Five Forms of Capital Map Older Adults’ Digital Health Information Divide

Older adults who score lower on measures of cultural and psychological capital also show less proactive health information behavior, according to a cross-sectional questionnaire study published in Frontiers in Public Health that used Bayesian network modeling to map how different forms of structural capital connect to the ways older people find, evaluate, and use digital health information. The study, conducted by researchers at Nanjing Medical University and affiliated institutions, analyzed survey responses from 535 Chinese adults aged 60 and above and identified fifteen directed pathways linking economic, cultural, social, political, and psychological capital to thirteen dimensions of information behavior.

The research team grounded their work in two theoretical frameworks: the Theory of Structural Poverty, which situates individual resource deprivation within broader social and institutional arrangements, and Wilson’s Information Behavior Model, which describes how information needs translate into seeking, evaluation, and use. Economic capital was measured through income, assets, consumption expenditure, and perceived economic stress. Cultural capital covered educational attainment, cultural participation, online information search proficiency, and ongoing learning. Social capital captured supportive networks, help available in difficulties, community participation, and interpersonal interaction. Psychological capital was assessed through self-efficacy, emotional regulation, resilience, and emotional stability, while political capital measured perceived access to public resources, sense of participation in decision-making, and perceived fairness in resource distribution.

Participants were recruited through non-probability convenience sampling, with questionnaires distributed both online via the Wenjuanxing platform and offline through in-person interviews at community health service centers, universities for older adults, and senior activity centers. Of 541 questionnaires collected, six were excluded for incomplete responses, leaving 535 complete datasets. Each latent variable was measured with three to four items on a five-point Likert scale, and reliability analysis showed Cronbach’s alpha coefficients above 0.7 for all latent variables. The authors acknowledge that this sampling approach did not ensure representativeness of the broader older population, and that individuals with limited digital access, lower educational attainment, or limited community participation may have been underrepresented.

Descriptive statistics revealed meaningful variation across the five capital dimensions. Psychological capital scored highest, with a mean of 14.24 and a standard deviation of 3.11, while economic capital scored lowest at 11.31 with a standard deviation of 3.68. Among the thirteen information behavior dimensions derived from Wilson’s model, active and ongoing search scored highest at 12.75, while environmental factors scored lowest at 8.32, suggesting that external conditions such as digital device accessibility and public information infrastructure may pose particular constraints for this population.

The core analytical method was a discrete Bayesian network, a probabilistic graphical model in which nodes represent variables and directed edges encode conditional dependencies. Dimension scores were averaged, placed on a one-to-five scale, and categorized into low, medium, and high states using cutoffs at 2.5 and 3.5. Structural learning employed the tabu search algorithm in the bnlearn R package with the Bayesian Information Criterion as the scoring measure, and 200 bootstrap samples assessed network stability, retaining arcs with strength values above 0.75. Spearman rank correlation analysis with Bonferroni correction complemented the network modeling.

Four behavioral variables were highlighted as proxy indicators of vulnerability to the digital health information divide: active and ongoing search, information source attributes, self-efficacy, and passive attention or search. The authors stress that these indicators were not intended to directly measure the divide itself but to capture behavioral manifestations associated with accessing, evaluating, and engaging with digital health information. Passive exposure, they note, does not necessarily indicate disadvantage and must be interpreted in context.

The network analysis positioned cultural capital as the most highly connected hub. Lower cultural capital was linked to lower psychological capital and weaker self-efficacy in one chain, and to reduced ability to recognize information source attributes in another. Cultural capital also connected through social capital and demographic and social learning variables to less frequent active and ongoing search, and through political capital directly to active search behavior. Psychological capital occupied an intermediate position across multiple chains, connecting to political capital, social capital, and social learning on routes toward both active searching and passive exposure. Political capital linked to economic capital and, through demographic and role-related and environmental factors, to passive attention and search.

Spearman correlations reinforced these patterns. Psychological capital showed a perfect positive correlation with stress adaptation capability, and self-efficacy was strongly correlated with active and sustained information search. Economic capital correlated moderately with cultural and social capital but showed weak, statistically insignificant correlations with behavioral variables such as demographic factors and interpersonal relationships, consistent with the network finding that economic capital operates mainly through indirect pathways tied to access and channel diversification rather than through direct behavioral links.

The authors are careful about interpretation. Because the study is cross-sectional, the directed edges in the Bayesian network represent statistical dependencies within the sample and cannot establish temporal ordering, causal effects, or mediation. No formal mediation analysis was conducted, so the apparent intermediary role of psychological capital should not be read as a statistically established mediation effect. The authors also note that discretizing continuous scores into three categories may lose information and affect the learned network structure, and that the findings may be sensitive to the chosen cutoffs. The sample was drawn from a specific geographic region, further limiting generalizability.

Despite these constraints, the study offers a framework for thinking about the divide as a networked phenomenon rather than a single deficiency. The authors propose governance strategies combining capital empowerment, pathway-based precision interventions, and institutionalized long-term support, and they call for longitudinal, intervention, and comparative studies to verify the direction and stability of the identified pathways and to test whether strengthening particular forms of capital actually changes information behavior over time.

Subject of Research: How economic, cultural, social, political, and psychological capital shape older adults' health information behavior and the digital divide.

Article Title: Study on the formation mechanism of the digital health information divide and capital-driven pathways among the older adults

Article References: Yang, X., Wang, B., Xu, M., Chen, H., Huang, L., Wang, H., Chen, W., Fan, Q., Yu, X., Zhang, S., Shen, S., & Pan, H. (2026). Study on the formation mechanism of the digital health information divide and capital-driven pathways among the older adults. Frontiers in Public Health, 14, Article 1913689. https://doi.org/10.3389/fpubh.2026.1913689

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1913689

Keywords: digital divide, older adults, health information, Bayesian network, structural capital, health literacy, information behavior, digital health, China, cross-sectional study, formation, mechanism