Evidence-based construction and validation of an agent-based model of community integrated care for older adults living with dementia.

Objective: Community dementia care involves interacting patients, caregivers, and formal care providers whose long-term dynamics are difficult to characterize with single-component empirical studies. We constructed and assessed a baseline agent-based model (ABM) to establish a transparent, inspectable reference scenario against which future computational experiments on organizational configurations can be evaluated. Methods: A fully rule-based ABM of community dementia care was developed using prior empirical studies, published literature, policy documents, and structured expert consultation. Each rule and parameter was documented with its evidence source and population provenance. The formal baseline used 1,000 synthetic care units, 100 independent replications, a 60-month horizon, and all configuration switches at baseline values. Behavioral and psychological symptoms of dementia (BPSD) prevalence at initialization was aligned with a prespecified Chinese meta-analytic anchor. Results: BPSD prevalence increased from 75.6% at initialization to 96.4% at month 60 (Monte Carlo 95% CI, 96.3-96.5%). Mean caregiver burden rose from 0.505 to 0.782. By month 60, a mean of 161 care units were in institutional care. Scale and replication diagnostics confirmed numerical adequacy; local sensitivity analysis showed limited changes in primary outcomes across the tested range. Conclusion: Under the documented formal baseline rules, the model generates conditional trajectories of disease progression, caregiver burden, closed-loop referral completion, and institutional-care membership. The resulting transparent baseline provides a comparison condition for future, explicitly scoped computational experiments on organizational configurations.