Spectrographic Voice Analysis Protocol (SAP): Convergent, Concurrent, and Accuracy Validity

Allan Carlos França da Silva; Luiz Medeiros de Araújo Lima-Filho; Anna Alice Almeida; Hemílio Fernandes Campos Coêlho; Vanessa Veis Ribeiro; Leonardo Wanderley Lopes

doi:10.1016/j.jvoice.2023.09.009

Spectrographic Voice Analysis Protocol (SAP): Convergent, Concurrent, and Accuracy Validity

J Voice. 2023 Oct 18:S0892-1997(23)00283-7. doi: 10.1016/j.jvoice.2023.09.009. Online ahead of print.

Authors

Allan Carlos França da Silva¹, Luiz Medeiros de Araújo Lima-Filho¹, Anna Alice Almeida¹, Hemílio Fernandes Campos Coêlho¹, Vanessa Veis Ribeiro², Leonardo Wanderley Lopes³

Affiliations

¹ Universidade Federal da Paraíba (UFPB), Decision Models and Health Program, João Pessoa, Paraíba, Brazil.
² Universidade de Brasília (UNB), Speech-Language and Hearing Department, Brasília, Federal District, Brazil.
³ Universidade Federal da Paraíba (UFPB), Decision Models and Health Program, João Pessoa, Paraíba, Brazil. Electronic address: lwlopes@hotmail.com.

PMID: 37863674
DOI: 10.1016/j.jvoice.2023.09.009

Abstract

Objective: To verify the convergent and concurrent validity of the Spectrographic Voice Analysis Protocol (SAP) and its accuracy to discriminate dysphonic from nondysphonic patients.

Method: The study used 82 vowel /Ɛ/ samples and their respective narrowband spectrograms, analyzed with SAP. Cepstral peak prominence (CPP) and cepstral peak prominence smoothed (CPPS) verified the convergent validity of the SAP total score, while the general grade of vocal deviation (GG) verified the concurrent validity of the SAP total score. The ROC (receive operator curve) curve and its accuracy, sensitivity, and specificity values, positive predictive value (PPV) and negative predictive value (NPV), and positive likelihood ratio (LR+) and negative likelihood ratio (LR-) verified the accuracy of the SAP score to discriminate dysphonic from nondysphonic individuals.

Results: Dysphonic and nondysphonic had different SAP total scores. In the convergent validity, the SAP score had a weak and moderate negative correlation, respectively, with CPP and CPPS, as well as a moderate positive correlation with GG. SAP performed well in discriminating dysphonic from nondysphonic individuals (area under the curve = 82.0%; sensitivity = 91.7%; specificity = 51.7%; PPV = 93.7%; NPV = 44.0%; LR+ = 6.21; LR- = 0.53) based on the 8-point cutoff score.

Conclusion: SAP has convergent validity with CPP and CPPS and concurrent validity with GG. The SAP total score performed well in discriminating dysphonic from nondysphonic individuals. However, the specificity, NPV, and LR- values justify cautiously using SAP, always in combination with other information in clinical voice assessment.

Keywords: Acoustic analysis; Spectrography; Validation study; Voice; Voice disorders.