Key points are not available for this paper at this time.
Background Early assessment of nutrition-associated health levels in patients with cancer—encompassing functioning, activities, and participation—is increasingly recognized as essential for nutrition assessment and cancer rehabilitation. Using a combined approach of item response theory (IRT) and graph modeling, our previous study developed a 32-item two-parameter logistic model (2PLM) for assessing the nutrition-associated health levels of cancer patients. By expanding the sample size and updating the methods, this study aimed to refine and extend the previous findings. Methods This study collected International Classification of Functioning, Disability, and Health (ICF) data from 300 cancer patients via a maximum variation sampling strategy, including 200 newly recruited patients and 100 from our previous study. Person abilities ( θ ) and item parameters were calculated by constructing an IRT model, and the benefit index was estimated using a graph model. Results The study constructed a 43-item three-parameter logistic model (3PLM) with high reliability (Cronbach’s α = 0.945, latent class reliability coefficient (LCRC) = 0.967). The estimated θ showed a significantly strong correlation with the Mini-Nutritional Assessment (MNA) ( p 0.001, r winsorized = 0.73), the Patient-Generated Subjective Global Assessment (PG SGA) ( p 0.001, r winsorized = −0.74), and the Nutrition Risk Screening 2002 (NRS 2002) ( p 0.001, r winsorized = −0.60), indicating high validity. In the graph model, 33 items formed a maximal connected subnetwork and 30 items exhibited a significant benefit index. Conclusion The study established a 43-item 3PLM for assessing nutrition-associated health levels in cancer patients and estimated both item parameters and the benefit index. By incorporating a larger sample size and updated analytical methods, the study extended prior findings and supports the utility of this integrated approach for nutritional assessment and cancer rehabilitation.
Chen et al. (Wed,) studied this question.