Health behavior profiles and potential support needs among community residents with clustered cardiovascular risk factors: a latent profile and explainable machine learning analysis for community-based prevention
- 1School of Medicine, Lishui University, No. 1 Xueyuan Road, Lishui, Zhejiang 323000 China
- 2Department of Nursing, The First Affiliated Hospital of Lishui University, Lishui People’s Hospital, Lishui, 323000 Zhejiang China
- 3School of Medicine, Aksu Vocational and Technical College, Aksu Prefecture, Xinjiang, 843100 China
Abstract
Background: Community residents with clustered cardiovascular risk factors may exhibit marked heterogeneity in health behavior patterns, posing challenges for community-based cardiovascular prevention and health behavior support. This study aimed to identify subgroups with different levels of health behavior and examine factors associated with subgroup membership using explainable machine learning.
Methods: This secondary analysis used cross-sectional questionnaire data from 475 community residents with clustered cardiovascular risk factors. Health behaviors were assessed using the six dimensions of the Health-Promoting Lifestyle Profile II (HPLP-II). Latent profile analysis (LPA) was used to identify latent health behavior profiles that may reflect different behavioral support needs. The sample was split into training and internal validation sets at an 8:2 ratio. Variable selection was performed using Least Absolute Shrinkage and Selection Operator regression and Random Forest. Five classification models were developed and compared, and the model retained for exploratory analysis was subsequently interpreted using Shapley Additive Explanations (SHAP).
Results: Four health behavior subgroups were identified: low health-promoting behavior profile (9.7%), low-to-moderate health-promoting behavior profile (45.5%), moderate-to-high health-promoting behavior profile (34.3%), and high health-promoting behavior profile (10.5%). In the internal validation set, the Random Forest (RF) model showed a comparatively balanced performance across multiple evaluation metrics, with an accuracy of 0.777, Kappa of 0.640, macro-averaged precision of 0.852, macro-averaged F1 score of 0.714, and area under the curve of 0.892; however, its minority-class identification requires further validation. SHAP analysis indicated that health behavior capability, social support, self-efficacy, and age were key contributors to subgroup classification.
Conclusions: Health behaviors among community residents with clustered cardiovascular risk factors exhibit marked heterogeneity. These findings provide descriptive evidence of heterogeneity in health behaviors and may support stratified approaches to health behavior support and community-based cardiovascular prevention strategies.
Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28868-y.
Related
- Bringing Cardiovascular Prevention into the Community: Protocol for an Observational Study of a Mosque-Based Health Initiative Combining Education and Cardiovascular Screening
- Prevalence and determinants of urban malaria among urban residents of Dilla Town, southern Ethiopia: a community-based cross-sectional study
- An explainable artificial intelligence framework using electrocardiograms to predict mortality and major adverse cardiovascular events
- Knowledge of antibiotic resistance and prevalence of antibiotic usage among urban residents of Kochi, India: a community-based cross-sectional study
- Cardiovascular and Cancer Mortality Among Texas Residents With Cancer: A Population-Based Study in Texas
BibTeX
@article{li2026health,
title = {{Health behavior profiles and potential support needs among community residents with clustered cardiovascular risk factors: a latent profile and explainable machine learning analysis for community-based prevention}},
author = {Kaihe Li and Xinyi Yan and Ying Zhou and Xuefen Lan and Xiaofen Li and Juanjuan Mai and Minqi Xia and Ruiming Chen},
journal = {BMC Public Health},
year = {2026},
month = jul,
doi = {10.1186/s12889-026-28868-y},
url = {https://doi.org/10.1186/s12889-026-28868-y},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Health behavior profiles and potential support needs among community residents with clustered cardiovascular risk factors: a latent profile and explainable machine learning analysis for community-based prevention
AU - Kaihe Li
AU - Xinyi Yan
AU - Ying Zhou
AU - Xuefen Lan
AU - Xiaofen Li
AU - Juanjuan Mai
AU - Minqi Xia
AU - Ruiming Chen
PY - 2026
DA - 2026/07/31
JO - BMC Public Health
DO - 10.1186/s12889-026-28868-y
UR - https://doi.org/10.1186/s12889-026-28868-y
AB - Background: Community residents with clustered cardiovascular risk factors may exhibit marked heterogeneity in health behavior patterns, posing challenges for community-based cardiovascular prevention and health behavior support. This study aimed to identify subgroups with different levels of health behavior and examine factors associated with subgroup membership using explainable machine learning. Methods: This secondary analysis used cross-sectional questionnaire data from 475 community residents with clustered cardiovascular risk factors. Health behaviors were assessed using the six dimensions of the Health-Promoting Lifestyle Profile II (HPLP-II). Latent profile analysis (LPA) was used to identify latent health behavior profiles that may reflect different behavioral support needs. The sample was split into training and internal validation sets at an 8:2 ratio. Variable selection was performed using Least Absolute Shrinkage and Selection Operator regression and Random Forest. Five classification models were developed and compared, and the model retained for exploratory analysis was subsequently interpreted using Shapley Additive Explanations (SHAP). Results: Four health behavior subgroups were identified: low health-promoting behavior profile (9.7%), low-to-moderate health-promoting behavior profile (45.5%), moderate-to-high health-promoting behavior profile (34.3%), and high health-promoting behavior profile (10.5%). In the internal validation set, the Random Forest (RF) model showed a comparatively balanced performance across multiple evaluation metrics, with an accuracy of 0.777, Kappa of 0.640, macro-averaged precision of 0.852, macro-averaged F1 score of 0.714, and area under the curve of 0.892; however, its minority-class identification requires further validation. SHAP analysis indicated that health behavior capability, social support, self-efficacy, and age were key contributors to subgroup classification. Conclusions: Health behaviors among community residents with clustered cardiovascular risk factors exhibit marked heterogeneity. These findings provide descriptive evidence of heterogeneity in health behaviors and may support stratified approaches to health behavior support and community-based cardiovascular prevention strategies. Supplementary Information: The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28868-y.
SN - 1471-2458
ER -