Applying bifactor modelling to improve the clinical interpretive values of Functional Independence Measure in adults with acquired brain injury.
Gao, F., Foster, M., Newcombe, P., Geraghty, T
To apply a modern robust approach, bifactor modeling, to critically examine psychometric properties of Functional Independence Measure (FIM) in adults with acquired brain injury and to propose a solution to improve the clinical interpretive values of the FIM to inform policy and clinical practice.
The data came from a state-wide specialist in-patient brain injury rehabilitation service in Queensland, Australia for adults with acquired brain injury and discharged between 2012 and 2017. The sample included 457 people. Three measurement models (unidimensional, correlated first-order and bifactor) for FIM were tested using confirmatory factor analysis with structural equation modeling. Then, model-based reliability and incremental validity were assessed.
The bifactor model best fit the data. When operationalized as latent factors under structural equation modeling framework, general care burden had a large predictive effect, while Motor and Cognitive showed medium and small predictive effects respectively on rehabilitation length of stay.
The total score of FIM was a reliable measure of general care burden, while the subscale scores were not. A solution is to apply a bifactor modeling approach based on structural equation modeling to disentangle the unique variance attributable to Motor and Cognitive factors. In the structural equation modeling framework, the FIM demonstrated good incremental validity to inform policy and clinical practice. Implications for rehabilitation Clinicians and researchers can confidently use FIM total score in adults with acquired brain injury. The current study proposed an alternative solution to the poor reliability of Motor and Cognitive scores, that is, by applying a bifactor modeling approach, the unique contributions of the Motor and Cognitive factors can be examined. The current study has demonstrated the strengths of bifactor modeling in the robust validation and interpretation of FIM to better inform clinical practice and policy decision-making. The current study has the potential to make an important contribution to enhance more equitable decision-making in the areas of national benchmarking of rehabilitation outcomes and other program eligibility criteria and funding allocation.