UC Irvine

Healthcare systems face significant resilience challenges during disasters, as hazards can simultaneously disrupt facilities, transportation networks, utilities, staffing, and resource availability. These challenges are especially acute in senior healthcare, where older adults have heterogeneous medical profiles, mobility limitations, equipment dependencies, and continuity-of-care requirements. This dissertation frames healthcare disaster resilience as a multi-scale geospatial decision-support problem, in which computational models integrate heterogeneous data about people, facilities, hazards, infrastructure, policies, and scarce resources to support efficient, fair, and operationally actionable decisions.This perspective is explored through three connected systems. First, at the facility scale, we develop iFair, a fairness-aware resource allocation framework for senior healthcare facilities. In this work, residents, caregivers, equipment, care tasks, facility layouts, and dynamic emergency conditions are modeled through a digital-twin-inspired representation. Allocation is formulated as a constrained scheduling problem that balances task efficiency, resident criticality, individualized needs, and preference satisfaction during urgent events such as evacuation. The results show that accounting for resident needs and preferences can improve both fairness and operational efficiency.Second, at the regional scale, we develop RADAR, a geospatial decision-support framework for disaster response. The framework integrates multisource GIS feeds, facility status, transportation conditions, resource availability, and policy-driven ranking functions to support evacuation, relocation, and resource-sharing decisions. It formulates regional coordination as a coupled matching-and-routing problem, combining stable matching with hazard-aware routing to generate interpretable recommendations for Emergency Operations Centers and healthcare stakeholders. This enables regional plans that account for facility compatibility, capacity, route safety, hazard evolution, and operational policy constraints.Third, at the infrastructure and community scale, we develop BlackOps, a GIS-informed preparedness framework for healthcare resilience amid prolonged power outages. The framework models facility autonomy, backup-power shortfall, infrastructure dependencies, road-network accessibility, hazard exposure, social vulnerability, and regulatory requirements to identify facilities most at risk of losing critical services during multi-day outages. It also uses bounded retrieval-augmented language-model workflows to extract facility-specific autonomy requirements from unstructured regulatory documents, while keeping risk computation grounded in geospatial analysis, infrastructure modeling, and interpretable preparedness indices.Together, these contributions demonstrate how geospatial systems can move beyond visualization to serve as computational foundations for healthcare disaster resilience. Across facility-level allocation, regional matching and routing, and infrastructure-aware preparedness planning, this dissertation shows that spatially grounded and human-centered models can help decision-makers reason more effectively about vulnerable populations, scarce resources, cascading disruptions, and time-sensitive operational constraints.