionnaire in MoBa, we used clinical data collected in a nested case-cohort study of ASD, the Autism Birth Cohort study (ABC). Children were selected for clinical assessment at approximately three years of age 

Methods: Subjects: The validation sample consisted of 444 children aged 36 to 42 months (286 boys, 158 girls; 198 controls, 246 cases). Measures: Language complexity [LC] (Dale and Bishop, 2003) and Expressive and Receptive language competence (Ages and Stages Questionnaire [ASQ-C], Bricker & Squires, 1999) was rated by mothers on the 36-month MoBa questionnaire. Instruments used at clinical assessment were: Child Development Inventory [CDI] (Ireton, 1992), an age-normed expressive language scale completed by both mothers and preschool teachers; Vineland Communication domain, a standardized interview conducted with mothers [VABS-C] (Sparrow et al 1984); a language subscale on the Stanford-Binet 5 Knowledge [KN] (Roid 2003) measuring the child’s vocabulary.

Results:  The correlation (Pearsons) between the two MoBa language measures LC and ASQ-C was 0.868** (p<0.0001). Low scores on LC (less than 3 word sentences) predicted low standard scores on VABS-C (β= .818**;p<0.0001) as did low scores for ASQ-C (β= .827**;p<0.0001). The children scored differently on VABS-C depending on parental rating on the LC-measure in the MoBa questionnaire. There were no overlapping confidence intervals. The relationship between the mothers’ and teachers’ scoring of expressive skills on the CDI showed high concordance (β=0.881*;p<0.001). A Bland-Altman plot of this relationship showed that there were low levels of disagreement, with little change across the level of expressive language skills. Mean difference between mother and teacher ratings on CDI constituted only 3 months in language age, less than 0.5SD.

Conclusions: Overall consistency in maternal reporting of their children’s language skills was high. MoBa questionnaire scores on language skills provide excellent measure of speech and language development.