Factors associated with hospitalization and critical illness among 4,103 patients with COVID-19 disease in New York City

Christopher M. Petrilli; Simon A. Jones; Jie Yang; Harish Rajagopalan; Luke F. O'Donnell; Yelena Chernyak; Katie Tobin; Robert J. Cerfolio; Fritz Francois; Leora I. Horwitz

was not yet determined. We excluded from this model four patients who expired in the emergency department before any information including vital signs could be collected. We obtained odds ratios from the models and bootstrapped confidence intervals for the odds ratios using the approach of Venables & Ripley, 17 since assuming normality of the maximum likelihood estimate to estimate confidence intervals can lead to biased estimates. 18 Finally, we constructed maximum information gain decision tree classifications for both hospital admission and severe complication to identify the variables that best classified patients into different outcome cohorts. For a given population, the decision tree classification method splits the population into two groups using one feature at a time, starting with the feature that maximizes the split between groups relative to the outcome in question. 19,20 Subsequent splits reevaluate each split subgroup for the next best feature. The final population in each end node . CC-BY-NC-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.