Use of item response theory to develop a shortened version of the EORTC QLQ-BR23 scales

Abstract

It is important that questionnaires are as short as possible while still capturing the scope of problems relevant in an effective and reliable manner, to minimize the response burden. The purpose of our study was to develop a shortened version of the EORTC QLQ-BR23 for using in breast cancer survivors. Our data come from 10794 breast cancer survivors who completed the EORTC QLQ-BR23. Two-thirds of the sample was randomly selected from the original sample for development, and the remaining was used for validation. Item response theory methods were applied to shorten scales. The graded response model of Samejima was used to fit the item responses. The shortened scale was evaluated with the validation set by examining the mean difference, the proportion of respondents correctly predicted, correlation and weighted kappa between the shortened form and the original observed scores. Results reveal that a three-item BRBI, a four-item BRST, a three-item BRBS and a two-item BRAS forecast the scores on the original scales with wonderful consistency and are alike in measurement precision with no loss or only little loss in detecting group differences. Prospective validation on new diagnosed breast cancer patients and with poor QOL is needed.