Clinimetric and psychometric strategies for development of a health measurement scale.
Academic Article
Overview
abstract
Clinimetrics and psychometrics, two accepted methods for developing multiitem health measurement scales, have fundamentally different aims and methods that have seldom been compared and never prospectively. The purpose of this study was to determine whether these two methodologies provided comparable scales in the development of an upper extremity disability measure. Psychometric analysis involved field testing a 70-item questionnaire on 407 patients. Equidiscriminatory item total correlation (EITC) was used to select the top 30 items. Clinimetric testing used the mean importance and severity ratings of the 70 items by 76 patients to select the top 30 items. Clinimetric and psychometric analyses were performed independently. Cronbach's alpha was 0.97 for the top 30 items selected by EITC and 0.96 for the items selected based on patient's ratings. The two scales (after clinician modification to improve face validity) shared 16 items in common (P=0.10). The intraclass correlation coefficient of the patient scores on the two 30-item scales was 0.93 before clinician input and 0.97 after. The mean (and standard deviation) difference between scales was 9.1 (8.8) before and 1.7 (5.2) after clinician input. A scale developed with a clinimetric strategy can measure a complex (so-called heterogeneous) clinical phenomenon (thought to be composed of several patient attributes) but still fulfill psychometric criteria for "homogeneity." Thus, these strategies for the development of health measurement scales, which have been considered potentially opposite or conflicting, may be complementary.
