A critique of Rasch analysis using the Dyspnoea-12 as an illustrative example


  • Janelle Yorke,

    1. Janelle Yorke PhD MRes RN Senior Lecturer School of Nursing and Midwifery, University of Salford, Greater Manchester, UK
    Search for more papers by this author
  • Mike Horton,

    1. Mike Horton BSc Research Assistant Psychometric Lab for Health Sciences, Faculty of Medicine and Health, Leeds General Infirmary, University of Leeds, UK
    Search for more papers by this author
  • Paul W. Jones

    1. Paul W. Jones MBBS PhD FRCP Professor Division of Cardiac and Vascular Science, St George's, University of London, Cranmer Terrace, UK
    Search for more papers by this author

J. Yorke: e-mail: j.yorke@salford.ac.uk


yorke j., horton m. & jones p. w. (2012) A critique of Rasch analysis using the Dyspnoea-12 as an illustrative example. Journal of Advanced Nursing68(1), 191–198.


Aim.  This paper is a report of a study of the application and critique of Rasch analysis to the development of the Dyspnoea-12 questionnaire; an instrument that measures breathlessness severity.

Background.  The development of questionnaires has traditionally involved application of classical test theory. Rasch analysis has gained international momentum as a robust application of ‘modern’ psychometric testing for the development of new instruments and the refinement of existing ones.

Method.  A total of 358 patients [mean age 62 (sd = 8); chronic obstructive pulmonary disease = 123, interstitial lung disease = 129, heart failure = 106], responded to an initial list of 81 items (between May 2006 and February 2008). Hierarchical modelling reduced the list to 34 items. Rasch analysis was used to inform decisions about further item removal and fit to the unidimensional model. Rasch analysis included tests of item response appropriateness, item residual, differential item functioning (including gender and diagnosis) and unidimensionality.

Results.  Twenty-two items failed to reach the requirements of the Rasch model and were removed. Reasons included high residuals and item bias associated with gender and diagnosis. The 12-items conformed to the Rasch unidimensional parameters (number of statistically significant t-tests 6·7%; confidence interval: 4·4–9·0%).

Conclusion.  We have presented the steps involved in reducing and refining a large item-set by identifying those items which possessed the most reliable measurement properties. The Dyspnoea-12 is reliable and simple to use and should find utility in both practice and research settings. We recommend that nurses consider Rasch analysis in the development of health-related questionnaires.