[Fool's gold standards in language screening. Sensitivity and specificity of the Hessian child language screening...

The German Kindersprachscreening (KiSS) is a universal speech and language screening test for large-scale identification of Hessian kindergarten children requiring special educational language training or clinical speech/language therapy.

To calculate the procedural screening validity, 257 children (aged 4.0 to 4.5 years) were tested using KiSS and four language tests (Reynell Development Language Scales III, Patholinguistische Diagnostik, PLAKSS, AWST-R). The majority or consensus judgements of three speech-language professionals, based on the language test results, served as a reference criterion. The base (fail) rates of the professionals were either self-determined or preset based on known prevalence rates.

Screening validity was higher for preset than for self-determined base rates due to higher inter-judge agreement. The confusion matrices of the overall index classification of the KiSS (speech-language abnormalities with educational or clinical needs) with the fixed base rate expert judgement about language impairment, including fluency or voice disorders, yielded a sensitivity of 88% and a specificity of 78%, for just language impairment 84% and 75%, respectively. Specificities for disorders requiring clinical diagnostics in the KiSS (language impairment alone or combined with fluency/voice disorders) related to the test-based consensus expert judgment was about 93%. Sensitivities were unsatisfactory because the differentiation between educational and clinical needs requires improvement.

Since the judgement concordances between the speech-language professionals was only moderate, the development of a comprehensive German reference test for speech and language disorders with evidence-based algorithmic decision rules rather than subjective clinical judgement is advocated.