Clinical, laboratory data and inflammatory biomarkers at baseline as early discharge predictors in hospitalized SARS-CoV-2 infected patients

Abstract

Background The SARS-CoV-2 pandemic has overwhelmed hospital services due to the rapid transmission of the virus and its severity in a high percentage of cases. Having tools to predict which patients can be safely early discharged would help to improve this situation. Methods Patients confirmed as SARS-CoV-2 infection from four Spanish hospitals. Clinical, demographic, laboratory data and plasma samples were collected at admission. The patients were classified into mild and severe/critical groups according to 4-point ordinal categories based on oxygen therapy requirements. Logistic regression models were performed in mild patients with only clinical and routine laboratory parameters and adding plasma pro-inflammatory cytokine levels to predict both early discharge and worsening. Results 333 patients were included. At admission, 307 patients were classified as mild patients. Age, oxygen saturation, Lactate Dehydrogenase, D-dimers, neutrophil-lymphocyte ratio (NLR), and oral corticosteroids treatment were predictors of early discharge (area under curve (AUC), 0.786; sensitivity (SE) 68.5%; specificity (S), 74.5%; positive predictive value (PPV), 74.4%; and negative predictive value (NPV), 68.9%). When cytokines were included, lower interferon-?-inducible protein 10 and higher Interleukin 1 beta levels were associated with early discharge (AUC, 0.819; SE, 91.7%; S, 56.6%; PPV, 69.3%; and NPV, 86.5%). The model to predict worsening included male sex, oxygen saturation, no corticosteroids treatment, C-reactive protein and Nod-like receptor as independent factors (AUC, 0.903; SE, 97.1%; S, 68.8%; PPV, 30.4%; and NPV, 99.4%). The model was slightly improved by including the determinations of interleukine-8, Macrophage inflammatory protein-1 beta and soluble IL-2R? (CD25) (AUC, 0.952; SE, 97.1%; S, 98.1%; PPV, 82.7%; and NPV, 99.6%). Conclusions Clinical and routine laboratory data at admission strongly predict non-worsening during the first two weeks; therefore, these variables could help identify those patients who do not need a long hospitalization and improve hospital overcrowding. Determination of pro-inflammatory cytokines moderately improves these predictive capacities.