Predicting patterns of developmental change: language development into primary school

Abstract

Objectives: Although it is possible to examine population subgroups using test cut point it is preferable to examine the performance of such groups across time if measures allow. This analysis will compare the profile of childrens language development between three and five years drawing on data (n=11788) in the UKs Millennium Cohort Study. Methods: Following clinical convention, the threshold for caseness was set at -1.5 standard deviations below the mean on the British Ability Scales, a well recognized standardized assessment of language development. We identified four groups based on their profiles across time on the British Abilities Scales. Group 1 comprised those whose scores fell 1.5 SD below the mean at both time points (n=199), Group 2 made up those whose scores deteriorated across time i.e. starting within normal limits and falling outside the normal range at five years (n=234), Group 3 comprised those who appeared to be resilient i.e. starting below the norm and improving across time (n=257) Group 4 comprised those whose scores remained within normal limits between 3 and 5 (n=11098). Multinomial regression was employed to assess the influence of these variables on the first three groups relative to the fourth. Potential predictors included gender, whether the child was small for dates, whether they had been in a special care baby unit, had long term health conditions, report of parental concern about hearing at nine months, attachment and developmental performance at nine months. Conclusion: The findings will be discussed in terms of their implications for our understanding of developmental change which has tended to emphasise continuities rather than discontinuities in child development. We will also interpret the data in terms of current health and educational policy and the popular desire to implement early screening procedures to identify at risk children. We argue that risk itself is not sufficient to warrant universal procedures which tend not to be very strong predictors even across such a narrow age span and using identical measures.