Author(s)
Amnie Ashour, BS
Jessica A. Koos, MLS, MSEd
Sunny C. Chung, MLIS, MS
Keith A. Chadwick, MD, MS
Affiliation(s)
SUNY – Stony Brook University Renaissance School of Medicine, Stony Brook, New York
Abstract:
Background/Objectives: The Grade-Roughness-Breathiness-Asthenia-Strain (GRBAS) scale is a perceptual measure of voice characteristics graded on a 4-point ordinal Likert scale. It is a reliable, efficient method of voice evaluation in clinical practice and research. However, peer-reviewed publications using the scale often utilize statistical methods inappropriate for ordinal outcomes data. The purpose of this study was to review literature utilizing the GRBAS scale as an outcome measure and assess the validity of statistical methods utilized.
Methods: Systematic review of articles using GRBAS as an outcome measure. Following PRISMA guidelines, databases (Ovid Medline, Cochrane Database of Systematic Reviews, CINAHL, Embase, and Web of Science) were searched using focused key terms.
Results: 846 unique abstracts were identified--full texts were obtained for 746, and 609 were ultimately included. The majority (86%) collect GRBAS scores as ordinal data. Most summarize GRBAS outcomes as categories (46%) or means (40%). 80% of studies utilize statistical tests for comparison, most commonly Wilcoxon signed-rank (32%), Mann-Whitney U/Wilcoxon rank-sum (18%), correlation (18%), and Fisher’s exact (11%) tests. Outcomes are not summarized or tested appropriately in 43% and 54% of peer-reviewed articles, respectively. Journal impact factor, sample size, geographic location and modifications of the GRBAS did not impact the likelihood of utilizing appropriate statistical methods.
Conclusions: The majority of peer-reviewed literature using the ordinal GRBAS scale as an outcome measure apply inappropriate statistical methods, calling into question the validity of the study’s conclusions. Authors should take care to use appropriate statistical methods, and readers should use caution when evaluating literature.