A random sample of German residents aged 16 and older was recruited as part of a broader cross-sectional questionnaire survey between October and December 2021. Since there is no directory available that contains the addresses of all private households or individuals in Germany, the “ADM Sampling System for face-to-face surveys” (18) was used to draw a representative German sample. A market and social research company (USUMA GmbH, Berlin, Germany) performed the subject acquisition and face-to-face interviews with trained interviewers. Overall, 5,901 persons were contacted, 2,526 participated (others did not respond or were ill or otherwise unavailable at the appointed interview date; all reasons for non-participation are listed at the end of the manuscript), and 2,509 completed the interview and the self-report battery. Data for comparison on demographics between participants and non-participants were not available, but the participants are representative regarding gender, age, and regional distribution (16 German federal states). All data were fully anonymized by USUMA prior to data analysis. All participants gave their verbal informed consent in accordance with the Helsinki declaration, which was documented by USUMA, and followed a structured face-to-face interview to collect the statistical and sociodemographic data. Participants were ≥ 16 years and are legally competent in Germany to participate in a study without the need for others' consent. The self-report questionnaires, such as the MFI-20, were then completed via paper and pencil independently by the subjects themselves, so as to not bias their personal mental and physical health answers. The study was approved by the Ethical Review Committee of the University of Leipzig (298/21-ek).

The Multidimensional Fatigue Inventory (MFI-20) is a 20-item self-report instrument, which consists of five subscales: 1. general fatigue (items 1, 5, 12, 16), 2. physical fatigue (items 2, 8, 14, 20), 3. reduced activity (items 3, 6, 10, 17), 4. reduced motivation (items 4, 9, 15, 18), and 5. mental fatigue (items 7, 11, 13, 19). Each subscale consists of four items with possible answers on a 5-point Likert scale (1 = yes, that is true; 5 = no, this is not true). Higher scores indicate higher levels of fatigue. Convergent validity with a Visual Analog Scale (VAS) assessing fatigue showed significant correlations ranging from 0.77 (general fatigue) to 0.23 (mental fatigue) (15), with the Fatigue Severity Scale ranging from 0.73 (general fatigue) to 0.16 (reduced activity) (19). The previous German validation study argued for using the total score of the 20 items because of the empirically uncertain factorial structure and the strong correlations between the subscales (1).

The Short Form-36 Health Survey (SF-36) (20) was used to provide a convergent validity criterion for the MFI-20. The 36 questions on the SF-36 are meant to reflect eight domains of health, including physical functioning, physical role, pain, general health, vitality, social functioning, emotional role, and mental health (range 0 to 100). Additionally, a Physical Composite Score (PCS) and a Mental Composite Score (MCS) can be calculated. These scores are T-values (mean = 50, SD = 10) in reference to a US norm sample. Especially, the SF-36 domain vitality has been widely used as a fatigue marker (21, 22).

IBM SPSS Statistics version 28 (23) was used for the statistical analyses. Negatively phrased items were inverted prior to further analyses. Because the questionnaire scores were not normally distributed (Shapiro-Wilk tests for all fatigue scales, p < 0.001), median and mode values are reported in addition to means and standard deviations (SD). Spearman's ρ correlations were calculated to investigate the association among the MFI-20 subscales and between those, the SF-36 domains, and age. Mann–Whitney U tests were calculated to compare MFI-20 scores between male and female subjects. Kruskal-Wallis tests with Dunn-Bonferroni post-hoc tests were conducted to compare seven age groups (≤ 24, 25–34, 35–44, 45–54, 55–64, 65–74, and ≥ 75 years, in accordance with other studies using this survey). A confirmatory factor analysis was computed using the lavaan R package for SPSS (24). The norm tables were conducted as percentile ranks. McDonald's omega (ω_t), an alternative measure of internal consistency that does not assume a tau-equivalent but a congeneric model, was computed using the SPSS Macro OMEGA (25).

Table 1 summarizes the sociodemographic characteristics, shows the MFI-20 subscale and total scores, and the SF-36 scores of the total sample. The descending order of the MFI-20 subscales, with general fatigue exhibiting the highest scores and mental fatigue the lowest scores, is comparable to a recent Dutch norm sample and other studies (11). The five MFI-20 subscales, the total score, and age did not follow a normal distribution. The MFI-20 total score had a profound right-tailed skewness (0.82) and a slight flat kurtosis (−0.13). The five subscales showed a strong floor effect, i.e., between 15% (e.g., general fatigue) and 20% (e.g., mental fatigue) of the subjects exhibited the lowest possible score.

Item scores ranged from 1.75 (item 18) to 2.38 (item 20). The selectivity (item-scale correlations) varied between 0.545 (item 9) and 0.814 (item 14), and the homogeneity (inter-item-correlations) varied between 0.362 (item 9 x item 15) and 0.733 (item 8 x item 20). The mean value of missing answers per item was 6 (SD = 2.77) or 0.24%, indicating a highly accepted questionnaire. Item characteristics are shown in Table 2.

Ascertaining reliability, Cronbach's α values for the subscales were 0.87 (General fatigue), 0.90 (Physical fatigue), 0.88 (Reduced activity), 0.79 (Reduced motivation), and 0.83 (Mental fatigue). These values are similar to the developers' 1995 values (15) and follow the identical ordinal ranking found in the first German 2003 validation study (1). The omission of items did not increase any Cronbach's α considerably. McDonald's ω_t for all items was 0.964, indicating high reliability (26, 27).

The convergent validity of the MFI-20 was examined by correlating the MFI-20 subscales/total score with the SF-36 domains and its PCS and MCS. All correlations between the MFI-20 subscales/total score and the SF-36 domains/PCS/MCS were significant (p < 0.001), using Spearman's ρ correlations due to the non-parametric distribution. Correlations between 0.40 and 0.60 can be seen as a medium in convergent validity. As expected, vitality exhibited the highest correlations with three of the five MFI-20 subscales and the total score, highlighted in bold in Table 3. The PCS correlated the most with physical fatigue (−0.72) and the lowest with mental fatigue (−0.41). The MCS correlated the most with the MFI-20 total score (−0.64), followed by general fatigue (−0.63) and mental fatigue (−0.61). Pairwise correlations of the MFI-20 subscales ranged between 0.71 and 0.85 and are slightly lower when controlled for age (refer to Table 3, values labeled ^p). The correlation between the MFI-20 total score and age was r_s = 0.35, with physical fatigue having the highest (r_s = 0.42) and mental fatigue having the lowest (r_s = 0.19) subscale correlation with age (all p < 0.001).

Mann–Whitney U tests showed significantly higher fatigue scores in females for all but one subscale and the total score (all p < 0.001, reduced activity p = 0.04) but not for reduced motivation (p = 0.08). The Kruskal-Wallis test showed significant differences between age groups for the different subscales and the total score: general fatigue [H(2) = 269.969], physical fatigue [H(2) = 466.147], reduced activity [H(2) = 318.654], reduced motivation [H(2) = 273.402], mental fatigue [H(2) = 118.925], and the total score [H(2) = 342.471]; all p < 0.001, df = 6. Following post-hoc tests (Dunn-Bonferroni) revealed which age groups significantly differed from each other (Table 4).

Two-way ANOVAs showed no interaction effect between age (seven age groups) and gender (male and female) on any subscale and the total score. However, this must be interpreted carefully due to non-parametric distributions. A descriptive, insignificant interaction effect of gender and age groups 3–44 and 45–54 years could be observed on all subscales (except mental fatigue) and the total score. Between these two age groups, the scores increase with age for men but decrease with age for women. Figure 1 shows the general fatigue scores to highlight the age and gender structure of the results. Figure 2 illustrates that in the oldest age group, reduced motivation and mental fatigue, two cognitive appraisal domains, show the relatively lowest scores. In the youngest age group (≤ 24 years), mental fatigue exhibited the highest and physical fatigue the lowest scores. In contrast, in the oldest age group (≥ 75 years), mental fatigue exhibited the lowest and physical fatigue the highest scores.

The data did not follow a multivariate normal distribution (28). A confirmatory factor analysis (CFA) for the five-factor structure showed poor model fit, comparative fit index (CFI) = 0.891, Tucker-Lewis index (TLI) = 0.871 (values > 0.95 indicate an acceptable fit), and root mean square error of approximation (RMSEA) = 0.105 (values < 0.08 indicate an acceptable fit). The Kaiser-Meyer-Olkin criterion of sampling adequacy was highly developed (0.972). The Bartlett test on sphericity was significant (p < 0.001, approximate χ² = 39625.349, df = 190). An exploratory factor analysis (EFA using principal component analysis) showed two disparate components with an eigenvalue (λ) > 1. The first component with λ = 11.872 explained 59.36% of the variance, and the second component with λ = 1.302 explained 6.51% of the variance. Another EFA, using varimax rotation, a factor loading > 0.70, and a cross-loading of < 0.40, led to the following solution: a general factor with items from all five subscales (items 3, 8, 20, 4, 1, and 15 in descending loading order, all items phrased positively) and a mental/motivational factor with items from the subscales mental fatigue and reduced motivation (items 13, 9, 18, and 19 in descending order, all items phrased negatively). Additionally, the goodness-of-fit test for a unifactorial solution, computed with Hayes' macro (25) with all items, was significant (p < 0.001, χ² = 5457.688, df = 170).

Given the significant differences between the MFI-20 subscales across gender and age, we calculated percentile ranks based on the MFI-20 subscale raw scores for gender and age separately. The percent rank norms as a non-linear and distribution-free test value transformation were applied since the distributions displayed a profound right-tailed skewness. A percent rank PR_n shows the percentage of the norm sample that scored lower and exactly as the subject n, i.e., PRn=100*freqcum(xn)N. The MFI-20 norm tables are shown in the Supplementary material for the subscales and the total score, separately by gender and age group.

The kish selection grid was used to identify randomly assigned target persons. A total of 57.2% of the target population did not participate due to the following reasons: household did not respond within four attempts (n = 791, 13.4%); household refused participation (n = 1374, 23.3%); target person did not respond within four attempts (n = 288, 4.9%); target person was on a business or vacation trip (n = 61, 1.0%); target person was ill (n = 75, 1.3%); and target person refused participation (n = 786, 13.3%).