Data-Driven Robust Healthcare Network Design under Disruption Considering Conditional Value at Risk

Abstract

This study proposes a distributionally robust two-stage risk-aversion model for comprehensive healthcare network design. The proposed model considers the weighted average of the expected losses to handle the disruption uncertainty affecting the network performance (i.e., service capacity, service demand, and route capacity) with conditional value-at-risk as a risk assessment measure. The proposed model is implemented in a stylized case study of healthcare network design in Iran. The results indicate the efficiency of the proposed distributionally robust optimization model compared to the traditional two-stage stochastic model. In case of a high level of uncertainty, the proposed distributionally robust network design tends to be more decentralized by establishing more new mobile facilities, selecting more backup facilities at low-risk regions, and assigning more inventory to established facilities. Results provide decision-makers with managerial insights about cost-effective decisions to handle uncertainty. Also, the sensitivity analysis shows that the type of ambiguous set and levels of confidence, risk aversion, and adjustment parameters affect the network performance.