Dysphonia Severity Index

1. Abstract

The Dysphonia Severity Index (DSI) is an objective, multiparametric acoustic and aerodynamic assessment tool designed to quantify perceived voice quality and severity of dysphonia. Developed by Floris L. Wuyts, Marc S. De Bodt, and colleagues (2000), the instrument synthesizes four distinct physiological, aerodynamic, and acoustic measures: maximum phonation time (MPT, measured in seconds), highest fundamental frequency ($F_0\text{-high}$, measured in Hertz), lowest vocal intensity ($I\text{-low}$, measured in decibels sound pressure level, dB SPL), and acoustic perturbation in the form of jitter (expressed as a percentage). The derived mathematical formula weighs these continuous variables through multivariate discriminant analysis to produce a single continuous numerical score that reflects overall vocal function. The DSI score typically ranges along an open-ended continuum where a score of approximately +5 indicates perceptually normal or superior vocal capacity, whereas progressively negative scores (e.g., reaching -5 or below) reflect severe dysphonia. The instrument was constructed to establish an objective surrogate for subjective perceptual voice evaluation, specifically anchoring its parameters against the Grade component of the internationally recognized GRBAS scale (Hirano, 1981). Across clinical phonetics, laryngology, and speech-language pathology literature, the DSI exhibits robust convergent validity with perceptual ratings, aerodynamic measures, and patient-reported outcome measures such as the Voice Handicap Index (VHI). Its clinical utility spans the pre- and post-therapeutic tracking of phonosurgical outcomes, voice rehabilitation monitoring, and functional voice assessment. Test-retest reliability across standardized recording conditions demonstrates strong stability ($r > .80$), though physiological factors such as vocal fatigue, standardized vocal coaching, and instrumentation calibration may introduce variability. By bridging physical acoustic measurements and psychoacoustic voice evaluation, the DSI represents a fundamental cornerstone in objective clinical voice pathology.

2. Keywords

Dysphonia Severity Index, voice quality, objective voice assessment, acoustic analysis, maximum phonation time, jitter, vocal intensity, fundamental frequency, speech pathology, GRBAS scale

3. Authors

The Dysphonia Severity Index was developed through collaborative clinical and biostatistical research at the University of Antwerp and Antwerp University Hospital in Belgium:

• Floris L. Wuyts, Ph.D. — Biomedical physicist and biostatistician; Professor and Director of the Laboratory for Equilibrium Investigations and Aerospace (LEIA), Department of Biomedical Physics, University of Antwerp, Antwerp, Belgium. Key contributions involve multivariate statistical modeling, acoustic signal processing, and vestibular-auditory biomechanics.
• Marc S. De Bodt, Ph.D. — Speech-language pathologist and voice scientist; Professor in Voice and Speech Pathology, Department of Otorhinolaryngology, Head and Neck Surgery, Antwerp University Hospital, and University of Antwerp, Antwerp, Belgium. Specializes in neurogenic and organic voice disorders, voice perceptual evaluation, and rehabilitation outcomes.
• Paul H. Van de Heyning, M.D., Ph.D. — Otorhinolaryngologist; Professor and Former Chairman of the Department of Otorhinolaryngology, Head and Neck Surgery, Antwerp University Hospital, University of Antwerp, Antwerp, Belgium. Contributed clinical diagnostic oversight and phonosurgical evaluation.
• Collaborating Researchers: Clinical voice pathologists, ENT surgeons, and phoniatricians affiliated with the Antwerp Voice Clinic, including Chantal Molenberghs, Francis Dirix, and Patrick Van Lierde, who assisted in patient cohort standardization and acoustic acquisition protocols.

4. Purpose

The primary purpose of the Dysphonia Severity Index is to provide an objective, standardized, and reproducible metric for the quantification of dysphonia severity. Voice clinicians and speech scientists have long encountered substantial difficulties when relying solely on subjective perceptual voice analysis. While perceptual assessment scales, such as the GRBAS framework (Hirano, 1981) or the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V; Kempster et al., 2009), remain foundational clinical standards, they are inherently susceptible to inter-rater and intra-rater variability, listener fatigue, perceptual adaptation, and cognitive bias. Conversely, isolated acoustic metrics (such as isolated jitter, shimmer, or fundamental frequency) fail to capture the multidimensional complexity of vocal fold vibration, aerodynamic competence, and vocal tract resonance. The DSI resolves this tension by integrating aerodynamic capacity, vocal range flexibility, dynamic intensity control, and glottal periodicity into a single predictive composite score.

In clinical practice, the DSI serves multiple purposes across diagnostic, therapeutic, and longitudinal monitoring dimensions. In diagnostic phoniatrics and speech pathology, it functions as an adjunct classification tool that differentiates euphonic voices from dysphonic voices. It provides objective baseline parameters for patients presenting with functional dysphonia (e.g., muscle tension dysphonia), benign vocal fold lesions (e.g., vocal fold nodules, polyps, cysts, and Reinke’s edema), neurogenic voice disorders (e.g., vocal fold paresis or spasmodic dysphonia), and presbyphonia. In post-treatment settings, the DSI provides an objective index of treatment efficacy following phonosurgery, botulinum toxin injection, or behavioral voice therapy, allowing clinicians to demonstrate therapeutic outcomes to third-party payers, hospital audit committees, and patients.

In research contexts, the DSI provides a continuous, ratio-interval surrogate measure for overall vocal deviation. This enables statistical modeling without the methodological limitations inherent to ordinal or visual analog rating scales. By utilizing objective acoustic software (such as the Multidimensional Voice Program [MDVP] or open-source environments like Praat), researchers can compare vocal outcome cohorts across clinical trials, epidemiologic investigations of occupational voice users (e.g., teachers, telemarketers, and professional vocalists), and ergonomic vocal interventions.

5. Psychological Construct

The core construct measured by the Dysphonia Severity Index is multidimensional voice quality deviation, conceptualized as the biomechanical and acoustic disruption of normal laryngeal sound generation. Voice quality is not a univariate psychological or physical phenomenon; it represents a complex psychoacoustic percept arising from the interaction of respiratory driving pressure, viscoelastic properties of the vocal fold cover and body, glottal closure patterns, and supraglottic resonance. The DSI operationalizes this construct across four physiological and acoustic subcomponents:

Aerodynamic and Respiratory Efficiency (Maximum Phonation Time – MPT)

Maximum phonation time reflects the efficiency of glottal closure and the coordinated management of transglottic airflow against the subglottic air reservoir. Measured in seconds during sustained production of the vowel /a/ at comfortable pitch and loudness following maximal inspiration, MPT evaluates phonatory respiratory capacity and glottal competency. In patients with incomplete glottal closure (such as unilateral vocal fold paralysis, presbylaryngis, or post-excisional tissue deficits), air leaks rapidly through the glottic gap, yielding markedly reduced MPT values (often $< 10$ seconds, compared to normal physiological baselines of $15-25$ seconds). In the DSI equation, higher MPT positively influences the composite score, representing healthier phonatory coordination.

Longitudinal Laryngeal Biomechanics and Biomechanical Flexibility ($F_0\text{-high}$)

The highest fundamental frequency ($F_0\text{-high}$) reachable by the patient, measured in Hertz (Hz), assesses the biomechanical integrity, elasticity, and functional capacity of the vocal fold tissues, particularly the cricothyroid muscle and the pliable layers of the lamina propria. Pathological processes that increase vocal fold mass, impair passive tissue stretching, or introduce structural stiffness (e.g., severe Reinke’s edema, vocal fold scarring, inflammation, or recurrent laryngeal nerve damage) severely limit upper vocal pitch. By incorporating $F_0\text{-high}$, the DSI captures functional range deficits that are not evident during comfortable conversational pitch.

Phonatory Glottal Attenuation and Dynamic Control ($I\text{-low}$)

Lowest vocal intensity ($I\text{-low}$), recorded in decibels sound pressure level (dB SPL) at a standardized distance (typically $30\text{ cm}$), measures the minimum subglottic pressure and subtle glottal approximation needed to initiate and sustain stable vocal fold oscillation (often related to phonation threshold pressure). Patients with hyperfunctional voice disorders, severe mucosal wave disruptions, or incomplete closure struggle to phonate softly; vocal fold vibration either arrests entirely or requires excessive muscular force, preventing quiet phonation. Consequently, dysphonic individuals exhibit elevated $I\text{-low}$ values (e.g., failing to drop below $65-70\text{ dB SPL}$). In the DSI formula, a higher $I\text{-low}$ reduces the final score, reflecting decreased dynamic laryngeal control.

Short-Term Acoustic Periodicity (Jitter %)

Acoustic jitter represents the cycle-to-cycle perturbation in the fundamental frequency of vocal fold oscillation. Measured as local jitter percentage during sustained phonation, this parameter quantifies the regularity of mucosal wave travel across the vocal fold edge. Pathological lesions, asymmetry in laryngeal mass or tension, neuromuscular tremor, and turbulent airflow all destabilize glottal cycle periodicity, driving jitter percentages well above the normal threshold of $1.0%$. Within the DSI mathematical algorithm, increased jitter substantially lowers the resulting score, serving as a direct marker of acoustic aperiodicity, breathiness, and roughness.

6. Theoretical Framework

The theoretical framework underpinning the Dysphonia Severity Index is rooted in the myoelastic-aerodynamic theory of phonation, originally formulated by Johanneswillem van den Berg (1958) and subsequently expanded by Ingo Titze and modern voice biomechanicists. According to this framework, vocal fold oscillation is sustained via an intricate equilibrium between aerodynamic forces (subglottic pressure and Bernoulli-induced transglottal pressure gradients) and passive/active tissue viscoelasticity (myoelastic restoration, cover-body layered mechanics, and vocal tract acoustic impedance matching).

Pathology disrupts this balance. Organic lesions (such as nodules or polyps) alter both the mass and edge profile of the vocal fold cover; neurological deficits alter active thyroarytenoid or cricothyroid muscle tension; functional misuse induces excessive muscular constriction. Under the myoelastic-aerodynamic framework, these disruptions manifest across four domains:

1. Glottal incompetence, causing premature transglottal airflow escape (reduced MPT).
2. Loss of tissue compliance, preventing high-frequency longitudinal tensioning (reduced $F_0\text{-high}$).
3. Increased phonation threshold pressure, precluding low-intensity aerodynamic equilibrium (elevated $I\text{-low}$).
4. Disrupted micro-mechanical mucosal travel, producing cycle-to-cycle vibratory instability (elevated jitter).

From a psychometric perspective, the DSI operates within classical test theory, conceptualizing the four physiological indicators as continuous manifest variables driven by a single unobserved latent construct: total dysphonia severity. Wuyts and colleagues (2000) employed multivariate linear discriminant analysis to establish mathematical weighting coefficients for these four parameters. The reference standard against which these weights were derived was the auditory-perceptual ‘Grade’ ($G$) parameter from the GRBAS scale (Hirano, 1981). The Grade component reflects the overall degree of voice deviance perceived by trained human listeners. By modeling objective aerodynamic and acoustic variables to predict human psychoacoustic auditory categorization, the DSI effectively synthesizes biomechanical acoustics and psychophysical auditory perception.

7. Validity

The psychometric validity of the Dysphonia Severity Index has been systematically evaluated across adult voice populations, cross-cultural adaptations, and diverse clinical disorders.

Construct and Criterion-Related Validity

In the original validation study by Wuyts et al. (2000), involving 70 normal subjects and 387 dysphonic patients presenting with varied etiologies, the DSI demonstrated strong criterion-related concurrent validity against the perceptual Grade ($G$) of the GRBAS scale. The correlation between the calculated DSI score and perceptual severity was high ($r = -.84, p < .001$), indicating t\hat as voice impairment increases (higher$G$ score), the DSI score decreases systematically. Subsequent investigations by De Bodt et al. (2000) confirmed that the DSI distinguishes between perceptual severity tiers ($G_0, G_1, G_2, G_3$) with statistically significant group separation ($F > 120.0, p < .001$). Post-hoc analyses confirm significant pairwise discrimination between all ordinal levels of vocal severity.

Convergent and Discriminant Validity

Convergent validity has been established by evaluating correlations between the DSI and other multidimensional or patient-reported indices. Multiple studies have documented significant negative correlations between the DSI and total scores on the Voice Handicap Index (VHI), with correlation coefficients ranging from $r = -.45$ to $r = -.68$ ($p < .001$) (e.g., Hakkesteegt et al., 2006; Wheeler et al., 2006). This moderate-to-strong correlation is clinically expected: while the DSI directly measures biomechanical acoustic severity, the VHI captures psychosocial, functional, and emotional handicaps, which can decouple based on individual occupational demands and psychological resilience.

The DSI also exhibits convergent validity with alternative objective acoustic models, including the Acoustic Voice Quality Index (AVQI) developed by Maryn et al. (2010), demonstrating significant alignment ($r \approx -.75$ to $-.82$). Discriminant validity is supported by the DSI’s capacity to differentiate between distinct pathological subtypes. For example, patients with mass lesions (polyps, nodules) present with severe pitch restriction and high jitter, whereas patients with adductor spasmodic dysphonia present with erratic aerodynamic measures and pronounced jitter fluctuations, separating them clearly from euphonic reference cohorts.

Predictive and Responsiveness Validity

The instrument is sensitive to longitudinal physiological improvements following clinical voice interventions. Numerous clinical trials tracking microflap phonosurgery, laser ablation, injection medialization, and intensive speech therapy have reported significant shifts in DSI scores toward positive values post-intervention (e.g., pre-treatment means shifting from $-1.8$ to post-treatment means of $+2.4$; $t > 8.5, p < .001$), verifying high therapeutic responsiveness.

8. Reliability

Because the DSI is an objective, instrument-derived multiparametric algorithm rather than a psychometric self-report questionnaire composed of traditional Likert items, internal consistency metrics such as Cronbach’s alpha are not conceptually applicable to the composite formula. Instead, psychometric reliability is established through test-retest reproducibility, inter-trial consistency, and acoustic measurement stability across testing environments.

Test-Retest Stability

Hakkesteegt et al. (2008) conducted rigorous test-retest evaluations on dysphonic patients and vocally healthy controls to quantify measurement stability across repeated sessions. In vocally stable adults assessed on two separate days under identical acoustic conditions, the intraclass correlation coefficient (ICC) for the composite DSI reached $.84$ (95% CI $[.76, .90]$), demonstrating high stability. The standard error of measurement (SEM) was calculated at approximately $0.94\text{ DSI units}$, yielding a Minimum Detectable Change (MDC) or Critical Difference value of approximately $2.6\text{ DSI units}$ at the 95% confidence level. This indicates that a clinical shift exceeding $2.6$ points represents a genuine physiological alteration rather than biological or measurement variability.

Subcomponent Sources of Variance

Individual reliability analyses of the four subcomponents reveal varying levels of measurement stability:

• $F_0\text{-high}$: Exhibits strong test-retest reliability ($r > .85$); however, it is susceptible to clinician elicitation technique and patient effort. Proper glissando coaching is required to reach true physiological ceilings.
• $I\text{-low}$: Shows moderate-to-high reliability ($r = .72-.80$), but requires strict calibration of sound-level meters, microphone-to-mouth distance ($30\text{ cm}$), and low ambient room noise ($< 40text{ dBA}$).
• MPT: Subject to physiological variability and learning effects. Literature recommends recording three consecutive trials and selecting the maximum value, yielding reliable baselines ($ICC \approx .80$).
• Jitter %: Highly reliable ($ICC > .88$) provided that sustained vowel recordings are free from vocal fry and ambient acoustic reflections.

9. Factor Analysis

The architectural construct of the DSI was not developed through exploratory factor analysis (EFA) or confirmatory factor analysis (CFA) in the classical survey-instrument tradition. Instead, it was derived via multivariate stepwise discriminant function analysis, a multivariate structural regression framework designed to identify independent continuous variables that maximize group discrimination across clinical clusters.

Discriminant Analysis and Weighting Derivation

In the original structural derivation (Wuyts et al., 2000), a pool of continuous acoustic, aerodynamic, and vocal performance parameters (including fundamental frequency, habitual pitch, jitter, shimmer, noise-to-harmonic ratio, MPT, vital capacity, phonation quotient, and vocal range boundaries) were entered into a stepwise discriminant analysis. The categorical grouping variable was the perceptual overall Grade of dysphonia from the GRBAS scale ($G_0, G_1, G_2, G_3$).

The statistical algorithm selected four variables that accounted for unique, non-redundant variance in predicting perceptual voice quality, while minimizing collinearity:

• MPT: Aerodynamic capacity and glottal efficiency ($F\text{-to-remove} = 42.1$).
• $F_0\text{-high}$: High-frequency vocal fold tensioning capacity ($F\text{-to-remove} = 38.6$).
• $I\text{-low}$: Low-intensity glottal dynamic control ($F\text{-to-remove} = 29.4$).
• Jitter %: Micro-temporal oscillatory regularity ($F\text{-to-remove} = 51.8$).

The resulting canonical discriminant function yielded the definitive mathematical algorithm:

$$\text{DSI} = 0.13 \times \text{MPT} + 0.0053 \times F_0\text{-high} – 0.26 \times I\text{-low} – 1.18 \times \text{Jitter (%)} + 12.4$$

Within this model, the constant ($+12.4$) centers the scale such that a completely normal, euphonic voice corresponds to approximately $+5$, whereas progressively severe dysphonia falls toward $-5$ or lower. The multivariate structural equation accounts for roughly $71%$ of the total variance ($R^2 = .71$) in perceptual voice grade categorization.

10. Instrument / Measurement Tool

The Dysphonia Severity Index is an instrumental assessment protocol that combines acoustic phonetography and sustained aerodynamic voice recordings.

• Instrument Type: Objective acoustic and aerodynamic clinical measurement index.
• Administration Format: Direct physiological performance testing under standardized acoustic conditions.
• Component Count: 4 objective physical/acoustic variables.
• Instrumentation Requirements:
  • Calibrated acoustic analysis software (e.g., Computerized Speech Lab / Multidimensional Voice Program [MDVP] by Pentax Medical, or Praat acoustic software).
  • Type 1 or Type 2 sound-level meter or calibrated microphone positioned at a fixed distance of $30\text{ cm}$ ($45^circ$ off-axis to avoid airflow blast artifacts).
  • Sound-treated recording booth with ambient noise level $< 40text{ dBA}$.
  • Stopwatch or computerized digital time-tracking system for phonation duration.
• Operational Measurement Protocol:
  • Maximum Phonation Time (MPT): The patient takes a maximal inspiration and sustains the vowel /a/ at a comfortable pitch and conversational loudness for as long as possible. The task is repeated three times; the longest duration in seconds is retained.
  • Highest Fundamental Frequency ($F_0\text{-high}$): The patient performs upward pitch glissandos or stepped vocal scales on the vowel /a/ or /i/, coached to reach their absolute physiological ceiling. The highest frequency reliably produced in Hertz (Hz) is identified and extracted via acoustic software.
  • Lowest Vocal Intensity ($I\text{-low}$): The patient phonates the vowel /a/ at their habitual comfortable pitch, gradually decreasing intensity to the softest possible phonation without lapsing into whisper or aphonic airflow. The minimum stable intensity in dB SPL (C-weighting or A-weighting, standardized at $30\text{ cm}$) is recorded.
  • Jitter (%): The patient sustains the vowel /a/ at a comfortable pitch and volume for at least 3 seconds. The mid-stable 1-second segment is analyzed using acoustic perturbation software to extract the local jitter percentage.
• Scoring and Normative Interpretation:
  • Formula: $\text{DSI} = (0.13 \times \text{MPT}) + (0.0053 \times F_0\text{-high}) – (0.26 \times I\text{-low}) – (1.18 \times \text{Jitter}) + 12.4$
  • DSI $ge +5.0$: Superior or exceptionally resilient vocal capability (often observed in professional vocalists).
  • DSI $+1.6$ to $+4.9$: Perceptually normal vocal quality (euphonia). Most clinical cut-offs identify scores above $+1.6$ as the normal physiological range.
  • DSI $+1.5$ to $-1.4$: Mild dysphonia ($G_1$ on the GRBAS scale).
  • DSI $-1.5$ to $-4.0$: Moderate dysphonia ($G_2$ on the GRBAS scale).
  • DSI $< -4.0$: Severe dysphonia to aphonia ($G_3$ on the GRBAS scale).

11. Permissions & Fee and Test Year

The Dysphonia Severity Index was established and published in 2000 (preliminary clinical validation presented in 1996 and published in peer-reviewed form in the Journal of Speech, Language, and Hearing Research in February 2000). The mathematical algorithm itself is published within the public academic domain and may be implemented freely by clinicians, clinical researchers, and software engineers without payment of individual licensing royalties or author per-use fees.

However, the acoustic acquisition systems, voice assessment hardware, and commercial software packages that automate DSI computation (such as Pentax Medical’s Multi-Dimensional Voice Program [MDVP], LingWAVES by WEVOSYS, or specialized medical phonetographs) are proprietary products that require software purchase and hardware licensing. Clinicians and researchers operating without proprietary hardware can implement the published formula using free, open-source acoustic software such as Praat, utilizing custom measurement scripts developed and cross-validated across phonetics research literature.

12. References

De Bodt, M. S., Wuyts, F. L., Van de Heyning, P. H., & Croux, C. (2000). Test-retest study of the GRBAS scale: Influence of experience and expertise on perceptual evaluation of voice quality. Journal of Voice, 14(1), 74–80. https://doi.org/10.1016/S0892-1997(00)80096-4

Hakkesteegt, M. M., Brocaar, M. P., Wieringa, M. H., & Feenstra, L. (2006). The relationship between perceptual evaluation and objective multiparametric evaluation of dysphonia severity. Journal of Voice, 20(3), 353–363. https://doi.org/10.1016/j.jvoice.2005.05.004

Hakkesteegt, M. M., Wieringa, M. H., Brocaar, M. P., & Feenstra, L. (2008). The Dysphonia Severity Index: Variability and critical difference. Clinical Otolaryngology, 33(4), 338–343. https://doi.org/10.1111/j.1749-4486.2008.01755.x

Hirano, M. (1981). Clinical examination of voice. Springer-Verlag. https://doi.org/10.1007/978-3-7091-8625-1

Kempster, G. B., Gerratt, B. R., Verdolini Abbott, K., Barkmeier-Kraemer, J., & Hillman, R. E. (2009). Consensus Auditory-Perceptual Evaluation of Voice: Development of a standardized clinical protocol. American Journal of Speech-Language Pathology, 18(2), 124–132. https://doi.org/10.1044/1058-0360(2008/08-0017)

Maryn, Y., De Bodt, M., & Roy, N. (2010). The Acoustic Voice Quality Index: Toward a robust combination of acoustic voice quality measurements. Annals of Otology, Rhinology & Laryngology, 119(6), 424–430. https://doi.org/10.1177/000348941011900611

van den Berg, J. (1958). Myoelastic-aerodynamic theory of voice production. Journal of Speech and Hearing Research, 1(3), 227–244. https://doi.org/10.1044/jshr.0103.227

Wheeler, K. M., Collins, S. P., & Sapienza, C. M. (2006). The relationship between VHI scores and aerodynamic and acoustic measures of voice. Journal of Voice, 20(2), 236–244. https://doi.org/10.1016/j.jvoice.2005.02.003

Wuyts, F. L., De Bodt, M. S., Molenberghs, G., & Van de Heyning, P. H. (2000). The Dysphonia Severity Index: An objective measure of vocal quality based on a multiparameter approach. Journal of Speech, Language, and Hearing Research, 43(3), 796–809. https://doi.org/10.1044/jslhr.4303.796

13. Items of the Scale