Tackling measurement issues in health predictors and outcomes using item response theory

• Main Author: Jackson, Jeanette
• Published: University of Aberdeen 2008
• Subjects: 616.89; Item response theory
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.521322

The Functional Limitation Profile (FLP), the Hospital Anxiety and Depression Scale (HADS) and the Recovery Locus of Control scale (RLOC) are three well established and useful measures used in Health Psychology. However, the reliable and valid measurement of these health predictors and outcomes has associated problems. The present thesis tackles measurement issues in all three instruments using item response theory (IRT). The Scientific Advisory Committee of the Medical Outcomes Trust has suggested the methodological and theoretical rationale for the conceptual and measurement model of available measurement instruments should be reported. The introduction chapter provides theoretical background in order to understand activity limitations and participation restrictions as behaviours affected by a certain health condition, as well as by thoughts and feelings. Within this theoretical framework, the present thesis investigates the measurement of mood using the HADS and functional limitations using the FLP in three different health conditions: (1) stroke patients, (2) patients with myocardial infarction, and (3) patients who underwent joint replacement surgery. The measurement of perceived personal control beliefs using the RLOC scale, and the relationship between control cognitions, mood and functional limitations were examined in stroke patients since all three measures were available for secondary analysis in this sample. The main findings are that (1) highly sensitive FLP items measure precisely different levels of disability and handicap, (2) removing 2 HADS items results in precise measurements of different levels of anxiety and depression, and (3) internal but not external perceived personal control beliefs measured sensitively different levels of the underlying construct.