Construct validity and reliability of the rivermead behavioural memory test in the Turkish population

Primary objective: To adapt the Rivermead Behavioural Memory Test (RBMT) for the Turkish adult population and to test its reliability and validity among acquired brain injury patients. Research design: Descriptive study of RMBT's normative scores, reliability and validity. Methods and procedures: One hundred and seventy-one acquired brain injury patients undergoing inpatient rehabilitation in a university hospital and 266 normal adults were assessed by RBMT. Reliability was tested by internal consistency and person separation index (PSI), internal construct validity by Rasch analysis and external construct validity by associations with physical and cognitive disability. Results: Reliability was good with Cronbach's = 0.84 and PSI = 0.85 at admission. Fit to the Rasch model showed one item (faces) to deviate from model expectation. After its removal, internal construct validity was confirmed (mean item fit = -0.394, SD = 0.796; mean person fit = -0.228, SD = 0.681). External construct validity was supported by expected correlations. Conclusion: The original 12-item version of RBMT adapted for use in acquired brain injury in Turkey has been shown to lack internal construct validity. An 11-item version (without 'faces') gives a valid unidimensional summed score and can be transformed to a linear metric for use in studies where change scores are required.