Building a COVID-19 vulnerability index
Dave DeCaprio, Joseph Gartner, Carol J. McCall, Thadeus Burgess, Kristian Garcia, Sarthak Kothari, Shaayaan Sayed
Abstract
Background: Coronavirus disease 2019 (COVID-19) is an acute respiratory disease that has been classified as a pandemic by the World Health Organization (WHO). Characterization of this disease is still in its early stages; however, it is known to have high mortality rates, particularly among individuals with preexisting medical conditions. Creating models to identify individuals who are at the greatest risk for severe complications due to COVID-19 will be useful for outreach campaigns to help mitigate the disease’s worst effects. Methods: We present the results for three models predicting such complications, with each model representing different tradeoffs between prediction accuracy and ease of implementation. To overcome a lack of validated COVID-19 case data, the models were trained using a proxy endpoint of complications due to other upper respiratory infections. The best performing model was validated against actual COVID-19 hospitalizations. Results: The survey risk factors model can be widely used because it is easy to implement and uses only a simple health history survey. It provides improved accuracy over a baseline Charlson comorbidity score, identifying 49.8% of vulnerable patients in the top 5% of the population. The diagnosis history and expanded features models are progressively harder to implement but provide improved accuracy (53.8% & 54.1% sensitivity at a 5% alert rate respectively) relative to the survey risk factors model. In validation on a Medicare population, the top 10% of patients predicted from the expanded features model had a mortality rate three times that of the full population. Conclusions: These models have been released as an open source package and a web-based survey. They are in use by dozens of organizations on millions of individuals. Having alternative models allows users to determine the balance of ease of implementation and overall accuracy that is most appropriate for their needs and the data they have available.