Purdue Engineering-IUSM healthcare analytics approach can help hospitals forecast COVID-19 demand, prepare for future outbreaks
Purdue Engineering-IUSM healthcare analytics approach can help hospitals forecast COVID-19 demand, prepare for future outbreaks
Purdue Engineering experts are collaborating with Indiana University School of Medicine (IUSM) colleagues to develop and implement a simulation model to help a major Indianapolis hospital overcome a key challenge in providing necessary care to all COVID-19 patients in need. The impact of this innovative healthcare analytics approach could extend to healthcare organizations in additional cities and countries and to other disease outbreaks.
In Indianapolis, as well as New York and many other cities, acute-care hospitals have been struggling to accurately forecast demand for intensive care unit (ICU) beds and ventilators, along with related staffing and supplies, during the COVID-19 pandemic. Hospitals and other types of healthcare organizations, such as nursing homes, also are grappling with how to cope with admission surges.
The better a hospital, for example, can pinpoint needs in advance, the more efficiently it can serve patients. But the task requires planning to allocate the right quantity of resources at the right time to each patient population subgroup – resulting to more lives saved, fewer readmissions, and better outcomes for survivors.
Analytics gap
Unfortunately, a gap in analytics capability has been thwarting these efforts, says Nan Kong, PhD, associate professor in the Weldon School of Biomedical Engineering at Purdue University, who is leading a project to benefit IUSM-supported Sidney & Lois Eskenazi Hospital, the flagship medical center for Eskenazi Health, founded in 1859 as Indiana’s oldest public health system.
“Amid volatile and uncertain conditions, and without good understanding of the disease transmission and progression, it has been very difficult for hospitals to precisely predict what resources they’ll need to address the COVID-19 needs,” Kong says.
“That’s largely because most models being used are based on aggregated public health data, on a state or country level, and don’t incorporate hospital- and patient-specific characteristics. Knowing how many people in a geographic area will be infected and how many will die isn’t that helpful to a hospital CEO. Rather, it’s critical for a hospital to receive a steady flow of timely, reliable and directly relevant information.”
Forecasting ICU bed, ventilator demand
In collaboration with IUSM and Eskenazi staff, the Purdue researchers have developed and calibrated a baseline model focused on forecasting daily demand for ICU beds and ventilators, which can help Eskenazi assign staff and allot supplies for the near future. A web app will be launched shortly to facilitate information sharing with hospital leadership.
“Our stochastic simulation model-based health analytics approach differs from what’s out there in that it is grounded on granular, real-world data collected in a timely manner from Eskenazi patients’ EHRs (electronic health records),” Kong says. “Our model enables predictions regarding two important variables with quantified confidence: 1) the number of patients who have been hospitalized and tested positive for COVID-19 but are not on ventilators, and 2) the number of COVID-19 patients on ventilators.”
The model further takes into account such factors as a patient’s length of stay, gender, age, and underlying conditions.
The project team includes Zhouyang Lou, a PhD student in Purdue’s School of Industrial Engineering, who is co-supervised by Paul Griffin, PhD, director of Purdue’s Regenstrief Center for Healthcare Engineering. These Purdue team members are collaborating with Christopher Callahan, MD, chief research and development officer of Eskenazi Health; Randall Grout, MD, MS, FAAP, research informatician and pediatrician at Eskenazi Health; and Wanzhu Tu, PhD, professor of biostatistics at IUSM.
The team is drawing on previous experience of Kong, the principal investigator for the interdisciplinary Purdue Biometrical Analytics and Systems Optimization Research Lab, in improving business analytics capabilities for healthcare organizations.
“Transformative impact”
Kong foresees broad applications of his team’s inventive work with Eskenazi and IUSM.
"Our provider-based approach could have a transformative impact, making healthcare organizations more resilient in rapidly-changing, uncertain times," Kong says.
“Precision health based on systems dynamics modeling could benefit hospitals and nursing homes, among various healthcare organizations, throughout the United States and in other countries, including in cities that haven’t reached COVID-19 peaks. Beyond responding to the current pandemic, models tailored to a particular healthcare organization have the power to help address subsequent waves of the novel coronavirus and manage other pandemics more proactively and effectively.”