Predictive Validity Analysis.

<div> The routine measurement of children’s developmental health varies across educational settings and systems. The Early Years Foundation Stage Profile (EYFSP) is a routinely recorded measure of a child’s development completed at the end of their first school year, for all children attending school in England and Wales. Despite widespread use for research and educational purposes, the measurement properties are unknown. This study examined the internal consistency and structural validity of the EYFSP, investigating whether the summed item-level scores, which we refer to as the ‘total score’, can be used as a summary of children’s developmental health. It also examined predictive validity of the total score with respect to later academic attainment and behavioural, social, and emotional difficulties. The data source was the longitudinal prospective birth cohort, Born in Bradford (BiB), and routine education data were obtained from Local Authorities. The internal consistency and structural validity of the EYFSP total score were investigated using Confirmatory Factor Analysis and a Rasch model. Predictive validity was assessed using linear mixed effects models for Key Stage 2 (Maths, Reading, Grammar/Punctuation/Spelling), and behavioural, social, and emotional difficulties (Strengths and Difficulties Questionnaire). We found that the EYFSP items demonstrated internal consistency, however, an Item Response model suggested weak structural validity (n = 10,589). Mixed effects regression found the EYFSP total score to predict later academic outcomes (n = 2711), and behavioural, social, and emotional difficulties (n = 984). This study has revealed that whilst caution should be applied for measurement of children with close to ‘average’ ability levels using the EYFSP, the EYFSP total score is an internally consistent measure with predictive validity. </div>