Lee’s Study Predicts Likelihood of COVID-19 Hospitalization

In a new study set to appear in a forthcoming issue of the Journal of International Technology and Information Management, Turner College assistant professor of management information systems Yoon Lee and his colleague Riyaz Sikora of the University of Texas – Arlington explain that the surge in COVID-19 infections during the early stages of the pandemic precluded access to needed intensive care treatment by many high-risk patients.  To better deal with limited availability of ICU resources, public health officials called for the development of mathematical models for predicting the demand for ICU resources as a way of improving organizational management.  While prior models predicting future medical events, including COVID-19-related symptom changes, focused on long-term horizons (more than 15 days), Lee and Sikora proposed the first method of predicting the likelihood of COVID-19 inpatients’ admission to the ICU within a time frame of 12 hours.  Tests of their model are based on the dataset posted on Kaggle by a hospital in Brazil, which includes 384 COVID-19 patients who provide five sets of measurements in five different time windows (from 0 to 2 hours, 2 to 4 hours, 4 to 6 hours, 6 to 12 hours, and above 12 hours from the time of hospital admission).  When their model is compared to other methods using various pre-processing techniques and classifiers, it achieves the best performance, making it a viable option for predicting the likelihood of COVID-19 inpatients’ admission to the ICU within only a few hours.