Speech Auditory Brainstem Response to Predict Language Delay.

Area of Science:

• Neuroscience
• Developmental Psychology
• Speech-Language Pathology

Background:

• Preterm birth is a known predictor of language delay but lacks precision for individual child assessment.
• Early intervention (EI) is highly effective for language delay but requires accurate, child-level prediction.
• Current predictive methods are insufficient for timely EI prescription before preschool.

Purpose of the Study:

• To develop and validate predictive models for language delay using infant neural data.
• To forecast language delay to enable preemptive early intervention.
• To utilize electroencephalography (EEG) neural speech encoding for early identification of at-risk children.

Main Methods:

• Recorded EEG neural speech encoding (speech auditory brainstem response [ABR]) from 423 infants (1-24 months).
• Collected language outcomes using the Bayley Scales of Infant and Toddler Development (7-32 months).
• Employed random forest models to compare predictive accuracy with and without EEG measures.

Main Results:

• Models incorporating EEG measures significantly outperformed non-neural clinical predictors.
• EEG-only models achieved >90% sensitivity and AUC, with sustained >80% sensitivity and >90% AUC upon external validation.
• Non-neural models (gestational age, birth weight) predicted outcomes above chance but were less accurate than EEG models.

Conclusions:

• Speech ABR shows promise as a novel screening tool for identifying infants at risk of language delay.
• Early identification via speech ABR enables timely EI, potentially enhancing language development.
• This neural-based approach offers a more precise method for child-level prediction compared to existing methods.