Voice Quality in Pediatric Dysphonia: Assessing Expert Rater Reliability Across Three Perceptual Methods.

OBJECTIVE

As with adult voices, the reliability of Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V) appears to vary widely for pediatric voices, especially for supraglottal voice sources (SGVS). The current study compared the reliability of estimates of breathy, rough, and strained voice qualities in pediatric dysphonic voices using laboratory-based methods, such as visual sort and rank (VSR) and magnitude estimation (ME), and the clinical CAPE-V method.

METHODS

A total of 124 samples (30 SGVS and 94 glottal vibratory source or GVS) from children aged 4 to 11 years were selected from the Cincinnati Children's Hospital Medical Center (CCHMC) database. Three expert listeners rated breathiness, roughness, and strain of 1000-ms /ɑ/ samples using VSR, ME, and CAPE-V. In the VSR task, experts were presented with subsets of voice samples and were encouraged to sort and rank them into a relative order by comparing each sample to another. In the ME task, experts assigned the voice quality magnitude a number between 1 and 1000.

RESULTS

High intra- and inter-rater reliability was observed for each of the perceptual measurement methods. For intra-rater reliability, the ME task provided the highest reliability across the two vibratory sources and the three voice quality dimensions. For inter-rater reliability, the ME task provided higher values for SGVS samples, while VSR provided higher values for GVS samples. There was also a strong relationship between the three methods with high correlations.

CONCLUSIONS

Among three expert voice clinician-scientists, inter- and intra-rater reliability was similar for the offline CAPE-V assessment and the laboratory methods of VSR and ME. However, there appears to be a slight variation in absolute reliability values between the two vibratory sources. It is unclear how this highly controlled comparison extends to typical clinical practice. High reliability among expert professionals can potentially be translated to other clinicians and inexperienced listeners using superior ratio-level scaling procedures provided by the ME task.