The high number of hospital cases during the COVID-19 pandemic have caused an overwhelming strain on clinical resources. Dr. Long Nguyen, assistant professor, computer science and data science, and colleagues have developed a model that can help health care workers quickly identify patients who need hospital treatment while still monitoring at-home patients.
Dr. Long partnered with Muzhe Guo, Dr. Hongfei Du, and Dr. Fang Jin from The George Washington University on this study. They used a dataset acquired from wearable technologies worn by COVID-19 patients. Their analysis introduced a model that combines long-short term memory and deep neural network to accurately classify and predict the COVID-19 disease stages of mild, moderate, severe and recovered.
They then identified vital indicators to help both patients and doctors determine their COVID-19 stage. Finally, they created case studies demonstrating the differences between severe and mild patients and to show the signs of recovery from COVID-19 by extracting shape patterns based on temporal features of patients.
Importantly the predictive model, by exploiting temporal stream data and attribute stream data simultaneously for disease stage classification, is able to categorize patients into stages for an improved, data-driven method for making treatment decisions.
“We hope that we can maximize the treatment for severe patients while minimizing the risk of disease transmission,” says Dr. Nguyen. “Our model will help doctors focus on hospitalized patients while still monitoring at-home patients.”
In addition to helping prioritize hospital resources on COVID-19 patients with the most severe cases, this model has the potential to help patients monitor their treatment needs at home.
“We see potential for remote monitoring of patients in many cases,” says Dr. Nguyen. “However, there are some edge cases that need further investigation.”
For example, some patients may quickly move through COVID-19 stages and need clinical treatment faster than they expect.
“We need to explore the timing for alerting patients and doctors for hospitalization to help those patients get proper treatment,” says Dr. Nguyen.
He and his colleagues also planned further research using larger data samples.