Lingraphica Co-founder and Chief Scientist, Dr. Richard Steele, summarizes research to improve the diagnosis of apraxia of speech (AOS) in adults with acquired aphasia.
The aphasiologists from the United States, Australia, and Germany collaboratively examined multiple variables as potential predictors for the presence of AOS following left-hemisphere stroke resulting in aphasia, with the goal of finding accurate, reliable, and clinically usable indicators of AOS.
The investigators recruited 72 individuals with acquired aphasia following left-hemisphere stroke, a subset of whom presented with co-occurring AOS. Presence or absence of AOS was established on the basis of expert judgment. Extensive complementary assessment data on subjects—documenting performances in linguistic, cognitive, non-speech oral motor, and motor speech tasks—were collected on a wide range of behavioral factors that could conceivably serve as AOS predictors. Where essential data were missing for application of their analytical model, investigators supplied them through imputation. The researchers then applied advanced modeling and analytical techniques to identify which performance factors were most closely aligned with the expert opinion on presence or absence of AOS in the subject sample.
The work flagged two items that, in combination, appear sufficient to distinguish participants with aphasia and co-occurring AOS from participants with aphasia alone. These were: [i] the number of speech errors observed as word length increased, and [ii] the relative vowel durations in three-syllable words with weak-strong-weak stress patterns (e.g., ‘banana’). Using these factors, the model showed good discriminative ability to distinguish between cases with and without AOS (c-index = 0.93) and good agreement between observed and predicted probabilities of presence of AOS (calibration slope = 0.94).
This work represents an innovative and valuable contribution to the field. It introduces advanced research methodologies to speech pathology research, and illustrates their use in laying the groundwork for new assessment approaches that could improve the accuracy and efficiency of AOS diagnosis. Follow-on work should focus on demonstrating replicability of results with larger sample sizes, on reducing its reliance on imputation to supply missing assessment data, and on promoting real-world introduction by supporting the development, validation, and dissemination of materials for clinical use in practice.
For further reading: K.J. Ballard, L. Azizi, J. Duffy et al., 2016. A predictive model for diagnosing stroke-related apraxia of speech. Neuropsychologia, 81:129-139.