Abstract
Objective: The study aimed to confirm the intertext robustness of the Acoustic Voice Quality Index for the Japanese-speaking population using reading tasks derived differently from those initially reported.
Methods: Two reading texts were extracted from a passage with 58 syllables taken from the Japanese translation of 'The North Wind and the Sun': one was the same as in the original study, namely 30 syllables from the first part of the passage, and the other was the remaining 28 syllables. The AVQI values were calculated from reading samples of these texts (AVQI30 and AVQI28). The concurrent validity and diagnostic accuracy of AVQI30 and AVQI28 for estimating hoarseness levels were evaluated on 311 combined samples of continuous speech and sustained vowels.
Results: Both AVQI30 and AVQI28 showed sufficient concurrent validity (r=0.850 and r=0.842, respectively). Subsequently, receiver operating characteristic analysis showed the area under the curve tobe 0.897 and 0.892, indicating the excellent discriminative capability of AVQI30 and AVQI28. In AVQI30, the original threshold of 2.06 corresponded to sensitivity and specificity of 80% and 85%, respectively, whereas in AVQI28, 83% and 75%, respectively. The absolute differences between them were judged to be negligible.
Conclusion: The results demonstrated the probability of the intertext robustness of AVQI.
