Vocal Fatigue Index (VFI): Development and Validation *Chayadevie Nanjundeswaran, †Barbara H. Jacobson, ‡Jackie Gartner-Schmidt, and ‡,§,k,{,#Katherine Verdolini Abbott, *Johnson City and yNashville, Tennessee, and zxk{#Pittsburgh, Pennsylvania Summary: Objective. To develop a psychometrically sound self-report questionnaire, the Vocal Fatigue Index (VFI), to help identify individuals with vocal fatigue (VF) and characterize their complaints. Study Design. Descriptive research—scale development. Methods. Four laryngologists and six speech-language pathologists specialized in voice created a beta version of the VFI (version 1), an index of 21 statements they considered to reflect VF. Two hundred patients presenting to two different clinics completed the VFI-1. Two items from VFI-1 were excluded because of poor item-to-total correlations. The final VFI of 19 items (version 2), completed by 105 patients with voice complaints and 70 vocally healthy individuals, was assessed for its psychometric properties. Results. Test-retest reliability for the final VFI was generally strong, as was sensitivity and specificity using the classification table under logistic regression for correctly distinguishing individuals with and without VF. Moreover, factor analysis indicated that VF may be characterized by three factors: (1) factor 1, related to tiredness of voice and voice avoidance, (2) factor 2, related to physical discomfort associated with voicing, and (3) factor 3, related to improvement of symptoms with rest. Conclusion. The VFI is a standardized tool that can identify individuals with probable VF with good reliability, validity, sensitivity, and specificity. Key Words: Vocal fatigue–Questionnaire–Reliability–Validity–Sensitivity–Specificity. INTRODUCTION Vocal fatigue (VF) is a term that has been frequently used in the clinical literature and is a phenomenon many clinicians recognize as widespread, potentially debilitating, and frequently puzzling. Unfortunately, the term has eluded any precise definition that is convergent across authors.1–3 A fundamental challenge has been whether to define fatigue in terms of a set of symptoms that an individual experiences or in terms of physiological changes that arise as a consequence of vocal work. According to a published clinical consensus document, clinicians generally appear to define VF as a feeling of local tiredness and weak voice after a period of voice use.4 More broadly, clinicians have described VF as a global syndrome identified by a series of symptoms during or after speech, including a person’s perception of increased vocal effort, laryngeal discomfort, neck or shoulder tension, throat or neck pain, reduced pitch range, loss of vocal flexibility, reduced vocal projection or power, reduced vocal control, voice loss, increase in symptoms across the speaking day, and typically, improvement of symptoms with rest.1,5–9

Accepted for publication September 9, 2014. From the *Department of Audiology and Speech-Language Pathology, East Tennessee State University, Johnson City, Tennessee; yDepartment of Hearing & Speech Sciences, Vanderbilt Bill Wilkerson Center, Vanderbilt University Medical Center, Nashville, Tennessee; zUniversity of Pittsburgh Voice Center, University of Pittsburgh, Pittsburgh, Pennsylvania; xDepartment of Communication Science and Disorders, University of Pittsburgh, Pittsburgh, Pennsylvania; kDepartment of Otolaryngology, University of Pittsburgh, Pittsburgh, Pennsylvania; {McGowan Institute for Regenerative Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania; and the #Center for the Neural Basis of Cognition, Carnegie Mellon University and University of Pittsburgh, Pittsburgh, Pennsylvania. Address correspondence and reprint requests to Chayadevie Nanjundeswaran, Department of Audiology and Speech-Language Pathology, College of Clinical and Rehabilitative Health Sciences, East Tennessee State University, Lamb Hall, Box 70643, Johnson City, TN 37614. E-mail: [email protected] Journal of Voice, Vol. 29, No. 4, pp. 433-440 0892-1997/$36.00 Ó 2015 The Voice Foundation http://dx.doi.org/10.1016/j.jvoice.2014.09.012

Vilkman10 defined VF more simply as a self-perceived condition involving negative sensations associated with voicing. The self-report approach is appealing considering that VF appears to be a complex clinical entity that is variable in its presentations and presumably mechanisms across individuals. Unfortunately, currently the self-report approach suffers from an agreed-on definition of this entity, and accordingly, no method exists that reliably and validly identifies individuals with the condition. An alternate approach has focused on perceptual, acoustic, or physiological consequences of prolonged voice use—which are assumed to reflect VF (eg, Refs. 3,11). The appeal of this approach is its potential to uncover mechanisms underlying putative fatigue and its ability to extricate itself from thornier self-report issues. The downside is that in this approach, we encounter a type of conceptual circularity: changes in voice after extended voice use are assumed to reflect VF, but VF is not defined independent of those changes. An approach converging self-report and external observations was taken by McCabe and Titze,12 whose proposals around VF built on notions in the physiology and kinesiology literature.13 In that literature, fatigue is defined as both a selfperceived and a behaviorally observed phenomenon linked to prolonged physical activity. Accordingly, McCabe and Titze12 defined fatigue as a ‘‘progressive increase in [presumably self-reported] phonatory effort accompanied by a progressive decrease in phonatory capabilities.’’ Finally, Solomon1 incorporated the concept of rest in her characterization of VF, proposing that such fatigue could be defined as a ‘‘perception by the voice user, manifested primarily as a sense of increased vocal effort that increases over time with voice use, and subsides with voice rest.’’ Despite the lack of broad-based consensus on how VF can or should be defined (eg, Refs. 1,2), research on this phenomenon

434

Journal of Voice, Vol. 29, No. 4, 2015

has proceeded nonetheless in an attempt to understand its underlying mechanisms and treatment. Various research strategies have been used, including (1) the study of individuals with and without complaints of VF6,14–16; (2) the study of voice changes that occur after a vocal workload3,17– 22 ; and (3) the treatment of self-reported VF.12,23–25 Outcomes from these studies generally have yielded mixed and inconclusive results, further complicating our ability to make inferences about VF and its mechanisms. Variability in the results can be attributed to (1) the lack of a clear and consistent definition of VF, and thus an inability to identify individuals with the clinical condition, (2) the study of VF in healthy individuals as opposed to individuals with the clinical condition, (3) variations in tasks across studies of VF, which have ranged from attempting to induce it with prolonged reading versus loud phonation, and (4) the use of measures that may not have sufficient sensitivity to detect VF, however defined (for detailed review, see Ref. 1). In sum, the gaps are clear. If an ultimate goal is to identify and understand mechanisms in VF and thus develop rational treatments for it, a first requirement is to have the ability to reliably identify people with this condition. However, to reliably identify such individuals, a basic working definition of VF is needed. The present study aligns with the view that clinically, VF is at bottom a self-perceived symptom set.1 Thus, a psychometrically validated self-report tool is needed to reliably identify people with VF, ultimately with the future goal of studying its underly-

ing mechanisms and treatments. The purpose of the present study was to address this gap and specifically to generate a cohesive and consensus description of primary VF symptoms and to develop a self-report tool that can reliably and validly identify and quantify symptoms of VF in humans. A third goal, which emerged somewhat fortuitously in the course of data collection, was to characterize component aspects of chronic VF. METHODS AND RESULTS Overview Two sites including the University of Pittsburgh Voice Center (UPVC) and Vanderbilt Voice Center (VVC) participated in the development and validation of the Vocal Fatigue Index (VFI). Institutional review board approval was obtained at both sites before initiating any formal work on the project. Development and validation of the index were carried out in three stages. In stage 1, index questions were generated by clinicians to create a beta (testing) version of the VFI (version 1). In stage 2, the beta version was administered to 200 patients presenting to the two voice clinics. Based on the results, questions having poor item-to-total correlations were removed from the questionnaire to create a second version of the VFI (version 2), and factor analysis was completed. In stage 3, the VFI-2 was administered to 105 patients and 70 vocally healthy individuals at both sites, and the tool’s psychometric properties were assessed. A flowchart of the methodology including development

Stage 1 : Survey Development UPVC

Stage 2 : Development of Scale

Site 1 UPVC N=100

Site 2 VVC N=100

Scale Reliability Analysis

Stage 3 Validation of Index

Test-retest reliability

Site 1 UPVC = 65

Construct validity

Site 2 VVC=40

Site 1 UPVC N-30

Site 2 VVC N=40

FIGURE 1. A flowchart of the development and validation protocol.

Chayadevie Nanjundeswaran, et al

435

Vocal Fatigue Index

and validation of the index is shown in Figure 1. Of note, procedures for the development and testing of the VFI replicated those used in the development of the Voice Handicap Index,26 an index of self-perceived vocal handicap that is widely used in the clinical voice community worldwide. Stage 1—index generation Participants, procedures, and results. A team of four experienced laryngologists and six voice-specialized speech-language pathologists (SLPs) at UPVC met as a group and generated 21 questions they considered to reflect the clinical entity, ‘‘VF.’’ Two standard questions were posed to the panel: (1) How would you describe the phenomenon of ‘‘vocal fatigue’’ in patients who have it? and (2) What are the examples of complaints that people with ‘‘vocal fatigue’’ bring to the clinic? In other words, the panel was asked to generate questions that, based on clinical experience, reflected symptoms patients typically report when presenting with a condition ultimately considered to involve VF. The panel met once for a duration of 2 hours during their weekly grand rounds. The SLPs ended up generating a significant portion of the items for the index. There was considerable discussion regarding the term ‘‘vocal effort’’ and whether it could be used interchangeably with ‘‘vocal fatigue.’’ After the discussion, the general consensus was that the terms were interchangeable. In fact, vocal effort is one example of the variability in how VF is defined in the literature, which further supports the need for this index. Overall, for the 21 items generated, there was a general consensus among the SLPs and laryngologists. After the initial generation of the 21 questions, the questions were circulated among the principal investigator and two other coauthors (K.V.A. and J.G-.S.) for wording and fine-tuning of the questions. The set of 21 questions generated in this way ended up targeting key symptoms, such as tiredness of voice, effort to produce voice, soreness associated with voicing, discomfort and pain in the throat, recovery with voice rest, and so on. (Appendix A, VFI beta version 1). To our knowledge, this is the first index related to vocal symptoms specifically including the effect of rest on voice (improvement or lack of improvement in vocal symptoms). After the generation of this question set at UPVC, the coauthor from VVC (B.H.J.) reviewed it further for wording and appropriateness. No further changes were made to the beta version of the VFI-1. The rationale for using experts in item generation instead of patient groups was primarily based on the rationale for the index. The VFI was developed with the intent to reliably and consistently identify individuals with VF. During the course of the development of the VFI, we considered using not only expert opinion but also patient focus groups to generate items for initial consideration as indicated in a review article by Branski et al.27 However, we quickly encountered a conceptual problem. Currently, widespread consensus lacks about how VF should be defined even by experts—indeed, such lack being one of the primary motivations for conducting the study in the first place. Further complicating the issue and more important was our collective perception that although patients may experience what clinicians call ‘‘vocal fatigue,’’ this vocabulary is unfamiliar to many patients, who instead may use a series of other terms to describe their condition. In sum, patient focus groups were not

used in the generation of items for two reasons: (1) there is currently no standard method to identify patients with ‘‘vocal fatigue’’ (the point of the present study) and (2) patients who many clinicians might consider have vocal ‘‘fatigue’’ are unfamiliar with the vocabulary. The authors acknowledge the importance of focus groups in item generation as indicated in the review by Branski et al.27 However, for the stated reasons, focus groups were not used in the generation of items for the VFI. Stage 2—index development and initial testing Participants. One hundred consenting patients (mean age, 51.76; standard deviation [SD], 19.54; 42 males and 58 females, and mean age, 50.94; SD, 16.01; 31 males and 69 females, respectively) who presented to the UPVC with voice complaints between October 2008 and January 2009, and 100 consenting patients presenting to the VVC between October 2008 and May 2009, completed the preliminary ‘‘beta’’ version of the VFI (VFI version 1). Individuals who came to the clinic for repeat BOTOX injections were excluded. Patients in this category were excluded for several reasons. First, patients of this type present with highly variable symptoms across different time points and would thus introduce unwanted variability in the data. Second, both of these referral centers hold BOTOX clinics once weekly, seeing about 12–20 patients during each clinic day. As such, the inclusion of these patients would have inflated their presence in the data set, unbalancing it with respect to a more representative sample. At both sites, participating patients were diagnosed with a broad range of pathologies affecting voice (Table 1). In

TABLE 1. Primary Diagnosis for Participating Patients at the Two Sites During the Development Phase of the Index—Stage 2 Diagnosis Atrophy Membranous lesion MTD/dysphonia Granuloma Paralysis Scar SD Laryngitis Cancer Tremor RRP Leukoplakia Immobility Paresis LPR Vocal fold hemorrhage Cricoid chondroma Unobtainable diagnosis Total

UPVC

VVC

20 22 15 1 9 5 2 3 3 1 2 5 4 2 1 1 1 3 97

9 24 33 2 9 1 14 1 1 1 1 4 0 0 0 0 0 0 100

Abbreviations: UPVC, university of pittsburgh voice center; VVC, vanderbilt voice center; MTD, muscle tension dysphonia; SD, spasmodic dysphonia; RRP, recurrent respiratory papilloma; LPR, laryngopharyngeal reflux.

436 Table 1, only primary diagnoses are indicated. Diagnostic information from three patients at the UPVC site was not obtained from their records during data collection. Data from all participants, including those with missing diagnostic data, were included in analyses. Three factors justified the query of ‘‘all comers’’ to the voice centers, as opposed to targeting only patients presenting with primary VF. First, there was the issue of circularity. How could one know which patients indeed had primary VF if the purpose of the index was to identify people with VF? In light of this problem, the net was cast wide to capture at least some individuals with VF. Second and equally important, a clinical impression is that many people presenting to voice clinics have complaints of VF at some level, regardless of their diagnosis of origin. Thus, presumably a fairly large proportion of patients completing the questionnaire would indeed have VF as reflected by their responses to questionnaire items. Third and most importantly, questionnaire items targeted VF, specifically. If some individuals had a voice problem but no complaints of VF, group results for patient respondents would underestimate rather than overestimate the presence and severity of VF in this cohort. However, stage 3 of the study, which compared test results of patients with those of vocally healthy individuals, would reveal whether index results would survive this dilution by examining the test’s sensitivity and specificity in its ability to distinguish patients with and without voice problems, while at the same time providing some window on the proportion of patients with and without VF, assuming face validity for questionnaire items. Procedures. Instructions were provided verbally to each subject as follows: ‘‘These are some symptoms usually associated with voice problems. Circle the response that indicates how frequently you experience the same symptoms.’’ All subjects were asked to read each item silently, and circle one of five responses on a five-point Likert scale. The scale was anchored by ‘‘never—0’’ and ‘‘always—4’’ as its end points. ‘‘Almost never’’ corresponded to a score of 1, ‘‘sometimes’’ was scored 2, and ‘‘almost always’’ was scored 3. Results. Data from all 200 subjects were analyzed for internal consistency and scale reliability using the Cronbach alpha coefficient. Based on preliminary analysis, two items—‘‘I run out of air when I talk’’ and ‘‘My shortness of breath with talking decreases after a period of rest’’—were removed because their item-to-total correlation was poor (

Vocal Fatigue Index (VFI): Development and Validation.

To develop a psychometrically sound self-report questionnaire, the Vocal Fatigue Index (VFI), to help identify individuals with vocal fatigue (VF) and...
185KB Sizes 1 Downloads 10 Views