.png){kind=icon}
.png){kind=icon}
Return
Longitudinal associations of lifestyle patterns on biomarkers of major chronic diseases and quality of life: comparing latent variable discovery methods
.png){kind=icon}
OA
AI
Q
D
H
R
J
E
G
DOI：10.1186/s13690-026-02028-5.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Lifestyle patterns cluster across multiple behavioural domains and contribute to the development of chronic diseases, yet existing evidence remains limited by heterogeneity in assessed lifestyle factors, methods used to identify patterns and study populations. We identified lifestyle patterns using six lifestyle domains (nutrition, physical activity, stress, sleep, social connection and substance use) in the Dutch Lifelines cohort (N = 112,842), by applying two latent clustering approaches: Latent Class Analysis (LCA) and Bayesian Latent Analysis for Tabular data (BLAT). We examined associations between these patterns and key physiological biomarkers of major chronic diseases, including blood pressure, lung function, markers of glucose homeostasis, and quality of life, using linear mixed-effects models. Across both methods, five lifestyle patterns were identified, of which three were highly consistent between LCA and BLAT (one “healthiest” pattern, one “unhealthiest” pattern, and one mixed healthy pattern). Association analytical samples ranged from 76,463 participants for lung function analyses to over 112,000 participants for other outcomes (60% women; mean age 44.5 years). Lifestyle patterns identified by both methods showed comparable associations with health outcomes, with overlapping patterns demonstrating similar effect sizes across physiological indicators. Differences and similarities in pattern structure between LCA and BLAT were reflected in corresponding health risks, supporting the comparability of these approaches not only in behavioural classification but also in epidemiological relevance. Both methods effectively identified meaningful lifestyle patterns in large-scale categorical data; selection between them may therefore depend on specific analytical objectives in future dimension reduction analyses.
Keywords:
Decision support
Machine learning
Multiple lifestyle factors
Dimension reduction
Non-communicable diseases
Bayesian Latent Analysis for Tabular data
Journal
IF:
3.2
Papers:
1.8K
Citations:
4.4K
Organization
Cited Papers