Derivation and Preliminary Internal Evaluation of a Candidate 10-Item Short Form of the Multidimensional Fatigue Symptom Inventory in Women With Breast Cancer

Objective To derive and preliminarily evaluate a candidate 10-item short form from the Chinese 27-item MFSI-SF for assessing cancer-related fatigue among women with breast cancer. Methods This psychometric study included 462 women with breast cancer. The sample was randomly split into exploratory and held-out validation samples (n = 231 each). Mokken scale analysis and exploratory graph analysis were used for item reduction and dimensionality assessment, and a multidimensional graded response model was fitted in the held-out validation sample. Associations between the candidate 10-item and source 27-item scores were examined using Pearson and Spearman correlations. Education-related differential item functioning (DIF) was evaluated using ordinal logistic regression. Results The 10 retained items formed 3 dimensions: Vigor, Physical and Mental Fatigue, and Emotional Fatigue. No distinct General Fatigue dimension was retained, and Physical and Mental Fatigue items clustered together. Loevinger H coefficients ranged from 0.624 to 0.698, and bootstrap EGA item stability was 1.00 for every item. Confirmatory model fit was inadequate, with three of the four prespecified fit indices failing to meet their criteria: CFI =.941, TLI =.877, SRMSR =.095, and RMSEA =.142 (90% CI,.110–.176); only the CFI met its prespecified criterion. Standardized factor loadings ranged from 0.798 to 0.946. The candidate 10-item score was strongly associated with the 27-item score (Pearson r =.934; Spearman ρ =.941), although this part–whole association should not be interpreted as evidence of agreement or interchangeability. Although several DIF tests were statistically significant, all ΔR² values were below the prespecified threshold for practical importance (range, 0.001-0.006). Conclusion The 10-item configuration showed strong scalability and stable exploratory item assignments but inadequate confirmatory model fit. It should therefore be regarded as a candidate short form requiring further structural evaluation, content validation, reliability assessment, and score interpretation in independent samples before clinical use.