Early Stage Prediction of US County Vulnerability to the COVID-19 Pandemic

Mihir Mehta; Juxihong Julaiti; Paul Griffin; Soundar Kumara

The variable importance for the overlapping predictors between the final classification and regression models for March 16 th is shown in Figure 1 . Total population (TOT_POP) was the most important variable for both the classification and regression models. Other important variables included population density, longitude, hypertension prevalence, chronic respiratory mortality rate, cancer crude rate, and diabetes prevalence. Latitude (we use this to identify neighboring counties and the presence or absence of positive class in the neighborhood) and percentage of populations older than 70 years were found to be the least important features of those considered, though still played a role. Table 2) is given by percentage of counties that had no confirmed cases but were identified as being among the 5% most vulnerable had at least one confirmed COVID-19 case five days later. The specificity (Table 3) is given by the percentage . CC-BY-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.