COVID-19 diagnosis prediction in emergency care patients: a machine learning approach

Andre Filipe de Moraes Batista; Joao Luiz Miraglia; Thiago Henrique Rizzi Donato; Alexandre Dias Porto Chiavegatto Filho

We measured predictive performance by calculating the area under the ROC curve (AUC), sensitivity, specificity, F1-score, Brier score, positive predictive value (PPV) and negative predictive value (NPV). All analyses were performed in Python using the scikit-learn library. The study was performed in accordance with the guidelines of Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) whenever applicable. 6 The study was approved by the Hospital Israelita Albert Einstein IRB (project number 4110-20) and by the National Commission for Ethics in Research (CONEP) of the National Health Council (CNS) from the Ministry of Health (CAAE: 30414720.0.0000.0071). Table 1 presents the descriptive results for the features included in the models, for all patients and separated according to COVID-19 diagnosis. The full sample was well balanced between males and females (51.1% and 48.9%, respectively), with a mean age of 49 years old. Within the COVID-19 positive group there were more men (65.7%), and lower mean values for leukocytes, lymphocytes, monocytes, basophil and eosinophils. Feature . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.