Improving Measurement Efficiency of the Voice Handicap Index-10 With Item Response Theory.

Abstract

To assess: (1) the Voice Handicap-10 (VHI) instrument with item response theory (IRT) and determine which items convey the most information, (2) the extent to which voice handicap can be measured with a subset of items for efficient information intake, and (3) whether a proposed shorter scale has differing discriminatory ability for key diagnoses made by fellowship-trained laryngologists, as compared to the 10-item instrument. Prospectively collected data from 3640 patients who completed the VHI-10 questionnaire during evaluation at a tertiary care otolaryngology clinic were utilized. IRT analysis provided discrimination and location parameters associated with each item. Residual item correlations were also assessed to assess redundant information. Based on these results, two 5-item subsets were further evaluated using item information function curves. Areas under receiver-operator characteristic curves (ROC-AUC) were also calculated to evaluate the discriminatory ability for findings from videolaryngoscopy/stroboscopy. Item discrimination parameter estimates ranged from 1.55 to 4.68, with higher values indicating more information. Residual item correlations were determined within item pairs, and location parameters were calculated. Based on these data, potential 5-item subsets were proposed, which preserved the capacity to reflect underlying voice handicap assessment. ROC-AUC analyses suggested no significant difference between the 5-item subset and 10-item instrument with regard to their ability to discriminate whether findings such as vocal fold paralysis were present. IRT data distinguished items for a proposed 5-item subset of the 10-item instrument, with the goal of reflecting the latent trait without significant loss of discrimination and precision. Level 2-Prospective validation study with reference standard.