Uncovering heterogeneous health burden in axial spondyloarthritis through latent class analysis of the ASAS health index.

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objective To identify and characterize unobserved subgroups of patients with axial spondyloarthritis (axSpA) based on response patterns to the Assessment of SpondyloArthritis International Society Health Index (ASAS HI) using latent class analysis (LCA). Methods We conducted LCA on 17 dichotomous ASAS HI items in a cohort of 180 axSpA patients. Models with 2–6 classes were evaluated using Bayesian and Akaike Information Criteria, selecting the optimal model based on the lowest BIC. Class profiles were defined by conditional item endorsement probabilities. Construct validity analyses included comparisons of BASDAI, BASFI, ASDAS, and total ASAS HI scores. Internal validation was performed via bootstrap resampling. Secondary multinomial logistic regression assessed the influence of disease indices on class membership. Results A five-class model showed distinct health impact profiles. Class 1 (n=25, 13.9%) exhibited the highest burden across physical, emotional, and motivational domains, while Class 2 (n=63, 35%) showed minimal impact. Intermediate classes reflected predominant physical (Class 3, n=35), fatigue (Class 4, n=18), or functional (Class 5, n=36)) impairment patterns. Significant differences in BASDAI, BASFI, ASDAS, and ASAS HI total scores were observed across classes (all p < 0.001). Combined multinomial regression models including BASDAI, ASDAS, and BASFI showed a pseudo-R² of 0.33 for class allocation. Posterior classification probabilities were high (mean = 0.88), and bootstrap analysis confirmed model stability. Conclusion LCA of ASAS HI responses identified five distinct and clinically meaningful health impact profiles in axSpA. These findings highlight the multidimensional burden of axSpA and support the potential role of PROM-based stratification.