Prognostic Value of Radiological and Laboratory Biomarkers for Assessing Risk of Adverse Outcome in Patients with COVID-19

Objective: to study associations between laboratory and radiological biomarkers of COVID-19, to develop prognostic model of deterioration and lethal outcome in a patient with COVID-19.Material and methods. The study included 162 patients with COVID-19 stratified according to the presence or absence of deterioration during hospitalization. We evaluated chest computed tomography (CT) data, assessed empirically and using a semi-quantitative scale, blood cell counts and parameters of biochemical blood test. The predictive model was built using gradient boosting and artificial neural network with sigmoid activation function.Results. Both CT signs (crazy-paving pattern, bronchial dilatation inside a lesion, peripheral distribution of symptoms, absence of a predominant distribution pattern, lesion grade and extent), and most of laboratory markers were associated with deterioration and its criteria. The CT severity index correlated positively with the levels of leukocytes, neutrophils, urea, aspartate aminotransferase, lactate dehydrogenase, creatine phosphokinase, glucose, C-reactive protein, and negatively with the concentrations of albumin, calcium and the number of lymphocytes. Based on the results of the selection and training of classifying models, the optimal method for stratifying patients with COVID-19 on the basis of deterioration during hospitalization, the need for transfer to the intensive care unit, mechanical ventilation, and adverse outcome was gradient boosting.Conclusion. The prognostic model obtained in our study, based on a combination of radiological and laboratory parameters, makes it possible to predict the nature of COVID-19 course with high reliability.