A-231 Prediction of Severe COVID-19 Based on Routine Biomarker Assessment

Abstract Background Disease courses in COVID-19 patients vary widely. Prior studies did not focus on initial values of various laboratory biomarkers, instead used data throughout the disease course. Prediction of disease severity based on parameters at the initial diagnosis would aid appropriate management. Objective of our study was to develop predictive models of COVID-19 severity based on demographic, clinical, and laboratory data collected at initial patient contact after diagnosis of COVID-19. Methods We evaluated de-identified demographic, clinical, and routine biomarkers data from 14 147 patients with COVID-19 diagnosed by polymerase chain reaction SARS-CoV-2 testing at Montefiore Health System from March 2020 to September 2021. We generated models predicting severe disease (death or more than 90 hospital days) vs mild disease (alive and fewer than 2 hospital days), starting with 58 parameters, by backward stepwise logistic regression. Results Of the 14 147 patients, 18% had severe, 24% had mild outcomes and the rest were moderate (cases who were alive but in the hospital for 2–89 days). We identified 4 proficient models in predicting patient outcomes: Inclusive model A, with the largest number (37) of parameters; Receiver Operating Characteristics (ROC) Model B, including 9 parameters with the highest Area under Curve (0.85); the Specific Model C, including 10 parameters with the highest specificity (81%) and Positive Predictive Value and the Sensitive Model D including 9 parameters with the highest sensitivity (91%) and Negative Predictive Value (Tables 1 and 2). The parameters that remained in all models were age, albumin, diastolic blood pressure, ferritin, lactic dehydrogenase, socioeconomic strength, procalcitonin, b-type natriuretic peptide, C reactive protein, d dimer and platelet count. Conclusions These findings suggest that certain clinical parameters and routine laboratory biomarkers within the specific and sensitive models could benefit healthcare providers, in their assessment of disease severity and clinical management of COVID-19 infection.