- Network-informed analysis prioritised depressive symptoms, poor self-rated health, ADL and IADL limitations, diabetes, and multisite pain as key indicators in older adults with symptomatic KOA.
- Latent class analysis identified three clinically interpretable profiles: lower-burden, psychological perceived-health burden, and multidimensional high-burden functional limitation.
- Higher-burden profiles were associated with rural residence, greater comorbidity and sleep disturbance; profiles were partially reproducible in 2018 and inform stratified care research.
Patient Prefer Adherence. 2026 Oct 25;20:631987. doi: 10.2147/PPA.S631987. eCollection 2026.
ABSTRACT
PURPOSE: Older adults with symptomatic knee osteoarthritis (KOA) may exhibit heterogeneous combinations of functional limitation, psychological burden, perceived health, comorbidity burden, and multisite pain. This study aimed to identify network-informed key indicators and derive clinically interpretable, management-oriented health-burden profiles to support multidimensional assessment and future preference-sensitive care research.
PATIENTS AND METHODS: In this cross-sectional secondary analysis, we analyzed 2011 baseline data from the China Health and Retirement Longitudinal Study (CHARLS) among adults aged 60 years or older with physician-diagnosed arthritis or rheumatism and knee pain. Eighteen indicators were modeled using an extended Bayesian information criterion graphical lasso network based on mixed-scale correlations. Three structural metrics were integrated to prioritize indicators, and the prioritized indicators were entered into latent class analysis (LCA). Multinomial logistic regression examined profile correlates. Robustness was assessed across five multiply imputed datasets, after removing diabetes from the LCA, and in an exploratory 2018 later-wave reproducibility analysis.
RESULTS: Among 749 participants, depressive symptoms, poor self-rated health, instrumental and basic activities of daily living limitations, diabetes, and non-knee pain burden ranked highest. Although the two-class model had the lowest Bayesian information criterion (BIC), the three-class model had the lowest sample-size-adjusted BIC (aBIC), acceptable classification quality, adequate minimum class size, and greater clinical interpretability. Profiles reflected lower overall health burden (n=93, 12.4%), psychological-perceived health burden (n=312, 41.7%), and multidimensional high-burden functional limitation (n=344, 45.9%). Rural residence, greater comorbidity burden, and sleep disturbance were associated with higher-burden profiles. The 2018 analysis (n=2280) yielded three broadly analogous qualitative profiles, although class proportions, some item-response patterns, and classification certainty differed from 2011.
CONCLUSION: Network-informed LCA identified three multidimensional health-burden profiles, with partial later-wave reproducibility. These profiles may inform multidimensional assessment and future stratified, preference-sensitive care research but require confirmation in independent clinical cohorts with direct measures of preference, treatment acceptability, and adherence.
PMID:42960188 | PMC:PMC13718463 | DOI:10.2147/PPA.S631987
Share Evidence Blueprint
Save to Google Notes

Search Google Scholar
⭐ My Revision List

