Leveraging Social Networks to Build Healthy Community for Low-Income Older Adults and Persons with Disabilities: Agent-Based Modeling

Background and Purpose: Prevalence of chronic illnesses, such as diabetes and hypertension, is high among low-income older adults and persons with disability residing in publicly subsidized housing (PSH). Implementing daily healthy behaviors (e.g., exercise, healthy eating) for successful chronic illness management (CIM) is critical for their health and quality of life. Anecdotal evidence suggests that tenants at PSH struggle with CIM due to limited resources in their social environment. Correlational studies showed that CIM is associated with indicators of good social environment, such as perceived social support, social capital, and social cohesion. However, those studies still have methodological limitations such as inability to model dynamics of social environment (e.g., network turnover) or subjective self-reporting. This study attempted to overcome those limitations by using agent-based modeling (ABM) with previously collected social network analysis (SNA) data to assess whether social environments influence community-level CIM among tenants in PSH.

Methods: ABM was employed to simulate the effects of social networks on CIM over time in a community. Parameters were drawn from data collected for a pilot study conducted at a PSH in Arizona (n=26). Using SNA in the pilot study, we identified three key community attributes measuring different aspects of social environments relevant to CIM: (a) proportion of community members who trust health professionals (sample mean=79.6%; range: 0%―100%); (b) proportion of community members who share health matters (sample mean=81.6%; range: 0%―100%); and (c) proportion of community members who exercise regularly (sample mean=57.6%; range: 0%―100%). NetLogo 6.0 was used to simulate four scenarios that measured changes to CIM adherence over time based on community characteristics.

Results: The baseline model slightly lowered the average CIM (mean=3.27; 95% CI [3.22, 3.32]) compared to the estimate from the cross-sectional survey (Mean=4.24; SD = 1.08) if accounting for dynamic interactions between agents. When the proportion of community members trusting health professionals declined from 79.6% (the baseline parameter) to 50%, the average adherence to CIM dropped from 3.27 to 1.78 (95% CI [1.75, 1.81]). When the proportion of community members willing to discuss health matters decreased from 81.6% (the baseline parameter) to 50%, we observed the average adherence to CIM was 2.10 (95% CI [2.07, 2.14]), dropping 1.17 points from the baseline value. Lastly, if this community managed to increase the proportion of community members regularly exercising from 57.6% to 75%, the average adherence to CIM increased by almost 1 point (mean=4.27; 95% CI [4.23, 4.32]).

Conclusions and Implications: Findings highlight dynamics of social contagion relevant to CIM among older adults and persons with disabilities at PSH who are socioeconomically disadvantaged. One notable area for intervention was that fewer people exercising might have devalued a culture of physical activity in this community, which can be improved through community programs promoting physical health and activity. Further studies are recommended to use more representative survey data with SNA to improve reliability and validity of ABM models, which account for interactive data observable in social environments.