Table 2.
Multivariate analysis of the 4495 combinations of 31 regulatory molecules used to predict ICU admission/Exitus in COVID-19 patients. For each molecule, the percentage of models in which the given molecule was significant (at 10%) and the average AUC contribution to ICU/Exitus prognosis calculated according to the clinical score are indicated. Also indicated are the selected molecules for the final prognostic score.
| Group | Soluble Molecule | Models with Significance | Average AUC Gain | Inclusion in Final Model |
|---|---|---|---|---|
| Adaptive immune response |
IL10 | 98.2% | 2.0% | YES |
| IFNg | 69.0% | 1.3% | NO | |
| sIL2Ra | 40.0% | 0.5% | NO | |
| sIL4R | 12.9% | 0.2% | NO | |
| IL9 | 0.2% | 0.0% | NO | |
| Growth factors | GCSF | 100.0% | 4.3% | YES |
| MCSF | 96.3% | 2.7% | YES | |
| IL3 | 72.6% | 1.1% | NO | |
| IL2 | 86.9% | 1.0% | NO | |
| sEGFR | 56.6% | 0.7% | NO | |
| GMCSF | 37.2% | 0.3% | NO | |
| FLT3L | 15.2% | 0.1% | NO | |
| IL7 | 0.0% | 0.0% | NO | |
| sVEGFR3 | 0.0% | 0.0% | NO | |
| Innate/inflammatory immune response | IL6 | 100.0% | 3.5% | YES |
| IL15 | 100.0% | 3.5% | YES | |
| sRAGE | 100.0% | 3.4% | YES | |
| IP10 | 99.8% | 3.2% | YES | |
| MCP3 | 98.4% | 2.9% | YES | |
| sIL1RII | 100.0% | 2.4% | YES | |
| IL8 | 97.5% | 2.0% | YES | |
| MCP1 | 80.5% | 1.9% | NO | |
| TNFa | 77.2% | 1.8% | NO | |
| sTNFRII | 77.0% | 1.5% | NO | |
| MIG | 63.0% | 1.1% | NO | |
| IL1RA | 70.1% | 0.8% | NO | |
| MIP1a | 29.7% | 0.3% | NO | |
| sIL6R | 4.8% | 0.1% | NO | |
| MIP1b | 0.7% | 0.0% | NO | |
| IL27 | 0.7% | 0.0% | NO | |
| EOTAXIN | 0.2% | 0.0% | NO |