Real-time Cohorting of Nursing Care into Bubbles

Jeffrey Keithley, Tinh Tran, Lucas Zach-Ryan, D M Hasibul Hasan, Brodie McCuen, Sriram Pemmaraju, Bijaya Adhikari

Keywords: online patient cohorting, healthcare-associated infections, greedy algorithms, integer linear programming

TL;DR: Minimizing healthcare-acquired infection spread in real-time by creating cohorts of HCPs and patients as patients are being admitted, discharged or transferred while maintaining reasonably low levels of excess mobility and HCP workload.

Abstract:

Healthcare facilities, such as hospitals and long-term care facilities, that house vulnerable populations are sites for pathogens (such as Antibiotic-Resistant Organisms, \textit{Clostridiodes Difficile}, influenza viruses, and SARS-CoV-2 among others) to spread. The contacts between healthcare providers (HCPs) and the patients that naturally emerge during the course of care delivery also serve as pathways for the infections to flow. Prior work has identified that cohorting patients and HCPs into nearly isolated groups leads to a reduction in infection. However, most of these works are either too simplistic (e.g. cohorting after observing infections) or have limited practical use (e.g. retrospective cohorting). In this paper, our goal is to minimize infection spread in real-time by creating cohorts of HCPs and patients on the fly as patients are being admitted, discharged or transferred.

Specifically, we formulate the novel \textsc{Online Bubble Clustering Problem}, which asks to create and maintain cohorts of HCPs and patients that have limited external contacts, yet meet the care demands of the patients and care capacity of the HCPs. We also theoretically demonstrate that the problem is very challenging in both deterministic and stochastic settings, implying no algorithm could achieve results close to the optimal solution. Despite the hardness, we propose natural heuristics and design offline an integer linear programming (ILP) approach to contrast the heuristics against the optimal solution. We also conduct extensive agent-based modeling experiments on granular HCP mobility data collected using sensor systems we previously deployed in a medical intensive care unit over a 30-day period, overlaid with a COVID-19 agent-based disease model. Our simulation results show that real-time cohorting leads to significantly lower disease prevalence, cutting cases in half in some scenarios, while maintaining reasonably low levels of excess mobility and HCP workload.

Area: Search, Optimization, Planning, and Scheduling (SOPS)

Generative A I: I acknowledge that I have read and will follow this policy.

Submission Number: 1405