Paired-Comparison Versus Cross-Sectional Approaches for Assessing Longitudinal Speech Change in Parkinson's Disease Following Deep Brain Stimulation of the Subthalamic Nucleus
Jordan L. Manes, Andrea H. Rohl, Mengyang Qiu, Kaila L. Stipancic, Daniel M. Corcos, Charity G. Patterson, Jeremy D. W. Greenlee, Kris TjadenPurpose:
Longitudinal changes in perceptual speech outcomes are typically derived from independent judgments at each time point rather than side by side within-speaker comparisons. This study evaluated a paired-comparison method for detecting longitudinal speech change in Parkinson's disease (PD) across three perceptual constructs (intelligibility, listener effort, and severity) relative to a cross-sectional method for detecting speech changes in the same cohort.
Method:
We collected recordings of the Speech Intelligibility Test at a baseline session and 12-month follow-up from speakers with PD after undergoing subthalamic nucleus deep brain stimulation implantation surgery and speakers with PD undergoing best medical management. Blinded naive listeners heard the recordings and provided cross-sectional judgments of intelligibility, listener effort, and speech severity. Another set of crowdsourced naive listeners completed a blinded online paired-comparison paradigm and judged changes in the same three constructs between baseline and 12-month sessions. We compared the sensitivity of paired-comparison measures to analogous cross-sectional measures for detecting 12-month speech changes. We then evaluated the relationship between paired-comparison ratings of intelligibility, listener effort, and speech severity using correlational analysis.
Results:
Sensitivity to speech change was comparable between paired and cross-sectional judgments of intelligibility. Cross-sectional ratings of listener effort and speech severity were more sensitive to speech change than paired judgments. All three constructs detected similar proportions of speech change at 12 months in the paired-comparison task. Paired judgments of intelligibility, listener effort, and speech severity were strongly correlated.
Conclusion:
Paired judgments are not advantageous over traditional cross-sectional approaches for measuring longitudinal speech change in PD.
Supplemental Material: