Diagnostic potential of a panel of acoustic/aerodynamic voice markers in patients with dysphonia: a cluster analysis.

PURPOSE: Dysphonia is a common condition impacting quality of life and requiring integrated/multidimensional clinical and therapeutic approaches. The main aim of this investigation was to develop a cluster analysis model to stratify dysphonic patients into homogeneous subgroups, based on a panel of acoustic voice markers. Secondary aims were assessing the association of clusters with clinical diagnoses and functional outcomes.

METHODS

This retrospective study involved 268 dysphonic patients. Demographics, laryngological diagnosis (functional dysphonia, nodules, cysts, vocal fold paralysis, chronic laryngitis, polyps/focal edema, other), acoustic (fundamental frequency, Jitter, Shimmer) and aerodynamic (maximum phonation time) parameters, GIRBAS, and Voice Handicap Index-10 [VHI-10] scores were evaluated. A cluster analysis (K-mean partitioning) was conducted to identify 4 clusters, based on Jitter, Shimer, and MPT values.

RESULTS

Diagnoses significantly varied across clusters (p < 0.0001). Dysfunctional dysphonia was most common in Clusters 1, 3 and 4, while in Cluster 2 vocal fold paralysis was more frequent. Higher prevalence of nodules was found in Cluster 3, while cysts were more frequent in Cluster 4, and chronic laryngitis was more common in Clusters 1 and 3. Perceptual voice parameters showed significant differences across clusters (p = 0.0001, p = 0.0025, p = 0.0001, p = 0.0001 for G, R, B and S parameters, respectively), with Cluster 2 having the less favorable scores.

CONCLUSIONS

In this study, cluster analysis seemed to highlight subjects with vocal patterns more at risk of presenting with specific laryngeal conditions. To better characterize the clinical potential of cluster analysis models, based on acoustic/aerodynamic parameters, further studies, based on multicentric prospective series, are required.