International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
Contents lists available at ScienceDirect
International Journal of Pediatric Otorhinolaryngology journal homepage: www.elsevier.com/locate/ijporl
Clinical relevance of speaking voice intensity effects on acoustic jitter and shimmer in children between 5;0 and 9;11 years Meike Brockmann-Bauser *, Denis Beyer, Jo¨rg Edgar Bohlender Department of Phoniatrics and Speech Pathology, Clinic for Otorhinolaryngology, Head and Neck Surgery, University Hospital Zurich, Zurich, Switzerland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 14 August 2014 Received in revised form 17 September 2014 Accepted 19 September 2014 Available online 28 September 2014
Background: Current voice assessment recommendations for dysphonic children comprise instrumental acoustic measurements of the perturbation parameters jitter and shimmer. In healthy adults and children changes in speaking voice sound pressure level (voice SPL) have significant confounding effects on both parameters. In adults these effects were considerably reduced in phonations with controlled voice SPL >80 dBA (10 cm distance). However, it is unclear if these findings apply to children and if children are able to control for their own voice intensity. Objective: This cross-sectional single cohort study investigates voice SPL effects on jitter and shimmer in children between 5;0 and 9;11 years phonating at individually ‘‘medium’’ (modeling ‘‘comfortable’’ loudness of the usual clinical protocol), ‘‘soft’’ and ‘‘loud’’ voice and a prescribed intensity level of ‘‘>80 dBA’’ (10 cm distance, with visual control). Further both their ability to phonate at a prescribed voice intensity level and the effect on SPL related confounding effects were studied. Subjects and methods: A total of 68 healthy children (39 f/29 m) aged 5;0 to 9;11 years were included. All phonated the vowel/a/for 5 s, three times at four defined voice intensity levels (soft/medium/loud/ >80 dBA) each. Jitter (%), shimmer (%) and voice SPL (dBA) were determined using PRAAT. Voice intensity level effects were assessed by descriptive statistics, Analysis of Variance (ANOVA) and Linear Mixed Models (LMM). Results: There were significant differences for jitter and shimmer between all voice tasks (p < .01). Jitter and shimmer were lowest and showed the smallest spread in controlled phonations ‘‘>80 dBA’’. 19 children below 7;0 years could not perform the voice tasks and were excluded from the study. Conclusions: This practical study demonstrated a significant effect of voice loudness and task on jitter and shimmer in children. Since the observed confounding effects were large compared to treatment effects, jitter and shimmer may not be meaningful without adequate control of voice SPL. In phonations at ‘‘>80 dBA’’ (10 cm distance) voice SPL related effects were considerably reduced. However, this assessment protocol was suitable only for children above 7;0 years. Application of this task to future studies of dysphonic children may yield clinically valuable information. ß 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Voice diagnostics Children Voice intensity Jitter Shimmer
1. Introduction According to pediatric voice assessment guidelines a comprehensive clinical voice examination usually includes visual, perceptual, patient based subjective and instrumental acoustic assessment techniques [1,2]. Studies investigating the incidence of
* Corresponding author at: Department for Phoniatrics and Speech Pathology; Clinic for Otorhinolaryngology, Head and Neck Surgery, University Hospital Zurich, Frauenklinikstrasse 24, 8091 Zurich, Switzerland. Tel.: +41 44 255 5830; fax: +41 44 255 4424. E-mail addresses: [email protected] (M. Brockmann-Bauser), [email protected] (D. Beyer), [email protected] (J.E. Bohlender). http://dx.doi.org/10.1016/j.ijporl.2014.09.020 0165-5876/ß 2014 Elsevier Ireland Ltd. All rights reserved.
voice disorders during childhood report a prevalence from 6% to 38% [3]. This wide prevalence range highlights the complexity of pediatric voice diagnostics and the difficulties to determine vocal pathology in children. A voice disorder may compromise the general well-being, communication behavior and the social and academic development of a child [4]. Teachers and parents tend to judge the personality and cognitive abilities of a dysphonic child more negatively than of a normophonic child [5,6]. Therefore an early diagnosis and treatment of any voice disorder is key to avoiding considerable and probably even long lasting negative effects on the child’s life. Up to 20% of children below the age of 10 years may not tolerate invasive assessment procedures such as laryngostroboscopy
2122
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
[1,7–10]. Further in patients with muscle tension dysphonia, one of the most frequent diagnoses in children, laryngostroboscopy has limited validity to determine pathologic voice production patterns [11]. In these cases the clinical diagnosis and treatment decision depends on perceptual, subjective and instrumental acoustic assessment methods. Instrumental acoustic measurements allow an objective and non-invasive assessment of uninfluenced voice function by analyzing the vocal output. For this the patient’s voice is recorded with a microphone, and acoustic analysis usually is done by a computer based software application. Even children of a very young age tolerate this procedure. Therefore instrumental acoustic assessments are considered as easily applicable in pediatric voice care [1]. The present work focuses on the acoustic perturbation parameters jitter and shimmer, indicating the irregularity of human voice pitch (jitter) and intensity (shimmer) respectively from one acoustic wave to the next. The broad clinical application of jitter and shimmer is based on the hypothesis that the acoustic waveform represents the vibratory characteristics of the vocal folds. A variety of studies in children and adults show that vocal pathology and/or dysphonia are associated with increased jitter and shimmer [1,5,12–16]. Also, the onset of mutation in children was accompanied by increased acoustic voice perturbation [17]. It has been suggested, that jitter and shimmer may even indicate subtle changes in the vibratory properties of the vocal folds [16,18,19]. Based on this jitter and shimmer have been characterized as informative and clinically valuable parameters to determine pathology and mutation onset in childrens’ voices [1,13]. However, in adults and children the reliability and usefulness of jitter and shimmer measurements has been questioned for a number of reasons [20–22]: both parameters are significantly influenced by technical confounding factors such as microphone quality, the analysis software type or background noise. The analysis system and program may affect mean shimmer and jitter by factors ranging from 1.2 to 3.1 [23]. Also the type of the analyzed voice signal determines the exactness and clinical usefulness of acoustic measurements. Several authors have argued that perturbation parameters are not meaningful in severely irregular voices, since jitter and shimmer depend on correct recognition of fundamental frequency and voice SPL [20,21,23–25]. However even under adequate measurement conditions both jitter and shimmer vary considerably within healthy adults and children during a day [26,27]. A practical reason for this might be how patients are instructed during acoustic assessments [22]. According to current guidelines, patients are usually asked to phonate at ‘‘comfortable loudness and pitch’’ [1]. Under this assessment protocol, adults have substantial interindividual differences in speaking voice SPL, which significantly influence jitter and shimmer [28–30]. A statistical analysis of data by means of etasquared in phonations at individually ‘‘normal’’ voice loudness showed that 62% of shimmer variance and 24% of jitter variance could be explained by changes in voice SPL. The effect sizes of vowel, gender and fundamental frequency (F0) were considerably smaller ranging from 0% to 4% [28]. Therefore the reliability of jitter and shimmer measurements in adults could be considerably improved with adequate control of voice SPL [22,28]. Children also vary substantially in their speaking voice intensity, when asked to phonate at ‘‘comfortable’’ loudness [31]. Further a study by Glaze et al. showed, that voice intensity differences have significant effects on both jitter and shimmer in children [32]. However it has not been investigated, how these confounding effects may be sufficiently controlled for in clinical practice. Preliminary findings in healthy adults suggest, that the effects due to differences in speaking voice SPL may be considerably reduced, when patients are asked to phonate at a minimum of 80 dBA (measured at 10 cm distance) [30,33].
However these findings may not apply to children, since vocal fold length and microstructure change with age. Currently it is not fully established, how these developmental changes affect the vibratory properties of the vocal folds and hence jitter and shimmer [6]. Also, from a practical point of view, it is unclear if children are able to phonate at a prescribed voice intensity level. Specifically younger children may be unable to control for their own voice SPL, since their conceptual skills and physical abilities are not fully developed [34]. To the best of our knowledge these issues have not been investigated in children to date, and therefore will be key aims of the present study. 1.1. Study aims The aims of the present study in children between 5;0 and 9;11 years were to characterize the effects of voice intensity changes on jitter and shimmer in a variety of voice tasks. Specifically it was tested if children of this age group are able to phonate at a prescribed voice intensity level of ‘‘>80 dBA’’, and if this minimizes confounding effects due to differences in voice SPL. 2. Materials and methods 2.1. Population studied A total of 87 children aged between 5;0 and 9;11 years, 47 girls and 40 boys, were recruited to the present cross-sectional cohort study. Of these, 19 children were not able to perform the voice tasks and were excluded. Please refer for further details about the specific exclusion reasons to Section 3. All children were students of three schools and kindergartens in Zurich, Switzerland. This study was approved by the responsible ethical review board under reference number KEK-ZH-Nr. 2010-0305/2. 2.2. Recruiting process and exclusion criteria Prior to the study information packages were send out to all parents of students between 5;0 and 9;11 years of age. All packages included detailed study information, an informed consent form and a participant questionnaire assessing the eligibility of children for the study. 142 parents returned the informed consent and questionnaires. Oral consent was obtained from all children on the assessment day. Children were excluded from the study if they met one or more of the following criteria: previous formal voice training or therapy; acute infection of the ears, nose and throat or allergic reaction on the recording day; a medical condition or medication that might affect normal voice function; surgery in the torso, head and neck region or intubation for any intervention within the last 18 months; inability to say the vowel/a/for 5 s in three auditively discernible loudness levels (’’soft’’, ‘‘medium’’, ‘‘loud’’ voice) and at a minimum of 80 dBA (measured at 10 cm distance) after a maximum training phase of 5 minutes; perceptual voice pathology, indicated by a mean of 1 in any GRBAS scale characteristic as assessed by the study examiner and a speech pathologist [35].
2.3. Recording and analysis technique All voice recordings were made at the children’s school or kindergarten during break time in a quiet room with ambient noise
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
2123
Table 2 Age distribution of the investigated children group per year (year;months) and gender (f;m).
below 50 dBA. A laptop with the recording software Audacity (Version 1.3.6. beta) and an external soundcard (EMU 0404 USB 2.0 Audio) set at a sampling rate of 48,000 Hz with 16 bits per sample were utilized [36]. An extra computer screen with monitoring facilities to indicate voice SPL in real time was connected to the laptop. The target voice intensity level of 80 dBA was marked with a red line and a smiley. A head-mounted microphone (AKG Acoustics, C444) with 10 cm microphone-mouth distance in an offaxis position was used [24]. Each single phonation was edited into an individual sound file and labeled anonymously using Audacity. Acoustic analysis was conducted with the software PRAAT Version 5.1.03 [37]. Only second 1.0 from voice onset to second 3.0 was analyzed to exclude the increased variability of the voice onset and offset phase. Comparison method was used to calculate calibrated SPL values from the voice recordings [38]. Prior to the experiment, speech weighted noise was recorded with 10 cm distance to sound source at 65 dBA, 75 dBA, 85 dBA and 95 dBA [39]. The difference between the known SPL values of the speech weighted noise and the uncalibrated values as indicated by PRAAT was used to calculate calibrated SPL values (dBA) of all voice recordings.
Age group (years;months)
5;0 to 5;11
6;0 to 6;11
7;0 to 7;11
8;0 to 8;11
9;0 to 9;11
Overall 5;0 to 9;11
Girls (f) Boys (m) Total
8 1 9
9 6 15
10 8 18
4 11 15
10 1 11
39 29 68
[40]. These mean values were used for further analysis. Thereafter, the mean, standard deviation (SD), minimum, maximum and range of all 3 parameters were determined with SPSS [41]. Analysis of Variance (ANOVA) was applied to determine the overall effects of voice intensity levels (soft/medium/loud/>80 dBA) on jitter and shimmer. As described above, in the present study our participants provided repeated phonations at a variety of intensity levels. Since measurements within individuals tend to be more similar than between individuals, Linear Mixed Models (LMM) were used to further investigate the relation between each voice intensity level (soft/medium/loud/>80 dBA) and jitter or shimmer.
2.4. Voice tasks and recording procedure 3. Results During a preparation phase of 10 min maximum, the children were able to practice the voice tasks. First, all participants were asked to ‘‘sustain/a/for 5 s at medium pitch and loudness’’. This task was used to model ‘‘comfortable’’ voice intensity as usually applied in voice clinics. Thereafter they were asked to sustain/a/ for 5 s, ‘‘as softly as possible’’ and then ‘‘as loudly as possible’’. When the children were able to phonate at three recognizable different loudness levels (soft/medium/loud), all were recorded phonating/a/ for 5 s at each level three times in randomized order. The second voice task was to sustain the vowel /a/ for 5 s while controlling for voice intensity under visual feedback. Live monitoring of voice SPL was provided with a second monitor, which was turned toward the children. The target level of 80 dBA was marked with a red line and a smiley. This way every participant could see, if her or his own voice intensity reached the requested intensity level. All children were asked to ‘‘please sustain/a/for 5 s and please be louder than indicated by the red line with the smiley’’. When the children were able to do this, they were recorded three times. 2.5. Main outcome measures Main outcome measures were the instrumental parameters voice SPL (dBA), jitter (%) and shimmer (%) (Table 1). The indices jitter (%) and shimmer (%) were chosen, since both are normalized for an individual’s voice SPL and F0 [37].
A total of 87 children between 5;0 and 9;11 years were originally recruited to the present study. Of these, 19 participants, who were not able to perfom the given voice tasks, had to be excluded. Thus 68 participants, 39 girls and 29 boys with a mean age of 7;6 years, were included for voice analysis. Table 2 summarizes the age and gender distribution of the studied group. Each subject provided 12 phonations, giving a total of 816 phonations for analysis. 3.1. Compliance to voice task As described before, 19 children were unable to perform the voice tasks and had to be excluded from this study. Notably, most of these children were still in kindergarten and below 6 years of age (Fig. 1). The main exclusion reasons were a too short phonation time, the inability to produce different loudness levels or the prescribed level of ‘‘>80 dBA’’. There was a significant difference in voice SPL (dBA) between the four examined intensity levels ‘‘soft’’,’’ medium’’, ‘‘loud’’ voice and ‘‘>80 dBA’’ (ANOVA and LMM: p < 0.01). As shown in Fig. 2 and Table 3 mean voice SPL was lowest in ‘‘soft’’ phonations (69.3 dBA, SD 5.8) and highest in phonations ‘‘>80 dBA’’ (91.4 dBA, SD 3.6). Further the intersubject range between the individually softest and
[(Fig._1)TD$IG]
25
First, the mean of the three repetitions (phonations) per child and task type were determined for the instrumental acoustic parameters voice SPL (dBA), jitter (%) and shimmer (%) using Excel
No. of children
20
2.6. Statistical analysis
15 10 5 0
Table 1 Main outcome parameters with labels as indicated by the software PRAAT.
4;0 - 4;11
5;0 - 5;11
6;0 - 6;11
7;0 - 7;11
8;0 - 8;11
9;0 - 9;11
Age (years;months) Instrumental parameter
Label software PRAAT
Jitter (%) Shimmer (%) Voice SPL (dBA)
Local jitter Local shimmer Mean energy intensity
Included children
Excluded children
Fig. 1. Number of excluded children per age group (years;months). The majority of children below 6;0 years of age was not able to perform the given voice tasks.
[(Fig._2)TD$IG]
2124
[(Fig._3)TD$IG]
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
Fig. 2. Mean voice SPL (dBA) with standard deviation (SD) and outliers per voice task (soft/medium/loud/>80 dBA) for 68 children between 5;0 and 9;11 years. The difference in produced voice SPL between individuals is considerably reduced in phonations with a prescribed intensity level of >80 dBA.
Fig. 3. Mean jitter (%) with standard deviation (SD) per voice task (soft/medium/ loud/>80 dBA). Outliers are marked with a cross. In phonations >80 dBA jitter is lowest and shows the smallest spread.
4. Discussion loudest phonation was with 15.6 dBA clearly smallest in the voice task to phonate at ‘‘>80 dBA’’, as compared to ranges from 27.7 to 31.1 dBA in all other voice tasks (Fig. 2, Table 3). Also the standard deviation (SD), a summary measure for the differences of each observation from the mean, was clearly smallest in phonations at >80 dBA (Table 3). 3.2. Effects of voice intensity levels on jitter and shimmer According to the statistical analysis with ANOVA there were significant effects of loudness level (soft/medium/loud/>80 dBA) on jitter (%) (p > 0.01). Notably, with LMM there was no significant difference between the levels ‘‘loud’’ voice and ‘‘>80 dBA’’ (p = 0.45). Jitter was smallest (0.27) and showed the lowest standard deviation and range (SD: 0.1; range: 0.49) in phonations ‘‘>80 dBA’’ (Fig. 3 and Table 1). With both ANOVA and LMM there were significant differences for shimmer (%) between the four examined intensity levels ‘‘soft’’,’’ medium’’, ‘‘loud’’ voice and ‘‘> 80 dBA’’ (p < 0.01). Analogously to jitter, shimmer was lowest in phonations ‘‘>80 dBA’’. Also for shimmer the spread between measurements was with a SD of 1.68 and range of 6.31% by far smallest in phonations at ‘‘>80 dBA’’, as compared to a SD from 2.34 to 4.43 and range from 10.05% to 17.6% in ‘‘loud’’ and ‘‘soft’’ phonations (Fig. 4 and Table 3).
Changes in voice intensity have significant confounding effects on acoustic jitter and shimmer measurements in children. In addition, the voice task we ask children to perform, significantly affects both acoustic parameters. The present study findings suggest that requesting phonations ‘‘>80 dBA’’ (at 10 cm distance) reduces the substantial confounding effects of differences in voice intensity between individuals. Especially for jitter, there was a considerably lower spread in phonations at a controlled voice SPL. To ensure the reliability of jitter and shimmer measurements, assessment guidelines for children should include a detailed characterization of the voice recording procedure and specify oral instructions. However, to date it is unclear, if by applying this procedure clinically relevant information or differences between individuals would be masked. 4.1. How important were the confounding effects of voice SPL? Especially Fig. 4 and Table 3 show the consequences of differences in habitual voice SPL between children. Under the same clinical protocol to phonate at ‘‘medium’’ voice intensity, the softest individual phonation was 62.7 dBA, whereas the loudest was 93.8 dBA. This wide range covered nearly the total range from the softest phonation at a ‘‘soft’’ level (55.1 dBA) and the loudest phonation with ‘‘loud’’ voice (97.6 dBA). Thus voice intensity control is comparatively small in phonations at ‘‘medium’’ voice
Table 3 Mean values with standard deviation (SD), minimum (min) and maximum (max) plus range for the instrumental parameters voice intensity (dBA), jitter (%) and shimmer (%) in the four examined intensity levels (soft/medium/loud/>80 dBA). Soft
Medium
Loud
>80 dBA
Voice SPL (dBA) Mean (SD) Min–max (range)
69.3 (5.8) 55.1–84.8 (29.7)
76.2 (5.6) 62.7–93.8 (31.1)
85.5 (6.0) 69.9–97.6 (27.7)
91.4 (3.6) 83.8–99.4 (15.6)
Jitter (%) Mean (SD) Min–max (range)
1.27 (1.11) 0.24–6.9 (6.66)
0.50 (0.26) 0.22–2.03 (1.81)
0.34 (0.13) 0.16–0.77 (0.61)
0.27 (0.10) 0.12–0.61 (0.49)
Shimmer (%) Mean (SD) Min–max (range)
14.40 (4.43) 3.81–21.4 (17.60)
9.47 (3.47) 3.84–19.57 (15.73)
6.09 (2.34) 2.77–12.82 (10.05)
4.34 (1.68) 1.61–7.92 (6.31)
[(Fig._4)TD$IG]
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
2125
computer screen. As shown in Fig. 1, the majority of younger volunteers below 6;0 years of age were not able to perform this voice task. This may be explained by the fact, that school children have more mature conceptual skills and physical abilities [34]. This might enable older children to better control their phonation behavior. Thus in clinical practice, requesting phonations ‘‘>80 dBA’’ appears suitable for children in school age and/or developmental state. Based on the study findings we suggest providing visual feedback with real-time monitoring of voice SPL. This can be done with a supplemental computer screen or a separate SPL meter indicating the produced voice SPL and the target intensity level (indicated by a red line/a smiley). In addition, children should be always instructed in the same way to ensure comparability between measurements. Based on the study experiences we suggest to say ‘‘On this screen you can see how loud your voice is. Please speak louder than the indicated intensity level as shown by the red line/smiley’’. 4.3. Implications for our understanding of vocal function Fig. 4. Mean shimmer (%) with standard deviation (SD) per voice task (soft/medium/ loud/>80 dBA). Shimmer is smallest in phonations >80 dBA. Also the data spread is smallest in phonations with controlled voice SPL.
loudness. It is reasonable to assume that this is also true when children are prompted to phonate at ‘‘comfortable’’ voice intensity. The observed large differences in voice SPL between individuals may even be bigger in dysphonic children with reduced control of voice function. Jitter was 9 times higher (range: 0.22–2.03) from the softest to the loudest phonation in the same voice task to phonate at ‘‘medium’’ voice intensity. The effect was smaller for shimmer, where the results were 4 times higher in the softest phonation at ‘‘medium’’ voice loudness as compared to the loudest phonation (range 3.84–19.57). By comparison Valadez et al. reported a reduction in jitter and shimmer by factors of around 5 and 2 in children with nodules after voice therapy. As described in the introduction, the analysis system and program affect mean shimmer and jitter by factors ranging from 1.2 to 3.1 [23]. Thus the influence due to voice SPL changes is enormous compared to the effects of treatment or even assessment system and program. Furthermore, some treatment approaches aim to increase the speaking voice intensity of a patient. In these cases lower jitter and shimmer might be side effects of louder phonations, but not necessarily indicate improvements in vocal fold vibration patterns or the biomechanical vocal fold properties. Therefore the large confounding effects due to voice SPL might mask the treatment effects we are clinically interested in. As suggested by a clearly smaller standard deviation and range for all acoustic measures, using phonations with controlled voice SPL >80 dBA (10 cm distance) increases the reliability of both jitter and shimmer. However a better comparability between measurements and individuals may come at the cost of sensitivity. For example, in loud phonations clinically important information may be lost. Also, we do not know, if clinically relevant differences between individuals are leveled out, when all children phonate at the same intensity level. Even though dysphonic adults have been shown to reach a voice intensity level of 80 dBA it is unclear if this is true for dysphonic children [29,30]. Further it is still unclear, if and how this might affect acoustic perturbation. This calls for further investigations in children with a variety of voice disorders. 4.2. Preliminary clinical voice recording protocol Most children older than 7;0 years were able to control for their own voice intensity using visual feedback provided with a
From a physiologic perspective, it is not fully established, why different voice intensity levels impact on jitter and shimmer in children. With LMM analysis, there was no distinct difference between jitter in subjectively ‘‘loud’’ voice as compared to phonations at a prescribed voice intensity level of ‘‘>80 dBA’’. However this was not the case for shimmer. This highlights, that not all influencing factors in jitter and shimmer measurements are sufficiently understood. For adult voices it has been proposed, that the naturally increased vocal fold tone in louder phonations might lead to less deviant vibration patterns and hence lower acoustic perturbation [28]. This may explain lower jitter and shimmer with louder voicing, regardless of task type. Due to physiologic differences in vocal fold length and microstructure, this phenomenon has to be investigated separately in children [6]. 4.4. How useful are acoustic measurements in children? Instrumental acoustic measurements of jitter and shimmer have clear advantages in pediatric voice care: the assessment procedure is non-invasive, relatively easily applicable and allows an investigation of uninfluenced voice function. However as shown by the comparatively large confounding effects due to voice SPL differences, jitter and shimmer may not be meaningful without suitable control of voice intensity in children [22,30]. Until now there is only preliminary evidence, how voice SPL related effects are best controlled for in clinical practice. This suggests a practical question: how useful are jitter and shimmer at present in pediatric voice care? Based on the described study findings and the ongoing critical discussion about the limited reliability and sensitivity of jitter and shimmer, both parameters appear not appropriate to detect discrete vocal pathology [16,20–22,27]. Future work should establish the suitability of perturbation measurements to indicate voice disorders and therapy success in children. 5. Conclusions In children aged 5;0 to 9;11 years, differences in voice SPL significantly influence jitter and shimmer. Further the voice task we ask children to perform during instrumental assessments considerably affects vocal perturbation. These confounding effects were large compared to the jitter and shimmer changes we would expect after voice treatment. Jitter and shimmer variability is considerably reduced when children are instructed to phonate at ‘‘>80 dBA’’ (10 cm measuring distance) and provided with visual feedback during instrumental acoustic assessments. This protocol is suitable only for school children above 7;0 years. Future works in
2126
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
dysphonic children will have to establish if jitter and shimmer provide clinically more relevant information under these measurement conditions. Acknowledgment We thank Dr. Malzorgata Roos from the Dept. of Biostatistics, University Zurich, Switzerland, for assistance with the statistical analysis of the here presented data. References [1] W. Cohen, D.M. Wynne, H. Kubba, E. McCartney, Development of a minimum protocol for assessment in the paediatric voice clinic. Part 1: Evaluating vocal function, Logoped. Phoniatr. Vocol. 37 (1) (2012) 33–38. [2] L.N. Kelchner, S.B. Brehm, A. de Alarcon, B. Weinrich, Update on pediatric voice and airway disorders: assessment and care, Curr. Opin. Otolaryngol. Head Neck Surg. 20 (3) (2012) 160–164. [3] P.N. Carding, S. Roulstone, K. Northstone, A.S. Team, The prevalence of childhood dysphonia: a cross-sectional study, J. Voice 20 (4) (2006) 623–630. [4] N.P. Connor, S.B. Cohen, S.M. Theis, S.L. Thibeault, D.G. Heatley, D.M. Bless, Attitudes of children with dysphonia, J. Voice 22 (2) (2008) 197–209. [5] L.W. Lopes, I.L. Barbosa Lima, L.N. Alves Almeida, D.P. Cavalcante, A.A. de Almeida, Severity of voice disorders in children: correlations between perceptual and acoustic data, J. Voice 26 (6) (2012), 819 e7-12. [6] S. Maturo, C. Hill, G. Bunting, C. Ballif, R. Maurer, C. Hartnick, Establishment of a normative pediatric acoustic database, Arch. Otolaryngol. Head Neck Surg. 138 (10) (2012) 956–961. [7] P. Campisi, T.L. Tewfik, J.J. Manoukian, M.D. Schloss, E. Pelland-Blais, N. Sadeghi, Computer-assisted voice analysis: establishing a pediatric database, Arch. Otolaryngol. Head Neck Surg. 128 (2) (2002) 156–160. [8] M. Fuchs, Diagnostics and therapy of dysphonia suitable for the ages and developmental stages of children and adolescents (part 2), Laryngorhinootologie. 87 (2) (2008) 86–91. [9] J. Hirschberg, P.H. Dejonckere, M. Hirano, K. Mori, H.J. Schultz-Coulon, K. Vrticka, Voice disorders in children, Int. J. Pediatr. Otorhinolaryngol. 32 (Suppl. 1) (1995) S109–S125. [10] M. Wolf, A. Primov-Fever, O. Amir, D. Jedwab, The feasibility of rigid stroboscopy in children, Int. J. Pediatr. Otorhinolaryngol. 69 (8) (2005) 1077–1079. [11] I. Smillie, K. McManus, W. Cohen, E. Lawson, D.M. Wynne, The paediatric voice clinic, Arch. Dis. Child. 99 (10) (2014) 912–915. , http://dx.doi.org/10.1136/ archdischild-2, Epub 2014 May 28. [12] R. Shah, S. Engel, S. Choi, Relationship between voice quality and vocal nodule size, Otolaryngol. Head Neck Surg. 139 (5) (2008) 723–726. [13] V. Valadez, A. Ysunza, E. Ocharan-Hernandez, N. Garrido-Bustamante, A. SanchezValerio, M.C. Pamplona, Voice parameters and videonasolaryngoscopy in children with vocal nodules: a longitudinal study, before and after voice therapy, Int. J. Pediatr. Otorhinolaryngol. 76 (9) (2012) 1361–1365. [14] R. Baken, R.F. Orlikoff, Frequency perturbation (jitter), in: R. Baken, R.F. Orlikoff (Eds.), Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Albany, New York, 2000, pp. 190–213. [15] R. Baken, R.F. Orlikoff, Amplitude perturbation (shimmer), in: R. Baken, R.F. Orlikoff (Eds.), Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Albany, NY, 2000, pp. 130–137. [16] D.D. Mehta, R.E. Hillman, Voice assessment: updates on perceptual, acoustic, aerodynamic, and endoscopic imaging methods, Curr. Opin. Otolaryngol. Head Neck Surg. 16 (3) (2008) 211–215.
[17] M. Fuchs, M. Fro¨ehlich, B. Hentschel, I.W. Stuermer, E. Kruse, D. Knauft, Predicting mutational change in the speaking voice of boys, J. Voice 21 (2) (2007) 169–178. [18] R. Baken, R.F. Orlikoff, Amplitude perturbation (shimmer), in: Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Clifton Park, NY, 2000, pp. 130–137. [19] R.J. Baken, R.F. Orlikoff, Frequency perturbation (jitter), in: Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Clifton Park, NY, 2000, pp. 190–213. [20] P.N. Carding, I.N. Steen, A. Webb, K. Mackenzie, I.J. Deary, J.A. Wilson, The reliability and sensitivity to change of acoustic measures of voice quality, Clin. Otolaryngol. 29 (2004) 538–544. [21] P.N. Carding, J.A. Wilson, K. Mackenzie, I.J. Deary, Measuring voice outcomes: state of the science review, J. Laryngol. Otol. 123 (8) (2009) 823–829. [22] M. Brockmann-Bauser, M.J. Drinnan, Routine acoustic voice analysis: time to think again? Curr. Opin. Otolaryngol. Head Neck Surg. 19 (3) (2011) 165–170. [23] Y. Maryn, P. Corthals, M. Bodt De, P. Cauwenberge van, D. Delyiski, Perturbation measures of voice: a comparative study between multi-dimensional voice program and praat, Folia Phoniatr. Logop. 61 (4) (2009) 217–226. [24] I.R. Titze, Workshop on Acoustic Voice Analysis: Summary Statement, National Center for Voice and Speech, Iowa City, USA, 1995. [25] D. Deliyski, H. Shaw, M. Evans, R. Vesselinov, Regression tree approach to studying factors influencing acoustic voice analysis, Folia Phoniatr. Logop. 58 (2006) 274–288. [26] I.D. Bough Jr., R.J. Heuer, R.T. Sataloff, J.R. Hills, J.R. Cater, Intrasubject variability of objective voice measures, J. Voice 10 (2) (1996) 166–174. [27] C.A. Hill, S. Ojha, S. Maturo, R. Maurer, G. Bunting, C.J. Hartnick, Consistency of voice frequency and perturbation measures in children, Otolaryngol. Head Neck Surg. 148 (4) (2013) 637–641. [28] M. Brockmann, M.J. Drinnan, C. Storck, P.N. Carding, Reliable jitter and shimmer measurements in voice clinics: the relevance of vowel, gender, vocal intensity, and fundamental frequency effects in a typical clinical task, J. Voice 25 (1) (2011) 44–53. [29] M. Brockmann, C. Storck, P.N. Carding, M.J. Drinnan, Voice loudness and gender effects on jitter and shimmer in healthy adults, J. Speech Lang. Hear. Res. 51 (5) (2008) 1152–1160. [30] M. Brockmann-Bauser, Improving Jitter and Shimmer Measurements in Normal Voices, Schulz-Kirchner Verlag GmbH, Idstein, 2012p. 205. [31] W.S. Brown, R. Shrivatsav, Comfortable effort level in young children’s speech, Folia Phoniatr. Logop. 59 (5) (2007) 227–233. [32] L.E. Glaze, D.M. Bless, R.D. Susser, Acoustic analysis of vowel and loudness differences in children’s voice, J. Voice 4 (1) (1990) 37–44. [33] M. Brockmann, M.J. Drinnan, P.N. Carding, How to minimise gender and voice intensity effects in acoustic assessments,(Wie der Einfluss von Geschlecht und Lautsta¨rke in der klinischen akustischen Stimmuntersuchung minimiert werden kann), in: 25. Wissenschaftliche Jahrestagung der Deutschen Gesellschaft fu¨r Phoniatrie und Pa¨daudiologie e. V, German Medical Science GMS Publishing House, Du¨sseldorf, 2008. [34] S. Weintraub, S.S. Dikmen, R.K. Heaton, D.S. Tulsky, P.D. Zelazo, P.J. Bauer, et al., Cognition assessment using the NIH toolbox, Neurology 80 (11 Suppl 3) (2013) S54–S64. [35] M. Hirano, in: G.E. Arnold, F. Winckel, B.D. Wyke (Eds.), Clinical Examination of Voice. Disorders of Human Communication, vol. 5, Springer-Verlag, Wien/New York, 1981, pp. 81–84. [36] D. Mazzoni, M. Brubeck, J. Crook, V. Johnson, M. Meyer, J. Habermann, et al., Audacity, Free Software Foundation, 2008. [37] P. Boersma, D. Weenink, PRAAT, University of Amsterdam, Amsterdam, 2009. [38] W.T.I. Winholtz, Conversion of a head-mounted microphone signal into calibrated SPL units, J. Voice 11 (4) (1997) 417–421. [39] F. Wagener, V. Ku¨hnel, B. Kollmeier, Entwicklung und Evaluation eines Satztestes in deutscher Sprache I: Design des Oldenburger Satztestes (German),(Development and evaluation of a sentence based hearing test in german language I: design of the Oldenburger Satztest), Z. Audiol. 38 (1) (1999) 4–15. [40] Microsoft Office Excel, 2003, Microsoft Corporation, Software. [41] SPSS Statistics, Version 17, 2008, Software.
Contents lists available at ScienceDirect
International Journal of Pediatric Otorhinolaryngology journal homepage: www.elsevier.com/locate/ijporl
Clinical relevance of speaking voice intensity effects on acoustic jitter and shimmer in children between 5;0 and 9;11 years Meike Brockmann-Bauser *, Denis Beyer, Jo¨rg Edgar Bohlender Department of Phoniatrics and Speech Pathology, Clinic for Otorhinolaryngology, Head and Neck Surgery, University Hospital Zurich, Zurich, Switzerland
A R T I C L E I N F O
A B S T R A C T
Article history: Received 14 August 2014 Received in revised form 17 September 2014 Accepted 19 September 2014 Available online 28 September 2014
Background: Current voice assessment recommendations for dysphonic children comprise instrumental acoustic measurements of the perturbation parameters jitter and shimmer. In healthy adults and children changes in speaking voice sound pressure level (voice SPL) have significant confounding effects on both parameters. In adults these effects were considerably reduced in phonations with controlled voice SPL >80 dBA (10 cm distance). However, it is unclear if these findings apply to children and if children are able to control for their own voice intensity. Objective: This cross-sectional single cohort study investigates voice SPL effects on jitter and shimmer in children between 5;0 and 9;11 years phonating at individually ‘‘medium’’ (modeling ‘‘comfortable’’ loudness of the usual clinical protocol), ‘‘soft’’ and ‘‘loud’’ voice and a prescribed intensity level of ‘‘>80 dBA’’ (10 cm distance, with visual control). Further both their ability to phonate at a prescribed voice intensity level and the effect on SPL related confounding effects were studied. Subjects and methods: A total of 68 healthy children (39 f/29 m) aged 5;0 to 9;11 years were included. All phonated the vowel/a/for 5 s, three times at four defined voice intensity levels (soft/medium/loud/ >80 dBA) each. Jitter (%), shimmer (%) and voice SPL (dBA) were determined using PRAAT. Voice intensity level effects were assessed by descriptive statistics, Analysis of Variance (ANOVA) and Linear Mixed Models (LMM). Results: There were significant differences for jitter and shimmer between all voice tasks (p < .01). Jitter and shimmer were lowest and showed the smallest spread in controlled phonations ‘‘>80 dBA’’. 19 children below 7;0 years could not perform the voice tasks and were excluded from the study. Conclusions: This practical study demonstrated a significant effect of voice loudness and task on jitter and shimmer in children. Since the observed confounding effects were large compared to treatment effects, jitter and shimmer may not be meaningful without adequate control of voice SPL. In phonations at ‘‘>80 dBA’’ (10 cm distance) voice SPL related effects were considerably reduced. However, this assessment protocol was suitable only for children above 7;0 years. Application of this task to future studies of dysphonic children may yield clinically valuable information. ß 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Voice diagnostics Children Voice intensity Jitter Shimmer
1. Introduction According to pediatric voice assessment guidelines a comprehensive clinical voice examination usually includes visual, perceptual, patient based subjective and instrumental acoustic assessment techniques [1,2]. Studies investigating the incidence of
* Corresponding author at: Department for Phoniatrics and Speech Pathology; Clinic for Otorhinolaryngology, Head and Neck Surgery, University Hospital Zurich, Frauenklinikstrasse 24, 8091 Zurich, Switzerland. Tel.: +41 44 255 5830; fax: +41 44 255 4424. E-mail addresses: [email protected] (M. Brockmann-Bauser), [email protected] (D. Beyer), [email protected] (J.E. Bohlender). http://dx.doi.org/10.1016/j.ijporl.2014.09.020 0165-5876/ß 2014 Elsevier Ireland Ltd. All rights reserved.
voice disorders during childhood report a prevalence from 6% to 38% [3]. This wide prevalence range highlights the complexity of pediatric voice diagnostics and the difficulties to determine vocal pathology in children. A voice disorder may compromise the general well-being, communication behavior and the social and academic development of a child [4]. Teachers and parents tend to judge the personality and cognitive abilities of a dysphonic child more negatively than of a normophonic child [5,6]. Therefore an early diagnosis and treatment of any voice disorder is key to avoiding considerable and probably even long lasting negative effects on the child’s life. Up to 20% of children below the age of 10 years may not tolerate invasive assessment procedures such as laryngostroboscopy
2122
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
[1,7–10]. Further in patients with muscle tension dysphonia, one of the most frequent diagnoses in children, laryngostroboscopy has limited validity to determine pathologic voice production patterns [11]. In these cases the clinical diagnosis and treatment decision depends on perceptual, subjective and instrumental acoustic assessment methods. Instrumental acoustic measurements allow an objective and non-invasive assessment of uninfluenced voice function by analyzing the vocal output. For this the patient’s voice is recorded with a microphone, and acoustic analysis usually is done by a computer based software application. Even children of a very young age tolerate this procedure. Therefore instrumental acoustic assessments are considered as easily applicable in pediatric voice care [1]. The present work focuses on the acoustic perturbation parameters jitter and shimmer, indicating the irregularity of human voice pitch (jitter) and intensity (shimmer) respectively from one acoustic wave to the next. The broad clinical application of jitter and shimmer is based on the hypothesis that the acoustic waveform represents the vibratory characteristics of the vocal folds. A variety of studies in children and adults show that vocal pathology and/or dysphonia are associated with increased jitter and shimmer [1,5,12–16]. Also, the onset of mutation in children was accompanied by increased acoustic voice perturbation [17]. It has been suggested, that jitter and shimmer may even indicate subtle changes in the vibratory properties of the vocal folds [16,18,19]. Based on this jitter and shimmer have been characterized as informative and clinically valuable parameters to determine pathology and mutation onset in childrens’ voices [1,13]. However, in adults and children the reliability and usefulness of jitter and shimmer measurements has been questioned for a number of reasons [20–22]: both parameters are significantly influenced by technical confounding factors such as microphone quality, the analysis software type or background noise. The analysis system and program may affect mean shimmer and jitter by factors ranging from 1.2 to 3.1 [23]. Also the type of the analyzed voice signal determines the exactness and clinical usefulness of acoustic measurements. Several authors have argued that perturbation parameters are not meaningful in severely irregular voices, since jitter and shimmer depend on correct recognition of fundamental frequency and voice SPL [20,21,23–25]. However even under adequate measurement conditions both jitter and shimmer vary considerably within healthy adults and children during a day [26,27]. A practical reason for this might be how patients are instructed during acoustic assessments [22]. According to current guidelines, patients are usually asked to phonate at ‘‘comfortable loudness and pitch’’ [1]. Under this assessment protocol, adults have substantial interindividual differences in speaking voice SPL, which significantly influence jitter and shimmer [28–30]. A statistical analysis of data by means of etasquared in phonations at individually ‘‘normal’’ voice loudness showed that 62% of shimmer variance and 24% of jitter variance could be explained by changes in voice SPL. The effect sizes of vowel, gender and fundamental frequency (F0) were considerably smaller ranging from 0% to 4% [28]. Therefore the reliability of jitter and shimmer measurements in adults could be considerably improved with adequate control of voice SPL [22,28]. Children also vary substantially in their speaking voice intensity, when asked to phonate at ‘‘comfortable’’ loudness [31]. Further a study by Glaze et al. showed, that voice intensity differences have significant effects on both jitter and shimmer in children [32]. However it has not been investigated, how these confounding effects may be sufficiently controlled for in clinical practice. Preliminary findings in healthy adults suggest, that the effects due to differences in speaking voice SPL may be considerably reduced, when patients are asked to phonate at a minimum of 80 dBA (measured at 10 cm distance) [30,33].
However these findings may not apply to children, since vocal fold length and microstructure change with age. Currently it is not fully established, how these developmental changes affect the vibratory properties of the vocal folds and hence jitter and shimmer [6]. Also, from a practical point of view, it is unclear if children are able to phonate at a prescribed voice intensity level. Specifically younger children may be unable to control for their own voice SPL, since their conceptual skills and physical abilities are not fully developed [34]. To the best of our knowledge these issues have not been investigated in children to date, and therefore will be key aims of the present study. 1.1. Study aims The aims of the present study in children between 5;0 and 9;11 years were to characterize the effects of voice intensity changes on jitter and shimmer in a variety of voice tasks. Specifically it was tested if children of this age group are able to phonate at a prescribed voice intensity level of ‘‘>80 dBA’’, and if this minimizes confounding effects due to differences in voice SPL. 2. Materials and methods 2.1. Population studied A total of 87 children aged between 5;0 and 9;11 years, 47 girls and 40 boys, were recruited to the present cross-sectional cohort study. Of these, 19 children were not able to perform the voice tasks and were excluded. Please refer for further details about the specific exclusion reasons to Section 3. All children were students of three schools and kindergartens in Zurich, Switzerland. This study was approved by the responsible ethical review board under reference number KEK-ZH-Nr. 2010-0305/2. 2.2. Recruiting process and exclusion criteria Prior to the study information packages were send out to all parents of students between 5;0 and 9;11 years of age. All packages included detailed study information, an informed consent form and a participant questionnaire assessing the eligibility of children for the study. 142 parents returned the informed consent and questionnaires. Oral consent was obtained from all children on the assessment day. Children were excluded from the study if they met one or more of the following criteria: previous formal voice training or therapy; acute infection of the ears, nose and throat or allergic reaction on the recording day; a medical condition or medication that might affect normal voice function; surgery in the torso, head and neck region or intubation for any intervention within the last 18 months; inability to say the vowel/a/for 5 s in three auditively discernible loudness levels (’’soft’’, ‘‘medium’’, ‘‘loud’’ voice) and at a minimum of 80 dBA (measured at 10 cm distance) after a maximum training phase of 5 minutes; perceptual voice pathology, indicated by a mean of 1 in any GRBAS scale characteristic as assessed by the study examiner and a speech pathologist [35].
2.3. Recording and analysis technique All voice recordings were made at the children’s school or kindergarten during break time in a quiet room with ambient noise
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
2123
Table 2 Age distribution of the investigated children group per year (year;months) and gender (f;m).
below 50 dBA. A laptop with the recording software Audacity (Version 1.3.6. beta) and an external soundcard (EMU 0404 USB 2.0 Audio) set at a sampling rate of 48,000 Hz with 16 bits per sample were utilized [36]. An extra computer screen with monitoring facilities to indicate voice SPL in real time was connected to the laptop. The target voice intensity level of 80 dBA was marked with a red line and a smiley. A head-mounted microphone (AKG Acoustics, C444) with 10 cm microphone-mouth distance in an offaxis position was used [24]. Each single phonation was edited into an individual sound file and labeled anonymously using Audacity. Acoustic analysis was conducted with the software PRAAT Version 5.1.03 [37]. Only second 1.0 from voice onset to second 3.0 was analyzed to exclude the increased variability of the voice onset and offset phase. Comparison method was used to calculate calibrated SPL values from the voice recordings [38]. Prior to the experiment, speech weighted noise was recorded with 10 cm distance to sound source at 65 dBA, 75 dBA, 85 dBA and 95 dBA [39]. The difference between the known SPL values of the speech weighted noise and the uncalibrated values as indicated by PRAAT was used to calculate calibrated SPL values (dBA) of all voice recordings.
Age group (years;months)
5;0 to 5;11
6;0 to 6;11
7;0 to 7;11
8;0 to 8;11
9;0 to 9;11
Overall 5;0 to 9;11
Girls (f) Boys (m) Total
8 1 9
9 6 15
10 8 18
4 11 15
10 1 11
39 29 68
[40]. These mean values were used for further analysis. Thereafter, the mean, standard deviation (SD), minimum, maximum and range of all 3 parameters were determined with SPSS [41]. Analysis of Variance (ANOVA) was applied to determine the overall effects of voice intensity levels (soft/medium/loud/>80 dBA) on jitter and shimmer. As described above, in the present study our participants provided repeated phonations at a variety of intensity levels. Since measurements within individuals tend to be more similar than between individuals, Linear Mixed Models (LMM) were used to further investigate the relation between each voice intensity level (soft/medium/loud/>80 dBA) and jitter or shimmer.
2.4. Voice tasks and recording procedure 3. Results During a preparation phase of 10 min maximum, the children were able to practice the voice tasks. First, all participants were asked to ‘‘sustain/a/for 5 s at medium pitch and loudness’’. This task was used to model ‘‘comfortable’’ voice intensity as usually applied in voice clinics. Thereafter they were asked to sustain/a/ for 5 s, ‘‘as softly as possible’’ and then ‘‘as loudly as possible’’. When the children were able to phonate at three recognizable different loudness levels (soft/medium/loud), all were recorded phonating/a/ for 5 s at each level three times in randomized order. The second voice task was to sustain the vowel /a/ for 5 s while controlling for voice intensity under visual feedback. Live monitoring of voice SPL was provided with a second monitor, which was turned toward the children. The target level of 80 dBA was marked with a red line and a smiley. This way every participant could see, if her or his own voice intensity reached the requested intensity level. All children were asked to ‘‘please sustain/a/for 5 s and please be louder than indicated by the red line with the smiley’’. When the children were able to do this, they were recorded three times. 2.5. Main outcome measures Main outcome measures were the instrumental parameters voice SPL (dBA), jitter (%) and shimmer (%) (Table 1). The indices jitter (%) and shimmer (%) were chosen, since both are normalized for an individual’s voice SPL and F0 [37].
A total of 87 children between 5;0 and 9;11 years were originally recruited to the present study. Of these, 19 participants, who were not able to perfom the given voice tasks, had to be excluded. Thus 68 participants, 39 girls and 29 boys with a mean age of 7;6 years, were included for voice analysis. Table 2 summarizes the age and gender distribution of the studied group. Each subject provided 12 phonations, giving a total of 816 phonations for analysis. 3.1. Compliance to voice task As described before, 19 children were unable to perform the voice tasks and had to be excluded from this study. Notably, most of these children were still in kindergarten and below 6 years of age (Fig. 1). The main exclusion reasons were a too short phonation time, the inability to produce different loudness levels or the prescribed level of ‘‘>80 dBA’’. There was a significant difference in voice SPL (dBA) between the four examined intensity levels ‘‘soft’’,’’ medium’’, ‘‘loud’’ voice and ‘‘>80 dBA’’ (ANOVA and LMM: p < 0.01). As shown in Fig. 2 and Table 3 mean voice SPL was lowest in ‘‘soft’’ phonations (69.3 dBA, SD 5.8) and highest in phonations ‘‘>80 dBA’’ (91.4 dBA, SD 3.6). Further the intersubject range between the individually softest and
[(Fig._1)TD$IG]
25
First, the mean of the three repetitions (phonations) per child and task type were determined for the instrumental acoustic parameters voice SPL (dBA), jitter (%) and shimmer (%) using Excel
No. of children
20
2.6. Statistical analysis
15 10 5 0
Table 1 Main outcome parameters with labels as indicated by the software PRAAT.
4;0 - 4;11
5;0 - 5;11
6;0 - 6;11
7;0 - 7;11
8;0 - 8;11
9;0 - 9;11
Age (years;months) Instrumental parameter
Label software PRAAT
Jitter (%) Shimmer (%) Voice SPL (dBA)
Local jitter Local shimmer Mean energy intensity
Included children
Excluded children
Fig. 1. Number of excluded children per age group (years;months). The majority of children below 6;0 years of age was not able to perform the given voice tasks.
[(Fig._2)TD$IG]
2124
[(Fig._3)TD$IG]
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
Fig. 2. Mean voice SPL (dBA) with standard deviation (SD) and outliers per voice task (soft/medium/loud/>80 dBA) for 68 children between 5;0 and 9;11 years. The difference in produced voice SPL between individuals is considerably reduced in phonations with a prescribed intensity level of >80 dBA.
Fig. 3. Mean jitter (%) with standard deviation (SD) per voice task (soft/medium/ loud/>80 dBA). Outliers are marked with a cross. In phonations >80 dBA jitter is lowest and shows the smallest spread.
4. Discussion loudest phonation was with 15.6 dBA clearly smallest in the voice task to phonate at ‘‘>80 dBA’’, as compared to ranges from 27.7 to 31.1 dBA in all other voice tasks (Fig. 2, Table 3). Also the standard deviation (SD), a summary measure for the differences of each observation from the mean, was clearly smallest in phonations at >80 dBA (Table 3). 3.2. Effects of voice intensity levels on jitter and shimmer According to the statistical analysis with ANOVA there were significant effects of loudness level (soft/medium/loud/>80 dBA) on jitter (%) (p > 0.01). Notably, with LMM there was no significant difference between the levels ‘‘loud’’ voice and ‘‘>80 dBA’’ (p = 0.45). Jitter was smallest (0.27) and showed the lowest standard deviation and range (SD: 0.1; range: 0.49) in phonations ‘‘>80 dBA’’ (Fig. 3 and Table 1). With both ANOVA and LMM there were significant differences for shimmer (%) between the four examined intensity levels ‘‘soft’’,’’ medium’’, ‘‘loud’’ voice and ‘‘> 80 dBA’’ (p < 0.01). Analogously to jitter, shimmer was lowest in phonations ‘‘>80 dBA’’. Also for shimmer the spread between measurements was with a SD of 1.68 and range of 6.31% by far smallest in phonations at ‘‘>80 dBA’’, as compared to a SD from 2.34 to 4.43 and range from 10.05% to 17.6% in ‘‘loud’’ and ‘‘soft’’ phonations (Fig. 4 and Table 3).
Changes in voice intensity have significant confounding effects on acoustic jitter and shimmer measurements in children. In addition, the voice task we ask children to perform, significantly affects both acoustic parameters. The present study findings suggest that requesting phonations ‘‘>80 dBA’’ (at 10 cm distance) reduces the substantial confounding effects of differences in voice intensity between individuals. Especially for jitter, there was a considerably lower spread in phonations at a controlled voice SPL. To ensure the reliability of jitter and shimmer measurements, assessment guidelines for children should include a detailed characterization of the voice recording procedure and specify oral instructions. However, to date it is unclear, if by applying this procedure clinically relevant information or differences between individuals would be masked. 4.1. How important were the confounding effects of voice SPL? Especially Fig. 4 and Table 3 show the consequences of differences in habitual voice SPL between children. Under the same clinical protocol to phonate at ‘‘medium’’ voice intensity, the softest individual phonation was 62.7 dBA, whereas the loudest was 93.8 dBA. This wide range covered nearly the total range from the softest phonation at a ‘‘soft’’ level (55.1 dBA) and the loudest phonation with ‘‘loud’’ voice (97.6 dBA). Thus voice intensity control is comparatively small in phonations at ‘‘medium’’ voice
Table 3 Mean values with standard deviation (SD), minimum (min) and maximum (max) plus range for the instrumental parameters voice intensity (dBA), jitter (%) and shimmer (%) in the four examined intensity levels (soft/medium/loud/>80 dBA). Soft
Medium
Loud
>80 dBA
Voice SPL (dBA) Mean (SD) Min–max (range)
69.3 (5.8) 55.1–84.8 (29.7)
76.2 (5.6) 62.7–93.8 (31.1)
85.5 (6.0) 69.9–97.6 (27.7)
91.4 (3.6) 83.8–99.4 (15.6)
Jitter (%) Mean (SD) Min–max (range)
1.27 (1.11) 0.24–6.9 (6.66)
0.50 (0.26) 0.22–2.03 (1.81)
0.34 (0.13) 0.16–0.77 (0.61)
0.27 (0.10) 0.12–0.61 (0.49)
Shimmer (%) Mean (SD) Min–max (range)
14.40 (4.43) 3.81–21.4 (17.60)
9.47 (3.47) 3.84–19.57 (15.73)
6.09 (2.34) 2.77–12.82 (10.05)
4.34 (1.68) 1.61–7.92 (6.31)
[(Fig._4)TD$IG]
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
2125
computer screen. As shown in Fig. 1, the majority of younger volunteers below 6;0 years of age were not able to perform this voice task. This may be explained by the fact, that school children have more mature conceptual skills and physical abilities [34]. This might enable older children to better control their phonation behavior. Thus in clinical practice, requesting phonations ‘‘>80 dBA’’ appears suitable for children in school age and/or developmental state. Based on the study findings we suggest providing visual feedback with real-time monitoring of voice SPL. This can be done with a supplemental computer screen or a separate SPL meter indicating the produced voice SPL and the target intensity level (indicated by a red line/a smiley). In addition, children should be always instructed in the same way to ensure comparability between measurements. Based on the study experiences we suggest to say ‘‘On this screen you can see how loud your voice is. Please speak louder than the indicated intensity level as shown by the red line/smiley’’. 4.3. Implications for our understanding of vocal function Fig. 4. Mean shimmer (%) with standard deviation (SD) per voice task (soft/medium/ loud/>80 dBA). Shimmer is smallest in phonations >80 dBA. Also the data spread is smallest in phonations with controlled voice SPL.
loudness. It is reasonable to assume that this is also true when children are prompted to phonate at ‘‘comfortable’’ voice intensity. The observed large differences in voice SPL between individuals may even be bigger in dysphonic children with reduced control of voice function. Jitter was 9 times higher (range: 0.22–2.03) from the softest to the loudest phonation in the same voice task to phonate at ‘‘medium’’ voice intensity. The effect was smaller for shimmer, where the results were 4 times higher in the softest phonation at ‘‘medium’’ voice loudness as compared to the loudest phonation (range 3.84–19.57). By comparison Valadez et al. reported a reduction in jitter and shimmer by factors of around 5 and 2 in children with nodules after voice therapy. As described in the introduction, the analysis system and program affect mean shimmer and jitter by factors ranging from 1.2 to 3.1 [23]. Thus the influence due to voice SPL changes is enormous compared to the effects of treatment or even assessment system and program. Furthermore, some treatment approaches aim to increase the speaking voice intensity of a patient. In these cases lower jitter and shimmer might be side effects of louder phonations, but not necessarily indicate improvements in vocal fold vibration patterns or the biomechanical vocal fold properties. Therefore the large confounding effects due to voice SPL might mask the treatment effects we are clinically interested in. As suggested by a clearly smaller standard deviation and range for all acoustic measures, using phonations with controlled voice SPL >80 dBA (10 cm distance) increases the reliability of both jitter and shimmer. However a better comparability between measurements and individuals may come at the cost of sensitivity. For example, in loud phonations clinically important information may be lost. Also, we do not know, if clinically relevant differences between individuals are leveled out, when all children phonate at the same intensity level. Even though dysphonic adults have been shown to reach a voice intensity level of 80 dBA it is unclear if this is true for dysphonic children [29,30]. Further it is still unclear, if and how this might affect acoustic perturbation. This calls for further investigations in children with a variety of voice disorders. 4.2. Preliminary clinical voice recording protocol Most children older than 7;0 years were able to control for their own voice intensity using visual feedback provided with a
From a physiologic perspective, it is not fully established, why different voice intensity levels impact on jitter and shimmer in children. With LMM analysis, there was no distinct difference between jitter in subjectively ‘‘loud’’ voice as compared to phonations at a prescribed voice intensity level of ‘‘>80 dBA’’. However this was not the case for shimmer. This highlights, that not all influencing factors in jitter and shimmer measurements are sufficiently understood. For adult voices it has been proposed, that the naturally increased vocal fold tone in louder phonations might lead to less deviant vibration patterns and hence lower acoustic perturbation [28]. This may explain lower jitter and shimmer with louder voicing, regardless of task type. Due to physiologic differences in vocal fold length and microstructure, this phenomenon has to be investigated separately in children [6]. 4.4. How useful are acoustic measurements in children? Instrumental acoustic measurements of jitter and shimmer have clear advantages in pediatric voice care: the assessment procedure is non-invasive, relatively easily applicable and allows an investigation of uninfluenced voice function. However as shown by the comparatively large confounding effects due to voice SPL differences, jitter and shimmer may not be meaningful without suitable control of voice intensity in children [22,30]. Until now there is only preliminary evidence, how voice SPL related effects are best controlled for in clinical practice. This suggests a practical question: how useful are jitter and shimmer at present in pediatric voice care? Based on the described study findings and the ongoing critical discussion about the limited reliability and sensitivity of jitter and shimmer, both parameters appear not appropriate to detect discrete vocal pathology [16,20–22,27]. Future work should establish the suitability of perturbation measurements to indicate voice disorders and therapy success in children. 5. Conclusions In children aged 5;0 to 9;11 years, differences in voice SPL significantly influence jitter and shimmer. Further the voice task we ask children to perform during instrumental assessments considerably affects vocal perturbation. These confounding effects were large compared to the jitter and shimmer changes we would expect after voice treatment. Jitter and shimmer variability is considerably reduced when children are instructed to phonate at ‘‘>80 dBA’’ (10 cm measuring distance) and provided with visual feedback during instrumental acoustic assessments. This protocol is suitable only for school children above 7;0 years. Future works in
2126
M. Brockmann-Bauser et al. / International Journal of Pediatric Otorhinolaryngology 78 (2014) 2121–2126
dysphonic children will have to establish if jitter and shimmer provide clinically more relevant information under these measurement conditions. Acknowledgment We thank Dr. Malzorgata Roos from the Dept. of Biostatistics, University Zurich, Switzerland, for assistance with the statistical analysis of the here presented data. References [1] W. Cohen, D.M. Wynne, H. Kubba, E. McCartney, Development of a minimum protocol for assessment in the paediatric voice clinic. Part 1: Evaluating vocal function, Logoped. Phoniatr. Vocol. 37 (1) (2012) 33–38. [2] L.N. Kelchner, S.B. Brehm, A. de Alarcon, B. Weinrich, Update on pediatric voice and airway disorders: assessment and care, Curr. Opin. Otolaryngol. Head Neck Surg. 20 (3) (2012) 160–164. [3] P.N. Carding, S. Roulstone, K. Northstone, A.S. Team, The prevalence of childhood dysphonia: a cross-sectional study, J. Voice 20 (4) (2006) 623–630. [4] N.P. Connor, S.B. Cohen, S.M. Theis, S.L. Thibeault, D.G. Heatley, D.M. Bless, Attitudes of children with dysphonia, J. Voice 22 (2) (2008) 197–209. [5] L.W. Lopes, I.L. Barbosa Lima, L.N. Alves Almeida, D.P. Cavalcante, A.A. de Almeida, Severity of voice disorders in children: correlations between perceptual and acoustic data, J. Voice 26 (6) (2012), 819 e7-12. [6] S. Maturo, C. Hill, G. Bunting, C. Ballif, R. Maurer, C. Hartnick, Establishment of a normative pediatric acoustic database, Arch. Otolaryngol. Head Neck Surg. 138 (10) (2012) 956–961. [7] P. Campisi, T.L. Tewfik, J.J. Manoukian, M.D. Schloss, E. Pelland-Blais, N. Sadeghi, Computer-assisted voice analysis: establishing a pediatric database, Arch. Otolaryngol. Head Neck Surg. 128 (2) (2002) 156–160. [8] M. Fuchs, Diagnostics and therapy of dysphonia suitable for the ages and developmental stages of children and adolescents (part 2), Laryngorhinootologie. 87 (2) (2008) 86–91. [9] J. Hirschberg, P.H. Dejonckere, M. Hirano, K. Mori, H.J. Schultz-Coulon, K. Vrticka, Voice disorders in children, Int. J. Pediatr. Otorhinolaryngol. 32 (Suppl. 1) (1995) S109–S125. [10] M. Wolf, A. Primov-Fever, O. Amir, D. Jedwab, The feasibility of rigid stroboscopy in children, Int. J. Pediatr. Otorhinolaryngol. 69 (8) (2005) 1077–1079. [11] I. Smillie, K. McManus, W. Cohen, E. Lawson, D.M. Wynne, The paediatric voice clinic, Arch. Dis. Child. 99 (10) (2014) 912–915. , http://dx.doi.org/10.1136/ archdischild-2, Epub 2014 May 28. [12] R. Shah, S. Engel, S. Choi, Relationship between voice quality and vocal nodule size, Otolaryngol. Head Neck Surg. 139 (5) (2008) 723–726. [13] V. Valadez, A. Ysunza, E. Ocharan-Hernandez, N. Garrido-Bustamante, A. SanchezValerio, M.C. Pamplona, Voice parameters and videonasolaryngoscopy in children with vocal nodules: a longitudinal study, before and after voice therapy, Int. J. Pediatr. Otorhinolaryngol. 76 (9) (2012) 1361–1365. [14] R. Baken, R.F. Orlikoff, Frequency perturbation (jitter), in: R. Baken, R.F. Orlikoff (Eds.), Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Albany, New York, 2000, pp. 190–213. [15] R. Baken, R.F. Orlikoff, Amplitude perturbation (shimmer), in: R. Baken, R.F. Orlikoff (Eds.), Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Albany, NY, 2000, pp. 130–137. [16] D.D. Mehta, R.E. Hillman, Voice assessment: updates on perceptual, acoustic, aerodynamic, and endoscopic imaging methods, Curr. Opin. Otolaryngol. Head Neck Surg. 16 (3) (2008) 211–215.
[17] M. Fuchs, M. Fro¨ehlich, B. Hentschel, I.W. Stuermer, E. Kruse, D. Knauft, Predicting mutational change in the speaking voice of boys, J. Voice 21 (2) (2007) 169–178. [18] R. Baken, R.F. Orlikoff, Amplitude perturbation (shimmer), in: Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Clifton Park, NY, 2000, pp. 130–137. [19] R.J. Baken, R.F. Orlikoff, Frequency perturbation (jitter), in: Clinical Measurements of Speech and Voice, Thomson Delmar Learning, Clifton Park, NY, 2000, pp. 190–213. [20] P.N. Carding, I.N. Steen, A. Webb, K. Mackenzie, I.J. Deary, J.A. Wilson, The reliability and sensitivity to change of acoustic measures of voice quality, Clin. Otolaryngol. 29 (2004) 538–544. [21] P.N. Carding, J.A. Wilson, K. Mackenzie, I.J. Deary, Measuring voice outcomes: state of the science review, J. Laryngol. Otol. 123 (8) (2009) 823–829. [22] M. Brockmann-Bauser, M.J. Drinnan, Routine acoustic voice analysis: time to think again? Curr. Opin. Otolaryngol. Head Neck Surg. 19 (3) (2011) 165–170. [23] Y. Maryn, P. Corthals, M. Bodt De, P. Cauwenberge van, D. Delyiski, Perturbation measures of voice: a comparative study between multi-dimensional voice program and praat, Folia Phoniatr. Logop. 61 (4) (2009) 217–226. [24] I.R. Titze, Workshop on Acoustic Voice Analysis: Summary Statement, National Center for Voice and Speech, Iowa City, USA, 1995. [25] D. Deliyski, H. Shaw, M. Evans, R. Vesselinov, Regression tree approach to studying factors influencing acoustic voice analysis, Folia Phoniatr. Logop. 58 (2006) 274–288. [26] I.D. Bough Jr., R.J. Heuer, R.T. Sataloff, J.R. Hills, J.R. Cater, Intrasubject variability of objective voice measures, J. Voice 10 (2) (1996) 166–174. [27] C.A. Hill, S. Ojha, S. Maturo, R. Maurer, G. Bunting, C.J. Hartnick, Consistency of voice frequency and perturbation measures in children, Otolaryngol. Head Neck Surg. 148 (4) (2013) 637–641. [28] M. Brockmann, M.J. Drinnan, C. Storck, P.N. Carding, Reliable jitter and shimmer measurements in voice clinics: the relevance of vowel, gender, vocal intensity, and fundamental frequency effects in a typical clinical task, J. Voice 25 (1) (2011) 44–53. [29] M. Brockmann, C. Storck, P.N. Carding, M.J. Drinnan, Voice loudness and gender effects on jitter and shimmer in healthy adults, J. Speech Lang. Hear. Res. 51 (5) (2008) 1152–1160. [30] M. Brockmann-Bauser, Improving Jitter and Shimmer Measurements in Normal Voices, Schulz-Kirchner Verlag GmbH, Idstein, 2012p. 205. [31] W.S. Brown, R. Shrivatsav, Comfortable effort level in young children’s speech, Folia Phoniatr. Logop. 59 (5) (2007) 227–233. [32] L.E. Glaze, D.M. Bless, R.D. Susser, Acoustic analysis of vowel and loudness differences in children’s voice, J. Voice 4 (1) (1990) 37–44. [33] M. Brockmann, M.J. Drinnan, P.N. Carding, How to minimise gender and voice intensity effects in acoustic assessments,(Wie der Einfluss von Geschlecht und Lautsta¨rke in der klinischen akustischen Stimmuntersuchung minimiert werden kann), in: 25. Wissenschaftliche Jahrestagung der Deutschen Gesellschaft fu¨r Phoniatrie und Pa¨daudiologie e. V, German Medical Science GMS Publishing House, Du¨sseldorf, 2008. [34] S. Weintraub, S.S. Dikmen, R.K. Heaton, D.S. Tulsky, P.D. Zelazo, P.J. Bauer, et al., Cognition assessment using the NIH toolbox, Neurology 80 (11 Suppl 3) (2013) S54–S64. [35] M. Hirano, in: G.E. Arnold, F. Winckel, B.D. Wyke (Eds.), Clinical Examination of Voice. Disorders of Human Communication, vol. 5, Springer-Verlag, Wien/New York, 1981, pp. 81–84. [36] D. Mazzoni, M. Brubeck, J. Crook, V. Johnson, M. Meyer, J. Habermann, et al., Audacity, Free Software Foundation, 2008. [37] P. Boersma, D. Weenink, PRAAT, University of Amsterdam, Amsterdam, 2009. [38] W.T.I. Winholtz, Conversion of a head-mounted microphone signal into calibrated SPL units, J. Voice 11 (4) (1997) 417–421. [39] F. Wagener, V. Ku¨hnel, B. Kollmeier, Entwicklung und Evaluation eines Satztestes in deutscher Sprache I: Design des Oldenburger Satztestes (German),(Development and evaluation of a sentence based hearing test in german language I: design of the Oldenburger Satztest), Z. Audiol. 38 (1) (1999) 4–15. [40] Microsoft Office Excel, 2003, Microsoft Corporation, Software. [41] SPSS Statistics, Version 17, 2008, Software.
