Machine Learning Successfully Detects COVID-19 Patients Prior to PCR Results and Predicts Their Survival Based on Standard Laboratory Parameters

Background: In the current pandemic, oversaturation of hospitals with patients with SARS-CoV-2 infection-like symptoms and an excess of hospitalized COVID-19 patients led to a global healthcare crisis. The aim of our study was to find a manageable set of decisive parameters that can be used to i) rapidly identify SARS-CoV-2 positive patients, ii) identify high-risk patients, and iii) recognize longitudinal warning signs of a possible fatal outcome.Methods: We trained several machine learning (ML) models using data on reported comorbidities, medications, symptoms, and laboratory parameters on hospital admission, and over the disease course in 201 SARS-CoV-2 positive and 314 SARS-CoV-2 negative subjects with a COVID-19-like clinical presentation.Findings: We identified a set of eight on-admission parameters: white blood cells, antibody-synthesizing lymphocytes, ratios of basophils/lymphocytes, platelets/neutrophils, and monocytes/lymphocytes, procalcitonin, creatinine, and C-reactive protein. The medical decision tree built using these parameters differentiated between SARS-CoV-2 positive and negative patients with up to 90-100% accuracy. Next, we determined that COVID-19 patients who on hospital admission were older, had higher procalcitonin, C-reactive protein, and troponin I together with lower hemoglobin and platelets/neutrophils ratio, were at highest risk of death from COVID-19. Finally, we identified patterns of changes in C-reactive protein, white blood cells, and D-Dimer that predicted the disease outcome.Interpretation: Our study provides sets of easily obtainable parameters that allow to assess a SARS-CoV-2 patient’s status prior to RT-PCR results and the dynamics of the disease, based on which the hospital logistics and treatment can be planned.Funding Information: This work has been supported by the Medical University of Lodz, Poland, by the Swiss Institute of Allergy and Asthma Research (SIAF), and by the Center for Data Analysis, Visualization and Simulation (DAViS) that is funded by the Swiss canton of Grisons. Declaration of Interests: MiSo reports research grants from Swiss National Science Foundation, GSK, Novartis and speakers fee from AstraZeneca and a board secretary position of the Basic and Clinical Immunology Section of the European Academy of Allergy and Clinical Immunology (EAACI). JM, FS, DZ, MaS, and KB report no competing interests.Ethics Approval Statement: The project has been accepted by the Ethical Committee of Medical University of Lodz, Poland (Nr. 228 RNN/126/20/KE).