Differential functioning of mini-mental test items according to disease.

INTRODUCTION Comparing the height of males and females would be impossible if the measuring device did not have the same properties for both populations. In a similar way, the cognitive level of diverse groups of patients should not be compared if the test has different measurement properties for these groups. Lack of Differential Item Functioning (DIF) is a condition for measurement invariance between populations. MATERIAL AND METHODS The most internationally used screening test for dementia, the MMSE (or Mini-mental State Examination), has been analysed using an advanced psychometric technique, the Rasch Model. The objective was to determine the invariance of mini-mental measurements from diverse groups: Parkinson's disease patients, Alzheimer's type dementia and normal subjects. The hypothesis was that the scores would not show DIF against any of these groups. The total sample was composed of 400 subjects. RESULTS Significant differences between groups were found. However, the quantitative comparison only makes sense if no evidence against measurement invariance was found: given the kind of items showing DIF against Parkinson's disease patients, the MMSE seems to underestimate the cognitive level of these patients. CONCLUSIONS Despite the extended use of this test, 11 items out of 30 show DIF and consequently score comparisons between groups are not justified.