Theoretical framework and methodological development of common subjective health outcome measures in osteoarthritis: a critical review

Subjective measures involving clinician ratings or patient self-assessments have become recognised as an important tool for the assessment of health outcome. The value of a health outcome measure is usually assessed by a psychometric evaluation of its reliability, validity and responsiveness. However, psychometric testing involves an accumulation of evidence and has recognised limitations. It has been suggested that an evaluation of how well a measure has been developed would be a useful additional criteria in assessing the value of a measure. This paper explored the theoretical background and methodological development of subjective health status measures commonly used in osteoarthritis research. Fourteen subjective health outcome measures commonly used in osteoarthritis research were examined. Each measure was explored on the basis of their i) theoretical framework (was there a definition of what was being assessed and was it part of a theoretical model?) and ii) methodological development (what was the scaling strategy, how were the items generated and reduced, what was the response format and what was the scoring method?). Only the AIMS, SF-36 and WHOQOL defined what they were assessing (i.e. the construct of interest) and no measure assessed was part of a theoretical model. None of the clinician report measures appeared to have implemented a scaling procedure or described the rationale for the items selected or scoring system. Of the patient self-report measures, the AIMS, MPQ, OXFORD, SF-36, WHOQOL and WOMAC appeared to follow a standard psychometric scaling method. The DRP and EuroQol used alternative scaling methods. The review highlighted the general lack of theoretical framework for both clinician report and patient self-report measures. This review also drew attention to the wide variation in the methodological development of commonly used measures in OA. While, in general the patient self-report measures had good methodological development, the clinician report measures appeared less well developed. It would be of value if new measures defined the construct of interest and, that the construct, be part of theoretical model. By ensuring measures are both theoretically and empirically valid then improvements in subjective health outcome measures should be possible.