Validation of the Voice Handicap Index Using Rasch Analysis
Introduction
The Voice Handicap Index (VHI) was introduced as a tool for measuring the psychosocial consequences of voice disorders1 and is now used in clinical research to evaluate the effectiveness of voice therapy. The VHI is a self-administered instrument consisting of 30 statements on voice-related dysfunction. The original version consists of three sub-domains measuring emotional-, physical-, and functional-related problems. The instrument has been translated and validated into many different languages including German, Taiwanese, Spanish, and Portuguese. In 2001, a Dutch translation was validated.2
Although the VHI is already widely used, some methodological problems exist.
First, the three-factor structure is not consistent across populations. For example, Nawka et al3 found four different domains in a German-speaking sample, whereas Wilson et al4 found only two domains using the original English version. These differences may, in part, result from the different factor rotation methods used, such as Varimax-rotation or oblique-rotation. However, the results of traditional correlation-based analysis are sample dependent, and the resulting factor structure, therefore, may vary from sample to sample.5 Second, despite the multi-dimensionality of the VHI, the domains are summed to arrive at a total score. However, a strict requirement for computing sum scores, ie, the additivity of test items, is the uni-dimensionality of the test.6 Third, the original psychometric evaluation of the VHI was performed in only 65 subjects with voice disorders and focused on the classic correlation-based evaluation of reliability and validity, which is, as noted before, highly sample dependent. The psychometric properties of the VHI needed to be assessed in a larger and more diverse sample of subjects with voice disorders. Such a group would provide the opportunity to apply modern methods of scale validation, such as Rasch analysis, that produces item and scale statistics that are less sample dependent.5, 7
Rasch analysis uses a probability model that gives directions of which test items to combine, and which test items to discard, to arrive at a truly uni-dimensional and thus additive (interval) scale.8 Unlike the classic correlation-based analysis, which focuses on overall summed scores, Rasch analysis focuses on the performance of items. Using the scores of subjects to the test items, it estimates item difficulty and transforms ordinal Likert-type rating scale scores into interval level measures. The resulting equal interval units of measurements are called “logits” or log-odds units. Rasch analysis has been applied for the validation and modification of existing scales, or for the validation of new scales in diverse areas of medicine including psychiatry,9 ophthalmology,10 medical rehabilitation,11 neurology,12 and overall health-related quality of life.13
The goal of this study was to use Rasch analysis to (1) provide estimates of VHI item “difficulties” that can be used to transform ordinal VHI scores of patients into interval level severity measures and (2) to re-examine the VHI-questionnaire to obtain more definite results regarding its dimension structure.
Section snippets
Patients
The VHI data used in this study were from 530 outpatients referred to the Departments of Otolaryngoly from two larger university teaching hospitals in the Netherlands in 2003 and 2004. All patients had voice-related complaints and were referred by general practitioners and ear, nose, and throat physicians for diagnostic examination. Phoniatric examination was performed by using stroboscopy to determine the possible cause of the voice problem. The subjects filled out the VHI before the
Results
The characteristics of the sample (N = 530) are shown in Table 1. Thirty-eight percent of the patients were men, and patients had a mean age (±SD) of 51.7 (±17.8) years. For 244 patients, the diagnosis was unknown. In the patient group with known diagnosis, 35.0% of the patients had functional dysphonia followed by laryngitis/vocal fold edema (21.0%), nodules (11.9%), and unilateral paralysis (11.5%). Sum-scores on the VHI ranged from 0 to 112 points (mean 39.7, ±22.9). In a subsample of
Discussion
Our sample (N = 530) consisted of a diversity of laryngeal pathologies with a wide range of dysphonic severity levels. The mean initial sum scores on the original 30-item VHI and the acoustic parameters showed that our patients had substantial voice problems.
Previous research focused on the classic correlation-based analysis to explore the dimensionality of the VHI, which is highly sample dependent. Furthermore, classic techniques result in qualitative, ordinal, test scores rather than
References (21)
- et al.
Rating scale analysis of the Berg Balance Scale
Arch Phys Med Rehabil
(2004) - et al.
The Voice Handicap Index (VHI): development and validation
Am J Speech Lang Pathol
(1997) - et al.
De Voice Handicap Index: het kwantificeren van de psychosociale consequenties van stemstoornissen
Logopedie en Foniatrie
(2001) - et al.
Validation of the German version of the Voice Handicap Index
HNO
(2003) - et al.
The Voice Symptom Scale (VoiSS) and the Vocal Handicap Index (VHI): a comparison of structure and content
Clin Otolaryngol
(2005) - et al.
A procedure for sample-free item analysis
Educational Psychologic Measur
(1969) Additivity in psychological measurement
Applications of Item Response Theory to Practical Testing Problems
(1980)Fundamental measurement for outcome evaluation
Phys Med Rehabil State of the Art Rev
(1997)- et al.
The scoring scheme of the informant questionnaire on cognitive decline in the elderly needs revision: results of Rasch analysis
Dement Geriatr Cogn Disord
(2004)
Cited by (32)
Urica-VV Scale: A New Research Perspective of The Stage of Readiness for Treatment in Patients with Dysphonia
2023, Journal of VoiceCitation Excerpt :This type of analysis has already been suggested in a theoretical article that analyzed the flaws in the construction of self-assessment instruments for the population of individuals with voice disorders; however, little progress has been made in this direction.17,18 There are some instruments in the voice field that apply more contemporary analyses for the validation of self-assessment instruments, such as the proposal for the validation of the Voice Handicap Index (VHI) in the Netherlands based on Rasch analysis19; the validation of the Voice Symptom Scale (VoiSS), American English version, based on the Mokken model20; the validation of the Vocal Fatigue Index (VFI), original version, using the Mokken model21; and the validation of the Danish version of the Voice-Related Quality of Life (V-RQOL) instrument based on Rasch analysis.22 For this purpose, the present study started from guiding questions related to the psychometric evaluation of and application of IRT for the URICA-V scale: Will the number of items and domains/factors in the instrument be the same?
Factor Analysis of the Persian Version of the Voice Disability Coping Questionnaire
2020, Journal of VoiceCitation Excerpt :These results are stronger and higher than the results in all previous versions like Brazilian Portuguese and Persian versions.9,10 These results could be due to the large sample size in this study.14 In order to check the internal consistency of the questionnaire, Cronbach's alpha coefficient was used which was 0.74 for social support, 0.39 for avoidance, 0.54 for passive coping and 0.84 for information seeking.
Cross-cultural Adaption and Validation of the Danish Voice Handicap Index
2019, Journal of Voice