Progress Toward Estimating the Minimal Clinically Important Difference of Intelligibility: A Crowdsourced Perceptual...

Area of Science:

• Speech-language pathology
• Neurology
• Biostatistics

Background:

• Assessing speech intelligibility is crucial for individuals with neurological conditions like multiple sclerosis (MS) and Parkinson's disease (PD).
• Establishing a minimal clinically important difference (MCID) provides a benchmark for meaningful change in speech intelligibility.
• Current methods for evaluating intelligibility changes may lack sensitivity to clinically relevant differences.

Purpose of the Study:

• To estimate the MCID for sentence intelligibility in control speakers and in speakers with dysarthria due to MS and PD.
• To validate a novel experimental paradigm using crowdsourced perceptual data for MCID estimation.
• To establish thresholds for important changes in speech intelligibility.

Main Methods:

• Audio-recorded sentences from control speakers and individuals with MS and PD across various speaking conditions.
• Crowdsourced nonexpert listeners rated intelligibility using paired comparisons and the Global Ratings of Change (GROC) Scale.
• Listener ratings were correlated with objective intelligibility scores, and Receiver Operating Characteristic (ROC) curves were analyzed.

Main Results:

• MCIDs derived from ROC curves were not valid.
• Average magnitude of intelligibility difference provided valid thresholds for MCID.
• The MCID for intelligibility was determined to be approximately 7% for a small difference and 15% for a large difference.

Conclusions:

• The study demonstrated a feasible method for collecting crowdsourced perceptual data to estimate MCIDs.
• Findings support the development of clinical tools categorizing intelligibility changes into 'no difference,' 'a little bit of difference,' and 'a lot of difference.'
• This research is a step towards a universal language for evaluating intelligibility changes in neurological conditions and speech therapy.