Conditional Associations Among Chronic Disease Indicators, Risk Behaviors, Health Habits, and Disability Indicators in an Analytic ENDES Sample of Peruvian Adults: A Secondary Analysis Using Exploratory Network Analysis, 2021–2023

Vollständiger Abstract

Worum geht es in dieser Arbeit?

ABSTRACT Background Multidomain health indicators, including chronic disease indicators, risk behaviors, health habits, depressive symptomatology, disability indicators, and healthcare‐access indicators, may be interconnected through complex patterns of conditional association. Network analysis provides an exploratory framework for estimating such interdependencies without implying causality or clinically confirmed comorbidity patterns. Objective To estimate the network structure of conditional associations among selected ENDES health‐related indicators in an analytic sample of Peruvian adults from 2021 to 2023, and to descriptively explore sex‐specific patterns. Methods A secondary analysis was conducted using ENDES 2021–2023 data. A total of 92,101 respondents aged 18 years or older who completed the Health Questionnaire module and had complete data for the selected variables were included. Twenty‐one indicators selected a priori from five domains—chronic disease indicators, behavioral risks, health habits, disability/depressive symptomatology, and healthcare‐access/anthropometric context—were modeled using a mixed graphical model. Bootstrapped edge‐weight accuracy and Expected Influence stability diagnostics were added. Sex‐specific networks were examined descriptively, without a formal Network Comparison Test. Results The overall network showed positive conditional associations among smoking indicators and alcohol consumption, among disability indicators, and between age and hypertension, diabetes, visual disability, and motor disability. Negative conditional associations were observed between daily smoking and both health insurance coverage and fruit consumption, as well as between depressive symptomatology and fruit‐based indicators. In sex‐specific networks, men showed a descriptively denser disability‐related clustering than women, but these visual differences were not inferentially tested. Conclusions Selected multidomain ENDES health‐related indicators were conditionally interconnected in the analytic sample. Findings should be interpreted as exploratory and hypothesis‐generating and do not imply causality, directionality, nationally weighted network estimates, statistically confirmed sex differences, or clinically confirmed comorbidity patterns.