Table 3.
Univariate analyses of the clinical features for the reduced multivariate viral pneumonia prediction model for data from the first 24 h after admission.
| Feature | Percentile# | Non-COVID-19 intensive care patients (training dataset) | COVID-19 intensive care patients (test dataset) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Survivors | Deceased | p value | Survivors | Deceased | p value | ||||||
| n | Median (IQR)/count (%) | n | Median (IQR)/count (%) | n | Median (IQR)/count (%) | n | Median (IQR)/count (%) | ||||
| Age [a] | 50th | 656 | 62.0 (48.0–72.0) | 62 | 66.5 (56.5–75.0) | 0.0043 | 961 | 66.0 (56.0–76.0) | 93 | 73.0 (65.0–82.0) | < 0.0001 |
| Vasoactive inotropic score | 50th | 656 | 4.0 (0.0–17.9) | 62 | 25.0 (6.0–56.0) | < 0.0001 | 961 | 0.0 (0.0–5.0) | 93 | 4.1 (0.0–25.0) | < 0.0001 |
| Thrombocytes [/nl] | 10th | 654 | 163.5 (94.2–245.8) | 62 | 80.5 (29.5–160.8) | < 0.0001 | 955 | 213.0 (158.5–286.5) | 93 | 178.0 (104.0–238.0) | < 0.0001 |
| Heart rate [1/min] | 10th | 650 | 81.8 (69.3–95.1) | 61 | 89.1 (74.1–100.1) | 0.031 | 959 | 70.0 (61.0–81.0) | 93 | 78.0 (64.0–85.0) | 0.0009 |
| Neutrophils (%) [%] | 10th | 309 | 80.4 (69.2–87.4) | 27 | 84.9 (79.1–92.9) | 0.0263 | 731 | 80.9 (73.5–86.3) | 55 | 84.0 (77.2–88.8) | 0.0066 |
| Urine output [ml/d] | 50th | 656 | − 1160.0 (− 2032.5–− 489.0) | 62 | − 845.0 (− 1750.0–− 238.8) | 0.0519 | 961 | − 1200.0 (− 1860.0–− 700.0) | 93 | − 750.0 (− 1320.0–− 200.0) | < 0.0001 |
| Mean arterial pressure [mmHg] | 10th | 650 | 66.8 (61.0–72.4) | 61 | 62.0 (56.5–65.9) | < 0.0001 | 959 | 68.0 (62.0–75.0) | 93 | 62.0 (57.0–66.0) | < 0.0001 |
| P/F ratio [mmHg] | 10th | 557 | 149.1 (106.1–199.2) | 58 | 112.1 (79.1–150.0) | < 0.0001 | 780 | 130.1 (98.8–173.1) | 76 | 106.6 (82.9–145.3) | 0.0001 |
| Respiratory Rate [1/min] | 90th | 637 | 22.5 (18.4–27.6) | 60 | 24.0 (20.0–29.1) | 0.2167 | 937 | 26.0 (22.0–30.0) | 93 | 26.0 (22.0–31.0) | 0.3362 |
| Alanine Aminotransferase [U/l] | 10th | 603 | 29.0 (16.0–54.5) | 61 | 35.0 (19.0–87.0) | 0.0387 | 944 | 30.5 (19.0–54.0) | 92 | 36.5 (22.0–60.8) | 0.0272 |
| pH | 50th | 648 | 7.4 (7.3–7.4) | 61 | 7.3 (7.3–7.4) | 0.0011 | 959 | 7.4 (7.4–7.5) | 93 | 7.4 (7.3–7.4) | 0.0003 |
| Ventilation mode | 10th | 631 | 2.0 (1.0–4.0) | 61 | 3.0 (2.0–4.0) | 0.0435 | 913 | 2.0 (1.0–4.0) | 92 | 2.0 (1.0–4.0) | 0.1234 |
| Monocytes (%) [%] | 90th | 242 | 3.1 (1.0–7.0) | 33 | 2.3 (1.9–4.6) | 0.2979 | 297 | 3.6 (1.2–5.5) | 44 | 2.8 (1.8–5.0) | 0.3044 |
| Pulmonary infiltrates | 10th | 656 | 101 (15%) | 62 | 19 (31%) | 0.004 | 961 | 391 (41%) | 93 | 43 (46%) | 0.3214 |
| Sodium [mmol/l] | 90th | 564 | 141.0 (138.0–144.0) | 57 | 141.0 (138.0–147.0) | 0.109 | 950 | 142.0 (139.0–145.0) | 93 | 143.0 (139.0–148.0) | 0.015 |
| Lactate dehydrogenase [U/l] | 50th | 485 | 356.0 (274.0–555.0) | 51 | 601.0 (377.5–878.5) | < 0.0001 | 854 | 427.0 (322.2–563.5) | 83 | 475.0 (368.0–720.0) | 0.0012 |
| Oxygen saturation [%] | 10th | 650 | 92.7 (90.4–94.7) | 61 | 90.7 (87.8–92.1) | < 0.0001 | 959 | 92.0 (90.0–94.0) | 93 | 89.0 (86.0–92.0) | < 0.0001 |
| Red blood cell distribution width [%] | 10th | 654 | 15.0 (13.7–17.1) | 62 | 16.0 (14.4–17.9) | 0.0118 | 954 | 13.9 (13.0–15.1) | 93 | 14.8 (13.6–16.3) | < 0.0001 |
| Ventilation peak pressure [mbar] | 50th | 656 | 15.0 (0.0–28.0) | 62 | 27.1 (0.0–31.0) | < 0.0001 | 961 | 0.0 (0.0–27.0) | 93 | 0.0 (0.0–28.0) | 0.0326 |
| Height [cm] | 50th | 603 | 172.0 (165.0–180.0) | 51 | 170.0 (165.0–178.5) | 0.2775 | 749 | 175.0 (166.0–180.0) | 57 | 175.0 (167.0–180.0) | 0.392 |
The table shows the selected features, the number (n) of surviving and deceased patients within the next 5 days after admission having the feature available in the dataset, the median value and interquartile range (IQR) for continuous features or count (percentage) for binary features over all patients of the respective group for both non-COVID-19 and COVID-19 patients. P-values are calculated using Mann–Whitney-U non-parametric tests (continuous features) or Fisher’s exact test (binary features) between survivors and non-survivors. #As data is aggregated over the first 24 h, the percentile denotes whether the 10th-, 50th- or 90th-percentile of values within the first 24 h were used.