Assessing the consistency of psychometric properties of the HRQoL scales within the EORTC QLQ-C30 across populations by means of the Mokken Scaling Model.

This paper demonstrates how the Mokken Scaling Model and other statistical tools may be useful in assessing the consistency of psychometric properties of health-related quality of life (HRQoL) scales across various populations. The main focus is the psychometric performance of the scales proposed for the EORTC QLQ-C30 in seven patient groups totalling more than 2,000 cancer patients. All scales performed satisfactorily in the total sample with the exception of the role functioning and cognitive functioning scales, which failed in terms of reliability and item discriminant validity. The descriptive statistics for the scales show that several of them, particularly those that build upon only two items, have discrimination problems at the extremes, visible in the high percentages at the maximum or the minimum observed values. The scalability analysis in the subsamples showed that the essential assumption in the Mokken Scaling Model of equal item step order does not hold for the cognitive functioning, emotional functioning and physical functioning scales. We conclude that the Mokken Scaling Model is well suited to the purpose of examining the generalizability of HRQoL scales across subpopulations although a global statistical test of the fit of the measurement model is not available.