Overview
Flooding presents a significant challenge to healthcare infrastructure, potentially leading to facility damage, localized care capacity interruptions, and expensive patient evacuations. Addressing these multifaceted risks necessitates long-term resilience investments, which must account for inherent uncertainties. A bi-objective two-stage stochastic optimization model has been developed to address this challenge, simultaneously determining permanent facility hardening strategies and scenario-dependent patient evacuation protocols.
The model operates with two distinct objectives. The primary objective is the minimization of combined expected costs, encompassing evacuation expenses, physical damage repair, and business interruption losses. The secondary objective focuses on minimizing a service-disruption impact index. This index integrates both the duration and the scale of service loss, further weighting these factors by a place-based social vulnerability measure.
To solve this optimization problem, the researchers implemented an exact Benders decomposition method and a scalable Lagrangian-dual approach. The Lagrangian method incorporates tailored primal recovery techniques. Both methodologies are embedded within an adaptive procedure designed to construct informative Pareto frontiers, which illustrate the trade-offs between the two objectives. The framework was applied and evaluated through a case study involving 3,752 hospitals and nursing homes located in Texas, utilizing climate-informed tropical cyclone flood scenarios.
Research Context
Healthcare facilities are critical infrastructure components susceptible to damage from flooding events. Such damage can directly impair care delivery, reducing local capacity, and necessitating patient evacuations, which are often costly. Effective long-term resilience planning is essential for mitigating these risks, requiring strategic investments under conditions of uncertainty regarding future flood events.
Existing approaches to risk mitigation often focus on single objectives, potentially overlooking broader societal impacts or the nuanced interplay of economic and social factors. The integration of multiple objectives, specifically economic costs and social vulnerability, into a decision-making framework for infrastructure resilience planning represents an evolution in addressing complex natural hazard risks for critical public services.
Approach
The core of the research involves a bi-objective two-stage stochastic optimization model. This model is designed to optimize two primary decision types: permanent facility hardening investments and scenario-dependent evacuation plans.
The model's first objective function is formulated to minimize the total expected costs. These costs are a summation of several components:
- Expected evacuation costs
- Expected physical damage costs
- Expected business interruption costs
The second objective function aims to minimize a service-disruption impact index. This index is constructed to capture a holistic view of service loss by combining:
- The duration of service loss
- The scale of service loss
- A place-based social-vulnerability weight, which factors in the social vulnerability of the affected area
For solving the optimization model, two distinct computational methods were developed and employed:
- An exact Benders decomposition technique.
- A scalable Lagrangian-dual method, which includes tailored primal recovery procedures.
Both solution methods are integrated into an adaptive procedure. This adaptive procedure's purpose is to construct informative Pareto frontiers, allowing for the visualization and analysis of the trade-offs between minimizing economic costs and minimizing service disruption impacts.
The model's efficacy and performance were assessed through a practical case study. This involved analyzing a dataset of 3,752 hospitals and nursing homes situated in Texas. The evaluation utilized flood scenarios derived from climate-informed tropical cyclone projections, providing a realistic context for the application of the developed optimization framework.
Findings
The application of the bi-objective optimization framework to the Texas case study yielded several key findings:
- A strategy focused solely on minimizing economic losses can result in systematic disparities in protection. Specifically, facilities located in socially vulnerable areas tend to receive less protective hardening under such an approach.
- By allowing for a moderate movement along the Pareto frontier—that is, accepting a slight increase in expected economic costs—it is possible to achieve substantial reductions in service disruption impacts. This indicates a trade-off where improved social outcomes can be secured with relatively contained additional financial investment.
- The computational performance of the developed methods was also evaluated. The exact Benders decomposition proved capable of solving all tested instances. This included larger stress tests, for which an extensive formulation of the problem became memory-limited.
- The Lagrangian-dual method demonstrated significant computational advantages, providing substantially faster high-quality solutions compared to other approaches.
Why This Matters
The findings from this research are relevant for informing decision-making processes concerning healthcare infrastructure resilience planning. By demonstrating that prioritizing economic efficiency alone can disproportionately affect socially vulnerable communities, the study highlights the importance of multi-objective considerations in infrastructure investment. The ability to achieve significant reductions in service disruption impacts with only moderate additional economic outlay provides a practical basis for policymakers and planners to pursue more equitable and resilient outcomes for healthcare provision during flood events.
Key Limitations Mentioned by Researchers
The source mentions that the extensive formulation of the problem became memory-limited for larger stress tests, which led to the development of the exact Benders decomposition and Lagrangian-dual methods. This implies a limitation in directly solving very large instances without specialized decomposition techniques.