Use of Rasch analysis in the evaluation of the oropharyngeal mucositis quality of life scale
Cheng, K., Lee, J., Leung, S., Liang, R., Tai, J., Yeung, R. & Thompson, DR. (2011). Use of Rasch analysis in the evaluation of the oropharyngeal mucositis quality of life scale. Nursing Research (print version),60(4), 256-263. Retrieved from https://doi.org/10.1097/NNR.0b013e318221f731
Background: Oropharyngeal mucositis (OM) is a significant clinical problem causing profound impairment of health-related quality of life (HQoL) for patients undergoing cancer therapy. The Oropharyngeal Mucositis-Specific Health-Related Quality of Life Measure (OMQoL) was developed using classical test theory to measure the self-perceived HQoL of patients with mucositis.
Objectives: The aim of this study was to analyze the OMQoL according to the Rasch model and, on the basis of results, determine whether improvements could be made.
Method: A multicenter approach was used, and 210 patients treated with stomatotoxic chemotherapy (36%), high-dose myeloablative chemotherapy +/- total body irradiation (10%), or head and neck irradiation +/- chemotherapy (54%) completed the OMQoL. The Partial Credit Model of Rasch analysis was applied to evaluate the 31-item OMQoL using WINSTEPS and R software. Unidimensionality (measurement of a single construct), item fit, response category performance, person separation reliability, targeting of item difficulty to person ability, and differential item functioning (DIF) were examined.
Results: Of 31 items, 5 were removed due to misfit; the OMQoL was reduced to 26 items with acceptable information weighted fit/outlier-sensitive fit indices (within 0.7-1.3) and eigenvalue units (<=2.0), confirming the unidimensionality of the reduced OMQoL. The OMQoL and its four subscales showed ordered category thresholds, and the person separation reliability was high (person separation index >0.2 with reliability >.8). Nevertheless, some of the items in the OMQoL might not be targeted effectively to patients with low levels of OM. Significant uniform and nonuniform DIFs were not found for gender (uniform DIF, p = .26; nonuniform DIF, p= .24) and age (uniform DIF, p = .95; nonuniform DIF, p = .65).
Discussion: Rasch analysis reveals that the reduced 26-item OMQoL is unidimensional and is adequate to measure HQoL for patients with OM regardless of gender and age group. This improved version can provide a common platform for nurses to use in their assessment, caring, and treatment of patients with OM.
Nursing Research Institute