Predictive Model for Hospitalization Due to COVID-19

• Activity Number: 90 - Novel Statistical Methods for COVID Pandemic and Other Current Health Policy Issues
• Type: Contributed
• Date/Time: Monday, August 9, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Fei Han* and Ian Stockwell and Morgan Henderson and Lucy Wilson and Zach Dezman
• Companies: University of Maryland Baltimore County and University of Maryland Baltimore County and University of Maryland Baltimore County and University of Maryland Baltimore County and University of Maryland, Baltimore
• Keywords: Predictive model; Hospitalization due to COVID-19; Ranking; Health care

Abstract

Hospitalization due to COVID-19 is considered a severe event. Researchers at UMBC and UMB developed a risk prediction model that helps health care practices identify patients who are at a high risk of having a hospitalization due to COVID-19.

We use discrete time survival models to predict the patient-level probability of incurring a hospitalization due to COVID-19 in the next month using diagnostic, care utilization, procedure, demographic, and environmental covariates from Medicare claims data. This model creates a near-to- real-time tool for identifying high-risk patients and is very valuable in population health. Interpretable output is generated from a discrete time survival model while using time-variant covariates. Model validation is measured by a concentration curve and shows high prediction power. For example, in January 2021, the top 1% predicted riskiest patients accounted for approximately 17% of the patients who were hospitalized due to COVID-19, and the top 10% predicted riskiest patients accounted for approximately 54% of the patients who were hospitalized due to COVID-19.