Murtaza Nasir Places 4th in COVID-19 Risk Factor Modeling Challenge
The models developed by Management Science Ph.D. student Murtaza Nasir can predict patient death as well as ventilation requirements due to COVID-19 with an accuracy of more than 95 percent.
09/30/2020
Murtaza Nasir, a second-year Ph.D. student of Management Science in the Manning School of Business, placed fourth overall in a recent national competition to develop and evaluate computational models to predict COVID-19 related health outcomes in military veterans.The Veterans Health Administration (VHA) Innovation Ecosystem and the Food and Drug Administration (FDA) invited data analytics researchers to take part in the VHA Innovation Ecosystem and Precision FDA COVID-19 Risk Factor Modeling Challenge to develop machine learning-based analytical models to predict whether an individual would contract the virus based on the person’s health and medical history.
The competition also included predictions about whether a person could die of the virus, whether they would require ventilation, the length of their hospitalization, as well as the length of ICU care required.
Out of six judging criteria (deceased sensitivity, deceased specificity, ventilation sensitivity, ventilation specificity, days hospitalized RMSE, days hospitalized c-index, days in ICU RMSE, days in ICU c-index), Murtaza received two silver badges (second place), one bronze badge (third place), and placed fourth overall among all teams.
The models developed by Murtaza can predict patient death as well as ventilation requirements due to COVID-19 with an accuracy of more than 95 percent.
“Congratulations to Murtaza on his outstanding work,” says Manning School Dean Sandra Richtermeyer. “The COVID-19 pandemic is a health and economic crisis with severe impacts at the global, national and local levels. We’re so proud to see students like Murtaza applying their research to help during this difficult time — especially for our veterans.”
The VHA received a total of 34 entries from 21 unique teams. Participants included research teams of Ph.D. students and postdoctoral researchers from prestigious universities like MIT, the University of Minnesota, the University of South Florida, the University of Delaware and the University of Hawaii, as well as teams from well-known artificial intelligence and machine learning companies such as Teradata and Accenture.
The teams were required to develop machine learning models to predict outcomes for a population of 30,000 people, whose data was provided by the VHA. Entries were judged by a panel of VHA experts based on the accuracy of their predictions compared to the real outcomes.
Murtaza is advised by Asil Oztekin, an associate professor of Operations and Information Systems. Oztekin has established a strong working relationship with multi-disciplinary teams at the UMass Lowell Center for Population Health (CPH) and UMass Medical School’s Center for Data Driven Discovery in Healthcare (CD3H), through which he learned about this challenge.
Murtaza and Oztekin are now working on further analyses with these models and dataset to understand deeper relationships between different factors related to the outcomes of COVID-19. They will be submitting their findings to a prominent academic journal for review and publication consideration soon.