Psychometric Properties of the Multidimensional Fatigue Inventory (MFI-20), Derived From Seven Samples

Context Fatigue is a frequent symptom in patients suffering from chronic diseases. The Multidimensional Fatigue Inventory (MFI-20) is often used to measure fatigue. The aim of this article was to test the scale structure of the questionnaire. Methods The MFI-20 data were obtained from seven samples, including general population samples and samples of patients with different diseases (N between 122 and 1993). Five confirmatory factor analysis (CFA) models were tested for each sample. Results The scale structure postulated by the original test authors could not be confirmed by the CFAs. The inclusion of a method factor which considers the positive versus the negative orientation of the items yielded a better model fit. Cronbach's alpha was acceptable for most of the samples and scales: the total score of the MFI-20 reached alpha coefficients above 0.89. A short form of the MFI-20 which is restricted to the 10 positively oriented items (MFI-10) showed relatively good CFA indices. Conclusion The factorial validity of the MFI-20 is insufficient, an issue which is due, at least in part, to the inclusion of positively and negatively oriented items. Nevertheless, we recommend maintaining the scale structure of the MFI-20 and not searching for alternative structures.