As hospitals struggle with high demand and increasing complexity of patient illnesses, a team of researchers including John C. Malone Assistant Professor of Civil and Systems Engineering Kimia Ghobadi has developed a way to help hospitals more effectively manage their capacity during high demand periods, supporting timely and effective care.
“Hospitals have limited resources and beds, and yet we’re seeing a steady increase in the number of patients, compounding delays in patient care and burnout in nurses and providers,” says Ghobadi. “Hospital managers have to make complex capacity decisions on a daily basis—they can’t simply add more physical beds due to regulation, staffing limitations, and cost.”
The team’s approach offers health care networks an interactive decision-support dashboard to manage resources during sudden increases in patient admissions. Underpinning the interface are data-driven models developed in the team’s previous study, which consider critical factors like the number of incoming patients, their expected length of stay, current hospital capacity and occupancy, and the health network’s existing plans to mitigate a surge.
In collaboration with the Judy Reitz Capacity Command Center at the Johns Hopkins Hospital, the team tested its dashboard during the height of the COVID-19 pandemic and validated their methods with subsequent simulations that showed promising potential to improve resource allocation and capacity management in health care systems.
“Our results show that we can reduce the need for additional beds in a single hospital by roughly 40%,” Ghobadi says. “This reduction increases when we have a network of hospitals working together; for instance, transferring just one patient every other day from one hospital to another in a network of five hospitals could decrease the need for additional beds by over 90%.
“Currently, hospitals usually only transfer patients for medical reasons,” she says. “Our study highlights that occasional strategic transfers to balance capacity could help mitigate overcrowding, which would enable better care and more capacity for the clinical transfers, too.”
The researchers say that their models’ user-friendly interfaces, flexibility, and scalability make them particularly effective.
“We built our models to be flexible so that every hospital can incorporate their distinct practical and operational preferences and needs,” says Ghobadi. “And with an easy-to-use interface, hospital stakeholders can easily modify and run the complex underlying models on the fly, enabling them to test out their surge strategies in real time.”
Ghobadi posits that as the health care industry begins to integrate data and AI more extensively, methods such as these will become essential.
“There is an increasing amount of data in health care that can inform decisions. With challenges like rising patient numbers and potential for future fluctuations due to natural or man-made disasters, there is enthusiasm for using data and modeling to improve patient outcomes. These models can support human decision-makers with tailored, interpretable recommendations and actionable insights. Our method is a step in that direction,” she says.
The team plans to continue its work with improvements for the decision-support dashboard and its underlying models. The researchers aim to create forecasting models of patient demand, including patients admitted through emergency departments; they also plan to gain insight into health systems’ surge policies to better optimize them for high-demand scenarios.
Additional Johns Hopkins collaborators on this work include Diego A. Martínez, a Malone Center affiliate and an assistant professor of emergency medicine and health sciences informatics at the Johns Hopkins School of Medicine; doctoral students Felix Parker and Fardin Ganjkhanloo; and James Scheulen, the former chief administrative officer of emergency medicine and capacity management in the Department of Emergency Medicine.