Modeling and Evaluation of Clustering Patient Care into Bubbles

TL;DR

This paper introduces a novel model and optimization approach for clustering patients and healthcare professionals into 'bubbles' to minimize infection spread in healthcare settings, balancing infection control with care quality and operational costs.

Contribution

It presents the BubbleClustering problem, an ILP-based solution approach called CoRN, and demonstrates its effectiveness using real hospital data to reduce infection spread while maintaining care standards.

Findings

CoRN significantly reduces infection spread in hospital settings.

The ILP model can be solved optimally for typical hospital units.

Minimal additional costs are incurred by HCPs during clustering.

Abstract

COVID-19 has caused an enormous burden on healthcare facilities around the world. Cohorting patients and healthcare professionals (HCPs) into "bubbles" has been proposed as an infection-control mechanism. In this paper, we present a novel and flexible model for clustering patient care in healthcare facilities into bubbles in order to minimize infection spread. Our model aims to control a variety of costs to patients/residents and HCPs so as to avoid hidden, downstream adverse effects of clustering patient care. This model leads to a discrete optimization problem that we call the BubbleClustering problem. This problem takes as input a temporal visit graph, representing HCP mobility, including visits by HCPs to patient/resident rooms. The output of the problem is a rewired visit graph, obtained by partitioning HCPs and patient rooms into bubbles and rewiring HCP visits to patient rooms so that patient-care is largely confined to the constructed bubbles. Even though the BubbleClustering problem is intractable in general, we present an integer linear programming (ILP) formulation of the problem that can be solved optimally for problem instances that arise from typical hospital units and long-term-care facilities. We call our overall solution approach Cost-aware Rewiring of Networks (CoRN). We evaluate CoRN using fine-grained-movement data from a hospital-medical-intensive-care unit as well as two long-term-care facilities. These data were obtained using sensor systems we built and deployed. The main takeaway from our experimental results is that it is possible to use CoRN to substantially reduce infection spread by cohorting patients and HCPs without sacrificing patient-care, and with minimal excess costs to HCPs in terms of time and distances traveled during a shift.