Detection of Risk Predictors of COVID-19 Mortality with Classifier Machine Learning Models Operated with Routine Laboratory Biomarkers

Early evaluation of patients who require special care and have high death-expectancy in COVID-19, the effective determination relevant biomarkers on large sample-groups are important to reduce mortality. This study aimed reveal routine blood-value predictors COVID-19 mortality determine lethal-risk levels these during disease process. The dataset consists 38 blood-values 2597 died (n = 233) those recovered 2364) from August–December, 2021. In this study, histogram-based gradient-boosting (HGB) model was most successful machine-learning classifier detecting living deceased (with squared F1 metrics F12 1). efficient binary combinations with procalcitonin were obtained D-dimer, ESR, D-Bil ferritin. HGB operated feature pairs correctly detected almost all survived (precision > 0.98, recall 0.98). Furthermore, a single feature, features (F12 0.96) ferritin 0.91). addition, according two-threshold approach, values between 376.2 μg/L 396.0 0.91) 0.2 5.2 0.95) found be fatal risk for COVID-19. Considering results, we suggest that many combined features, especially ferritin, model, can used achieve very results classification live, die Moreover, strongly recommend clinicians consider critical properties, lethality disease.