Service Center Location with Decision Dependent Utilities with an Application to Early Stage Testing and Vaccination in Epidemic Planning

TL;DR

This paper develops a mixed-integer second-order cone programming model for locating service centers with ambiguous utilities, applying it to epidemic testing and vaccination site planning, and demonstrates its computational effectiveness.

Contribution

It introduces a novel MISOCP formulation for location problems with decision-dependent utility ambiguity, incorporating non-convex constraints and valid inequalities.

Findings

The model can handle non-convex utility constraints effectively.

Strengthened formulations improve computational performance.

Application to COVID-19 testing demonstrates practical relevance.

Abstract

We study a service center location problem with ambiguous utility gains upon receiving service. The model is motivated by the problem of deciding medical clinic/service centers, possibly in rural communities, where residents need to visit the clinics to receive health services. A resident gains his utility based on travel distance, waiting time, and service features of the facility that depend on the clinic location. The elicited location-dependent utilities are assumed to be ambiguously described by an expected value and variance constraint. We show that despite a non-convex nonlinearity, given by a constraint specified by a maximum of two second-order conic functions, the model admits a mixed 0-1 second-order cone (MISOCP) formulation. We study the non-convex substructure of the problem, and present methods for developing its strengthened formulations by using valid tangent inequalities. Computational study shows the effectiveness of solving the strengthened formulations. Examples are used to illustrate the importance of including decision dependent ambiguity. An illustrative example to identify locations for Covid-19 testing and vaccination is used to further illustrate the model and its properties.