A questionable factor structure of the multidimensional fatigue inventory in the general Dutch population

J Clin Epidemiol. 2021 Sep;137:266-276. doi: 10.1016/j.jclinepi.2021.05.005. Epub 2021 May 14.


Objective: One of the most commonly used tools to measure fatigue is the Multidimensional Fatigue Inventory (MFI). Studies into the scale structure of the MFI show discrepant findings. The objective of this study was to investigate the scale structure of the MFI in the general Dutch population.

Study design and setting: Using data from a Dutch probability-based internet panel (n = 2512), the original 5-factor model, a 4-factor, and a 5- and 4-bifactor model of the MFI were tested with confirmatory factor analyses. Additional models were investigated using exploratory factor analysis.

Results: Results neither confirmed a 5-factor (RMSEA = 0.120, CFI = 0.933, TLI = 0.920) nor a 4-factor model (RMSEA = 0.122, CFI = 0.928, TLI = 0.917). The two bi-factor models also showed a poor fit (bi-4-factor: RMSEA = 0.151, CFI = 0.895, TLI = 0.873; bi-5-factor: RMSEA = 0.153, CFI = 0.894, TLI = 0.871). Exploratory factor analysis did not support an alternative model, but seemed to show robustness in the loading of the original general fatigue items.

Conclusion: Our results did not provide empirical support for a four or five (bi-)factor structure of the MFI, nor for an alternative model. The most reliable scale of the MFI seems to be the general fatigue scale that could be used as a general indicator of fatigue.

Keywords: Factor analysis; Fatigue; Measurement model; Psychometrics.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Diagnostic Tests, Routine / methods
  • Factor Analysis, Statistical
  • Fatigue / diagnosis*
  • Humans
  • Netherlands