Measuring disability in neurological rehabilitation : psychometric evaluation of two outcome measures
Objectives: To evaluate comprehensively the psychometric properties of the Functional Independence Measure (FIM) and the Functional Independence Measure + Functional Assessment Measure (FIM+FAM), and to compare their performance in stroke and multiple sclerosis (MS) patients and with the Barthel Index. To evaluate the conceptual models of both instruments using item analysis, and determine the feasibility of developing a short-form measure. To compare five methods of evaluating responsiveness. Design: Psychometric study. Subjects: 209 inpatients with a variety of neurological disorders recruited from three neurorehabilitation units in Southeast England. Method: Standard methods were used to evaluate the acceptability, reliability, validity, and responsiveness of the FIM and FIM+FAM. Detailed item analyses were performed including internal consistency, intercorrelations between scales and subscales, item convergent and discriminant validity, and principal components analysis. Item reduction techniques were used to develop a short-form FIM. Five methods were used to evaluate responsiveness: t - statistics, relative efficiency, effect size, standardised response mean, and the responsiveness index. Results: The FIM and FIM+FAM are acceptable, reliable, valid, and responsive measures of disability in neurorehabilitation. However, they demonstrate no psychometric advantage over the Barthel Index, show item redundancy, limited item discriminant validity, and inadequate support for hypothesised subscales. An 8-item short-form FIM is developed that shows similar psychometric performance to the 18-item FIM and 30-item FIM+FAM. Five methods of evaluating responsiveness rank order scales similarly, but generate numerical estimates of different magnitude. Conclusions: Results demonstrate the need for a more systematic and rigorous approach to the development and psychometric evaluation of instruments before their introduction into practice to ensure the accurate measurement of patient-oriented outcomes in health care. This approach includes the development of appropriate conceptual and measurement models, the application of standard item analysis and item reduction techniques during questionnaire development, and comprehensive evaluation of the recommended full range of psychometric properties.