Modern psychometrics and the design and application of patient-reported outcome measures

Twiss, James (2015) Modern psychometrics and the design and application of patient-reported outcome measures. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (10MB) | Preview

Abstract

Creating accurate, high quality measurement with patient-reported outcomes (PRO) is a key challenge for developers. It is often the case that PRO measures fail to clearly define the constructs that they are intended to measure. Consequently, they fail to provide measurement that is valid and meaningful. Classical test theory has been applied in the development of most outcome measures currently in use. Such psychometric approaches to PRO measure development are being superseded by more powerful item response theory (IRT) methods. The Rasch model is the one parameter form of IRT that embodies fundamental measurement requirements. Scales that produce data fitting the Rasch model provide interval level measurement, improving their power and discrimination. The thesis argues that it is a combination of clear construct definition and application of Rasch analysis that lead to improved measurement. The aims of the thesis are to i) describe approaches to construct definition and psychometric measurement ii) evaluate my own research in relation to these approaches iii) critically assess the contribution of the research to the field. The thesis considers ten articles relating to the development and application of PROs. The articles in the thesis cover the following topics: New PRO scale developments. Three articles describe developments of new measures that are based on a clearly defined construct and apply Rasch analysis in their development. Application of PRO scales in international research. Two articles describe the adaptation of PRO scales into several additional languages. Such adaptations increase the value of the measures to international research. In addition, a minimal important difference (MID) study is described in one article. The MID estimations generated assist the interpretation of scores and sample size determination for future studies. Evaluation of existing PRO scales. Three studies describing the evaluation of PRO measures are discussed. Weaknesses were identified in each of the scales. The Mood Disorder Questionnaire, a screening tool for bipolar disorder, was found to screen patients more effectively when the symptoms section of the scale was used without the other sections of the questionnaire. The Dermatology Life Quality Index (DLQI) and the SF-36 had several measurement limitations and were not based on clearly defined constructs. Co-calibration of disease-specific PRO scales. A new method for combining scores from two disease-specific PROs using Rasch analysis is discussed. This method offers a means of combining PRO data from patients with different diseases that complete different diseasespecific measures. This approach was possible as both measures were developed based on the same clearly defined construct and both produced data fitting the Rasch model. The research makes a number of important contributions to the improvement of PRO measurement. The studies show that clear construct definition and application of Rasch analysis are central to improving the science. More work is necessary, particularly to understand in greater detail the needs-based model of quality of life that has been applied in the new measure development described in the thesis.