Abstract
This paper reports the reliability, validity and accuracy of the General Language Screen (GLS) as the first stage in its development as a potential screening instrument for the identification of speech and language disorder (SLD) in the primary care setting. Data reported were obtained during the Cambridge Language and Speech Project (CLASP) a longitudinal epidemiological study of primary SLD. The GLS was originally used in initial CLASP sample ascertainment when children were aged 36 months. This report utilises speech and language data collected at subsequent stages of this study (37-months, 45-months and 8-years) to determine the reliability, validity and accuracy of the GLS. The GLS demonstrated sufficient levels of reliability (internal consistency f =0.74) and validity (criterion validity r =-0.55). Adequate construct validity was demonstrated by a factor analysis in which two language related factors were identified. These data therefore justified exploration of the accuracy of the GLS as a screening instrument. Analysis of accuracy in this context is complicated by lack of a generally accepted method of defining SLD case status. This report utilises data on language function at different stages of development to provide measures of GLS accuracy over time. Concurrent accuracy, measured against parental report of language function at 37-months, was good (ROC area under the curve=0.93). Predictive accuracy measured against direct assessment of speech and language at 45-months and 8-years was, as expected, somewhat less (ROC=0.77 and 0.68 respectively). The implications for the use of the GLS as a screening instrument are discussed. The data indicate that the GLS would function most usefully as the first stage of a two-stage screen for SLD in the primary care setting. Recommendations are made as to the nature of the second stage of such a screen and plans for future research are outlined.