Objective Assessment of Pediatric Voice Disorders

With the Acoustic Voice Quality Index

*Victoria Reynolds, *Ali Buckland, *Jean Bailey, *Jodi Lipscombe, †Elizabeth Nathan, ‡,§Shyan Vijayasekaran,

kRona Kelly, {Youri Maryn, and k,#Noel French, *yzxk#Western Australia, Australia and {Bruges, Belgium

Summary: Objectives/Hypothesis. Instrumental measures of voice allow practitioners to assess the severity of

AccepFrom

WesternMemoriaogy, HeaAustraliaMargarettre, Westand AudCare UniAddre

Speech P6840, AuJourna0892-1� 201doi:10

voice disorders and objectively measure treatment outcomes. Instrumental measures should be calculated on both sus-tained vowel and connected speech samples to ensure ecological validity. However, there is a lack of appropriate, val-idated acoustic measurements for use in the pediatric population. The Acoustic Voice Quality Index (AVQI) isa multivariate acoustic measure of dysphonia that has been found to be reliable, valid, and have diagnostic accuracyand response to change in an adult population. This study aimed to evaluate the AVQI in a pediatric population.StudyDesign. This study was a prospective observational study of a sample of dysphonic and normophonic children.Methods. Sixty-seven preterm participants (born at less than 25 weeks gestation) aged between 6 and 15 years wererecruited. Participantswere excluded because of either inability to complywith task requirements or other speech-relatedfactors that affected acousticmeasurement. Forty normophonic term-born participants aged between 5 and 15 years werealso recruited. AVQI analysis was conducted on a prolonged vowel sample and a sample of continuous speech.Results. The AVQI was found to have diagnostic accuracy and specificity in this population of children with and with-out dysphonia. It was moderately correlated with ratings of severity on the GRBAS (overall grade of hoarseness (G),roughness (R), breathiness (B), aesthenicity (A), and strain (S)), a subjective rating scale. The threshold for pathologyof this sample of 3.46 showed strong sensitivity, specificity, and accuracy, with good-to-excellent likelihood ratios.Conclusions. This study found that the AVQI has diagnostic accuracy in a pediatric population, suggesting that it is anappropriate assessment tool to determine the presence and severity of pediatric voice disorders.Key Words: Pediatric–Dysphonia–Multivariate acoustic measurement–Continuous speech–Voice disorders.

INTRODUCTION

An individual is considered to have a dysphonic voice whentheir vocal quality differs perceptually from the norm, orwhen it no longer allows the individual to fulfill social, aca-demic, or employment tasks.1 Dysphonia can be caused by or-ganic and/or functional factors2 and individuals can experiencesignificant functional and participatory limitations associatedwith their voice-related difficulties.3 There are a number oftools available to the clinician for use in the assessment andmeasurement of voice disorders, which are reliable and valid,including perceptual evaluation, quality-of-life measures, andacoustic measurements.4,5 Acoustic measures are consideredthe most objective measure of voice quality. However, the opti-mal acoustic measure has yet to be determined. Issues with re-liability and validity, and confounding factors, have hamperedefforts to develop a single measure that reliably indicates thepresence and/or severity of a voice disorder.4,5

ted for publication February 22, 2012.the *Department of Speech Pathology, Princess Margaret Hospital, Subiaco,Australia, Australia; yWomen and Infants Research Foundation, King Edwardl Hospital, Subiaco, Western Australia, Australia; zDepartment of Otolaryngol-d and Neck Surgery, School of Surgery, University of Western Australia, Western, Australia; xDepartment of Otolaryngology, Head and Neck Surgery, PrincessHospital, Subiaco, Western Australia, Australia; kState Child Development Cen-Perth, Western Australia, Australia; {Department of Speech Language Pathologyiology, Sint-Jan General Hospital, Bruges, Belgium; and the #Neonatal Clinicalt, King Edward Memorial Hospital, Subiaco, Western Australia, Australia.ss correspondence and reprint requests to Victoria Reynolds, Department ofathology, Princess Margaret Hospital, GPO Box D184, Perth, Western Australiastralia. E-mail: victoria.reynolds@health.wa.gov.aul of Voice, Vol. 26, No. 5, pp. 672.e1-672.e7997/$36.002 The Voice Foundation.1016/j.jvoice.2012.02.002

The Acoustic Voice Quality Index (AVQI) is an algorithm,calculated from acoustic parameters of a voice sample consist-ing of a sustained vowel and connected speech.6 As ameasure ofdysphonia severity, it has been found to have diagnostic accu-racy, reliability, and validity for the adult population.7 Sixacoustic characteristics of a speech signal are used to calculatetheAVQI, further details ofwhich are reported byMaryn et al.6,7

Although other acoustic measures have also demonstrated reli-ability and validity including, for example, the DysphoniaSeverity Index,8 the AVQI may be considered more appropriatefor use in the pediatric population for several reasons.First, correlations of perceptual attributes with previous

acoustic measures have been mixed. This is particularly prob-lematic as perceptual measures are the primary manner inwhich a voice disorder is judged by an individual’s community.9

It has been suggested that acoustic evaluations should beclosely correlated with perceptual judgments of voice,10 tohave greater clinical utility. The AVQI has been found to corre-late well with subjective evaluations of dysphonia severity.10

Secondly, previous objective analysis of voice has tended tobe limited to sustained vowels.4 Inclusion of a connectedspeech sample increases the ecological validity of acousticanalysis, as the acoustic qualities of sustained vowel productionand connected speech vary.11 Although sustained vowels lackvariation of the speech signal because of prosodic and voicingfactors and allow for easier standardization of the sample, con-nected speech is a more natural speaking behavior, and is theenvironment in which perceptual judgments about the accept-ability of vocal quality are made.12,13 Therefore, inclusion ofboth sustained vowels and connected speech in the acousticanalysis of voice is highly desirable and provides a measure

Victoria Reynolds, et al AVQI in Pediatric Voice Disorders 672.e2

of dysphonia severity on a sample more closely related to theindividual’s conversational speaking voice.

Although some pediatric participants were included in thestandardization sample of the AVQI, the measure has yet to bevalidated on an exclusively pediatric sample.6 This study soughtto investigate the diagnostic accuracy of the AVQI in a sample ofparticipants with and without dysphonia in a pediatric context.This study formed part of a larger investigation into voice qual-ity in extremely preterm children to be published in future.

In summary, this study aims to investigate the diagnostic ac-curacy and specificity of the AVQI in a pediatric sample ofvoice-disordered and nonvoice-disordered participants. TheAVQI has previously been found to have these properties, andstrong ecological validity, in adult populations. Should thismeasure be found to retain these qualities in a pediatric popula-tion, it would be a useful research and clinical tool for cliniciansworking in the area of pediatric voice disorder, as a measure ofdysphonia severity.

METHOD

Participants

Seventy participants between 6 and 15 years of age were re-cruited from the follow-up program for preterm infants runby King Edward Memorial Hospital, which provides neonatalintensive care and follow-up services to all extremely pretermbirths inWestern Australia. Eligible participants were survivorsborn at less than 25 weeks gestation, all of whom required en-dotracheal intubation immediately after birth. Forty term-bornparticipants were recruited from the Orthopedic Fracture Clinicof Princess Margaret Hospital to provide a range of acousticmeasures in non voice-disordered children. Exclusion criteriafor the term children included no recent history of factorsknown to influence voice (eg, recent or recurrent upper respira-tory tract infection, recent intubation or previous surgery, or in-jury to the larynx). All voice samples were obtained with priorwritten informed consent from the parent or caregiver guardian.

Voice samples

Voice analysis was conducted on a prolonged vowel sample anda sample of continuous speech. Participants were asked to sus-tain the vowel /a/ for as long as possible and to read the story‘‘Level 1—Bird’’ from the Neale Analysis of Reading Abilityusing their everyday speaking voice.14 The samples were re-corded on a headset-mounted Røde electret microphone anddigitized, at a sampling frequency of 44.1 kHz with 16 bits ofresolution, using a Zoom Corporation H2 Handy Recorder.15,16

The microphone was positioned between 4 and 5 cm from eachparticipant’s mouth. Samples were saved in .wav format.

The speech samples were edited using the Praat computerprogram.17 Thevowel sample analyzed included only themiddle3 seconds (or the middle 2 seconds where participants were un-able to sustain the vowel for longer than 3 seconds). The readingsample analyzed included the middle two sentences. The voicesamples were edited in accordance with the procedure describedby Maryn et al9 to calculate the AVQI for each participant usingthe SpeechTools and Praat programs.17,18 Of the six parameters

used to calculate the AVQI, one (smoothed cepstral peak prom-inence) was calculated using the SpeechTools program and theremainder were calculated using Praat, in exact accordancewith the procedure described by Maryn et al.6

Overall dysphonia ratings

As discussed, perceptual data should support objective acousticanalysis. The GRBAS is a four-point scale that rates five aspectsof voice quality (overall grade of hoarseness (G), roughness(R), breathiness (B), aesthenicity (A), and strain (S))19 and isa widely used tool in research and clinical practice, which pro-vides useful information about the nature of voice disorders toassist in planning intervention.4,13 For the purposes of statisticalanalysis, the arithmetic mean of the sum of the ratings for eachparameter was used.

The mean G was considered the most appropriate measure ofdysphonia to assess the diagnostic accuracy of the AVQI as bothmethods are used to determine the presence and severity of dys-phonia. Neither the GRBAS nor the AVQI provide informationregarding underlying laryngeal pathology.

Two speech pathologists with postgraduate experience inclinical voice assessment were involved in data collection andrating. One joint rating session, of 1 day in length, was heldat the commencement of the data collection phase of the projectto calibrate the subjective ratings on the GRBAS by attemptingto establish an external standard based on six samples that werecollected. Inter- and intrarater reliability was calculated ona random sample of 10% of the raw voice samples.

Statistical analysis

Demographic and voice characteristics were summarized usingmedians, interquartile ranges and ranges, and categorical datausing frequency distributions. Univariate analysis was conduct-ed using theMann-Whitney test and the Chi-square test for con-tinuous and categorical data, respectively. The inter-rater andintrarater reliability were assessed using the intraclass correla-tion coefficient (ICC) and strength of agreement for ordinaldata using weighted kappa statistics. To incorporate the orderednature of the outcomes, weights were assigned to each cell ina two-way contingency table and then a kappa value was calcu-lated. Weightings were used to measure the importance of dis-agreement; the farther apart their classifications were, theworse the agreement. Linear weightings were assigned to dis-agreements using the following algorithm: 1� ji� jj/(k� 1),where i and j index the rows and columns of the ratings by thetwo raters, and k is the maximum number of possible ratings.The diagnostic precision of the AVQI for discriminating be-tween normal and dysphonic voices was assessed using receiveroperator curve (ROC) analysis. Data-driven methods were usedto calculate the point closest to (0, 1) on the curve and derive thebest cut-off score. The methods applied were minimizing[(1� sensitivity)2 + (1� specificity)2] and maximizing [sensi-tivity + specificity� 1]. Sensitivities, specificities, and likeli-hood ratios were calculated to measure the diagnosticcapabilities of theAVQI. Statistical analyseswere conducted us-ing PASW v18.0 statistical software (SPSS Inc, Chicago, IL)

Journal of Voice, Vol. 26, No. 5, 2012672.e3

and SAS Version 9.2 of the SAS System for Windows (Copy-right � 2002–2010 SAS Institute Inc., Cary, NC, USA).

RESULTS

Population characteristics

One hundred thirty extremely preterm survivors born at lessthan 25 weeks gestation were eligible for study participation.Seventy subjects agreed to participate, with three excluded be-cause of either inability to comply with task requirements orother speech-related factors that affected the acoustic measure.The remaining 67 recruits combined with the 40 term-born re-cruits gave a total of 107 participants included in the analysis.

Demographic and voice characteristics. The demo-graphic and voice characteristics of the participants can beseen in Table 1. The preterm group included 35 females and32 males with a median age of 11 years (range 6–15 years),and the term group, 17 females and 23 males with a medianage of 9 years (range 6–15 years). Fifty-nine percent (39 of67) of the children born preterm were classified withmoderate-severe dysphonia and one-term born child was classi-fied with moderate dysphonia based on the GRBAS rating.

Agreement between AVQI and GRBAS methods of

dysphonia measurement. The overall agreement betweenperceptual and acoustic classifications of speech in detectingdysphonic from normal voice patterns was 80.4% and the kappastatistic was 0.489 indicating moderate agreement beyondthat expected by chance (prevalence index¼ 0.49; bias in-dex¼ 0.065).20 The 95% confidence interval (CI) for kappa

TABLE 1.

Demographic and Voice Characteristics

Characteristic

Term Born

Median (Q1

Age 9 (8–11

Male (N%) 23

Mean G* 0 (0–0.

G score severity rating (N%)*

Normal 22

Mild 16

Moderate 1

Severe —

AVQI* 3.0 (2.5–

Cepstral peak prominence* 7.5 (7.0–

Harmonics-to-noise ratio (dB)* 18.4 (16.4

Shimmer local (%)* 3.7 (2.8–

Shimmer local (dB)* 0.4 (0.3–

Slope of LTAS (dB) �16.5 (�18

�2

Tilt of trendline through LTAS (dB)** �10.6 (�11

�1

Abbreviations: AVQI, acoustic voice quality index; LTAS, long-term average spec

*P value <0.001.

**P value <0.05.

was 0.299–0.679. A plot of theAVQI scores comparedwith dys-phonia severity ratings on the GRBAS can be seen in Figure 1.

Inter- and intrarater reliability. The average ICC betweenraters for the mean G was 0.914 (95% CI: 0.698–0.977).The overall severity (G-score) agreement between raters was85.0% and kappa 0.571 indicatingmoderate agreement. The av-erage ICC for the mean G was 0.846 (95% CI: 0.457–0.958) in-dicating a moderate level of consistency within raters. Overallseverity rating (G-score) agreement within raters was 84.9%,kappa 0.567, indicating moderate agreement. These results in-dicate that the G score was shown to be a reliable measure forthis cohort.

Applicability of AVQI to pediatric population. Crite-rion-related concurrent validity. The correlation betweenmean G scores and AVQI scores was 0.794 indicating a positivelinear relationship, rs¼ 0.794, P < .001, confirming strong con-current validity. Further, the linear relationship illustrates thathigher AVQI scores are correlated with more severe clinical rat-ings of the disordered voices. The proportional relationship be-tween the AVQI and the GRBAS can be seen in Figure 2.

Diagnostic precision—ROC analysis. The area under the re-ceiver operator curve (AUC) using mean G as the state variable(mean G¼ 0 indicating normal voice quality, mean G > 0 indi-cating disrupted voice quality) and AVQI scores as the test vari-able was 0.876 (95% CI: 0.810–0.942, P < 0.001) indicatinggood discriminatory potential to differentiate between normaland dysphonic voices.Application of an AVQI cut-off score of 2.95 to discriminate

between normal and dysphonic voices was recommended by

(N¼ 40) Extremely Preterm (N¼ 67)

–Q3; Min-Max) Median (Q1–Q3; Min-Max)

; 6–15) 11 (8–13; 6–15)

58% 32 48%

4; 0–1.8) 1.4 (0.6–1.8; 0–2.4)

56% 4 6%

41% 24 36%

3% 36 54%

— 3 5%

3.4; 1.9–5.7) 5.0 (4.1–6.2; 2.3–7.7)

8.1; 5.5–9.8) 5.5 (4.6–6.4; 1.8–9.0)

–19.4; 13.5–22.5) 13.8 (11.2–16.3; 3.0–21.2)

4.3; 1.6–6.9) 5.9 (4.0–9.1; 2.1–19.0)

0.5; 0.2–0.8) 0.6 (0.5–0.9; 0.2–1.7)

.9 to (�11.8);

4.8 to 18.9)

�17.0 (�19.4 to (�14.0);

�27.4 to (�6.3))

.3 to (�10.0);

2.1 to 11.7)

�9.9 (�10.7 to (�9.3);

�12.3 to (�4.7))

trogram.

FIGURE 1. AVQI scores and dysphonia severity ratings.

Victoria Reynolds, et al AVQI in Pediatric Voice Disorders 672.e4

Maryn et al,6 and gave the following diagnostic accuracy re-sults: sensitivity 89%, specificity 62%, and accuracy 82%with LR + 2.34 and LR� 0.18.

Both data-driven techniques used to choose the best AVQIcut-off score to discriminate dysphonic voices from normal voi-ces in this sample indicated the same score. The best AVQIcut-off scorewas 3.46, giving the following diagnostic accuracymeasures: sensitivity 82%, specificity 92%, and accuracy 84%,with LR + 10.25 and LR� 0.20 (Figure 3).

FIGURE 2. Scatterplot and linear regression line indicating the relationshi

term born.

DISCUSSION

Applicability of AVQI to pediatric voice

This study provides support for the use of the AVQI as an objec-tive measure of voice in a pediatric population, using a cohort ofextremely preterm infants with a very high prevalence of dys-phonia (data to be reported separately) and a term-born, normo-phonic cohort. The AVQI was found to be a valid measure ofdysphonia severity in this cohort. The AVQI values were

p between the AVQI and the GRBAS. Cross: extremely preterm. Circle:

FIGURE 3. ROC graph illustrating the diagnostic accuracy of the

AVQI. The dotted line represents chance-level determination between

normal and dysphonic voices.

Journal of Voice, Vol. 26, No. 5, 2012672.e5

correlated with GRBAS scores, indicating a strong, positive re-lationship between the two measures, providing evidence thatthe AVQI is an accurate indicator of dysphonia severity. Todate, the AVQI is the only instrumental measure of voice that in-cludes both a sustained vowel and a connected speech sample. Ithas been argued that both samples are essential for ecologicalvalidity, to capture the different voice characteristics of sustainedvowels versus connected speech.11 The threshold for pathologyof this sample of 3.46 showed strong sensitivity, specificity, andaccuracy, with good-to-excellent likelihood ratios.

However, there were participants who were rated as disor-dered on the GRBAS, yet had an AVQI score below the thresh-old for pathology. Additionally, there were cases where theAVQI score was above the threshold for pathology but theGRBAS was rated as normal. In clinical settings, in such cases,other measures such as quality-of-life questionnaires could beexamined, or ratings could be cross checked by another clini-cian, to determine whether or not the voice warrants clinical at-tention. Voice is a subjective entity, and patients approachclinical services when their voices sound different to the un-trained ear. In cases where there is disagreement between theAVQI and the GRBAS, it may be useful to have the clinicianrate the voice blinded, then adopting the second GRBAS ratingwould be appropriate. Although in such cases, careful examina-tion of the parameters of the GRBAS would be required to pin-point the precise nature of vocal quality and attempt to identifywhy the acoustic parameters fell within the average range.

The presence of some disagreement between the AVQI andthe GRBAS is not unexpected, as no study has yet found perfectcorrelation between the two measures, or indeed any subjectiveand objective measures of voice, to our knowledge. These re-sults lend support to the proposition that any voice assessmentshould encompass objective, subjective, and quality-of-lifemeasures.21

Despite the fact that the AVQI has been validated for an adultpopulation, separate validation studies are necessary for adultand pediatric voice. Structural differences in adult and child lar-ynxes lead to different acoustic characteristics of the respectivevoices.22 The larynx descends to the adult position between C4and C7 vertebrae by 6 years of age, whichwas the lower age limitfor participants in this study. Cartilaginous structures are softerand therefore not capable of supporting sustained and loudvoicing at the same level as adults, hence the differences in mea-sures such as maximum phonation time on sustained vowels be-tween adults and children.12,23 One study23 found that childrenhave a mean harmonics-to-noise ratio of 15.39 dB (standard de-viation [SD] 3.55 dB), whereas healthy adult speakers should beable to produce a value of 20 dB or higher. Children also havea higher shimmer value than adults, 6.28% (SD 4.33%) versus4%, but that jitter values fell within the normal adult range.In support of the differences between the adult and pediatric

voice, Maryn et al6 found that the threshold for pathology of theAVQI was 2.95, whereas, this study found a cut-off of 3.46 tohave better specificity and sensitivity for this sample. Thiscould be due to either different vocal qualities in children versusadults, differences in sampling techniques, or difference in thethreshold for pathology or a combination of all such factors. InMaryn et al’s studies, voice samples were collected in sound-treated rooms, but these were not available for this study. Allother sampling procedures (eg, regarding microphone place-ment) were similar and recording equipment was selected tominimize distortion and maximize sound quality.This study also provided information about the utility of the

AVQI in everyday clinical contexts. Collection techniques werenoninvasive and involve minimal discomfort. The microphoneheadset was generally well tolerated by participants. TheAVQI was found to be easy to collect and analyze. Calculationof the AVQImeasure, on completion of data collection, took ap-proximately 5 minutes and could potentially be done online, toprovide clients and their caregivers with immediate feedbackabout the severity of their voice disorder. Several childrenwith mild intellectual disabilities were included in this studyand were able to comply with the task instructions. The con-nected speech sample was obtained from the Neale Analysisof Reading Ability,14 requiring a reading age equivalent tothe first year of compulsory schooling in Western Australia,and will therefore be useful for children aged 6 years and older.

Limitations

Voice samples should preferably be collected in sound-treatedrooms as this allows speech to be recorded free of acoustic con-tamination. However, aside from this factor, the recordingequipment and procedures were standardized. Therefore, theintegrity of the voice recordings should be identical betweenparticipants.The methodology of the study does preclude children who

have not yet achieved basic literacy skills. Should the AVQIbe required for use in children of preschool age, such as in anearly intervention clinic, a method for eliciting a standardizedsample of connected speech that is not biased for literacy skillsshould be investigated.

Victoria Reynolds, et al AVQI in Pediatric Voice Disorders 672.e6

The response rate of the study was 54%. However, detailedinformation pertaining to the voices of the preterm childrenwho did not respond is not available. It is possible that childrenwith dysphonia are overrepresented in the cohort and confiningdysphonic subjects to be only those born extremely pretermmay not be representative of the wider pediatric population ofvoice disorders.

In addition, there is a further aspect of voice evaluation (ie,laryngoscopy) that was not considered in this study becauseof its invasive nature. To determine the precise nature of the un-derlying laryngeal pathology, endoscopic evaluation of the lar-ynx and the vocal folds, is considered to be a prerequisite forspeech pathology intervention.23 An endoscopic evaluation ofeach subject’s larynx would provide information in relation tothe precise anatomical and functional status of vocal output.Perceptual qualities could be linked to structural and functionalfactors and the presence of compensatory voicing strategiescould be detected. However, this is not essential as laryngeal pa-thology does not necessarily correlate with dysphonia severity.In psychogenic voice disorders, for example, dysphonia can besevere in the absence of vocal pathology.

Despite these limitations, the results of the data allow someconclusions to be drawn about the applicability of AVQI to pe-diatric voice.

Future directions

Confirmation of these results through a replication study usinga larger sample size and more diverse pediatric population isneeded to provide further evidence as to the applicability ofthe AVQI to pediatric voice.

The threshold for pathology for the AVQI in this study wasbased on data provided by Maryn (personal communication,21 May 2011) from a pilot study in a small sample of childrenand young people. Maryn et al found that the threshold for pa-thology originally calculated in the adult studies was valid forthat sample. However, this study found that a slightly higherthreshold, 3.46, was more accurate in this pediatric sample.Therefore, a future direction for research would be to investi-gate the diagnostic threshold of the AVQI in a larger andmore diverse sample of children.

Measuring change over time is essential to determine treat-ment efficacy and effectiveness; thus any assessment tool mustbe sensitive to change to have application in a clinical context.Maryn et al determined that the AVQI has specificity to changefollowing intervention in adults with voice disorders.7 As theAVQI has diagnostic validity and sensitivity to change in an adultpopulation, it can be hypothesized that, in this pediatric samplewhere diagnostic accuracy has been found, that the measure willbe similarly sensitive to change. This should be investigated witha voice-disordered, pediatric population in a treatment study.

In addition, some of the preterm cohort have undergone stro-boscopy as part of their ongoing medical management. A retro-spective chart review may provide information that allows theirstroboscopy results to be compared with their AVQI scores, todetermine the correlation between laryngeal pathology anddysphonia severity, as measured by the AVQI.

CONCLUSIONS

This is the first study to systematically investigate an acousticmeasure containing a connected speech sample in pediatric pa-tients both with and without dysphonia that measures dyspho-nia severity. The AVQI is an instrumental measure that iseasy and quick to administer and analyze. It has been foundto have diagnostic accuracy in this study of its applicability topediatric voice. It is correlated with scores on the GRBAS,a subjective assessment of the perceptual qualities of disorderedvoices. The AVQI is potentially a useful tool for the assessmentand diagnosis of pediatric voice disorders. Further study isneeded to replicate these results and to determine its sensitivityto change in pediatric voice.

Acknowledgments

The authors acknowledge Dr Suzanne Meldrum of the Schoolof Pediatrics and Child Health of the University of WesternAustralia for her valuable assistance with editing and articlepreparation, and Mr Colin Whitewood and the Orthopedic De-partment of Princess Margaret Hospital for Children, for theirassistance with the recruitment of comparison group partici-pants. The authors also acknowledge research funding fromthe Women and Infants Research Foundation.

REFERENCES1. Aronson A. Clinical Voice Disorders: An Interdisciplinary Approach. New

York, NY: Thieme; 1990.

2. Wilson DK. Voice Problems of Children. 3rd ed. Baltimore, MD: Williams

& Wilkins; 1987.

3. Ma EPM, Yiu EML, Abbott KV. Application of the ICF in voice disorders.

Semin Speech Lang. 2007;28:343–350.

4. Carding PN, Wilson JA, MacKenzie K, Deary IJ. Measuring voice out-

comes: state of the science review. J Laryngol Otol. 2009;123:823–829.

5. Oates J. Auditory-perceptual evaluation of disordered voice quality. Folia

Phoniatr Logop. 2009;61:49–56.

6. Maryn Y, De Bodt M, Roy N. The Acoustic Voice Quality Index: toward

improved treatment outcomes assessment in voice disorders. J Commun

Disord. 2010;43:161–174.

7. Maryn Y, Corthals P, Van Cauwenberge P, Roy N, De Bodt M. Toward im-

proved ecological validity in the acoustic measurement of overall voice

quality: combining continuous speech and sustained vowels. J Voice.

2010;24:540–555.

8. Hakkesteegt M, Wieringa M, Brocaar M, Mulder PG, Feenstra L. The in-

terobserver and test-retest variability of the dysphonia severity index. Folia

Phoniatr Logop. 2008;60:86–90.

9. Maryn Y, Roy N, De Bodt M, Van Cauwenberge P, Corthals P. Acoustic

measurement of overall voice quality: a meta-analysis. J Acoust Soc Am.

2009;126:2619–2634.

10. Oates J. The evidence base for the management of individuals with

voice disorders. In: Reilly S, Douglas J, Oates J, eds. Evidence Based

Practice in Speech Pathology. London, UK: Whurr Publishers; 2004:

110–139.

11. Wolfe V, Cornell R, Fitch J. Sentence/vowel correlation in the evaluation of

dysphonia. J Voice. 1995;9:297–303.

12. Shewell C. Voice Work: Art and Science in Changing Voices. 1st ed. Chi-

chester, UK: John Wiley & Sons; 2009.

13. Carding P, Carlson E, Epstein R, Mathieson L, Shewell C. Formal percep-

tual evaluation of voice quality in the United Kingdom. Logoped Phoniatr

Vocol. 2000;25:133–138.

14. Neale MD. The Neale Analysis of Reading Ability. 3rd ed. Melbourne,

Australia: Australian Council for Educational Research; 1999.

15. Rode Microphones. Sydney, Australia.

Journal of Voice, Vol. 26, No. 5, 2012672.e7

16. Handy Recorder H2. Tokyo: Zoom Corporation.

17. Boersma P, Weenink D. Praat: Doing Phonetics by Computer. 5.3.08 ed.

Amsterdam, The Netherlands: Institute of Phonetic Sciences; 2011.

18. Hillenbrand J. SpeechTools. 1.65 ed. Kalamazoo, MI: Author; 2008.

19. Hirano M. Psycho-acoustic evaluation of voice. In: Arnold GE, Winckle F,

Wyke BD, eds. Disorders of Human Communication 5, Clinical Examina-

tion of Voice. New York, NY: Springer-Verlang; 1981:81–84.

20. Landis J, Koch G. The measurement of observer agreement for categorical

data. Biometrics. 1977;1977:159–174.

21. Carding P.Evaluating Voice Therapy:Measuring the Effectiveness of Treat-

ment. 1st ed. London, UK: Whurr Publishers; 2000.

22. Andrews ML. Manual of Voice Treatment: Pediatrics Through Geriatrics.

San Diego, CA: Singular Publishing Group; 1999.

23. Hartnick CJ, Boseley ME. Pediatric Voice Disorders. 1st ed. San Diego,

CA: Plural Publishing; 2008.