Distribution and accuracy of selected mortality prediction tools
| Prediction tool | Total (N=166) | Died (n=47) | Survived (n=119) | Original AUC (95% CI) | Observed AUC (95% CI) |
| Clinical gestalt, confidence (IQR) | 30 (20–50) 80 (70–90) | 40 (30–70) 80 (60–90) | 30 (15–40) 80 (70–90) | – | 0.68 (0.59 to 0.77) |
| LOW-HARM (IQR) | 46 (8.4–83.8) | 86 (37.5–99.3) | 37.5 (6.4–69) | 0.96 (0.94 to 0.98) | 0.76 (0.69 to 0.84) |
| LOW-HARM V.2 (IQR) | 9.7 (0.9–52.7) | 49 (9.7–96.3) | 3.2 (0.5–28.1) | – | 0.78 (0.71 to 0.86) |
| NUTRI-CoV (IQR) | 9 (7–12) | 10 (8–12) | 9 (7–11) | 0.79 (0.76 to 0.82) | 0.60 (0.51 to 0.69) |
| MSL-COVID-19 (IQR) | 8 (7–10) | 8 (8–10) | 8 (7–9) | 0.72 (0.69 to 0.75) | 0.64 (0.55 to 0.73) |
| qSOFA (IQR) | 1 (1–1) | 1 (1–2) | 1 (1–1) | 0.74 (0.65 to 0.81) | 0.61 (0.53 to 0.69) |
| NEWS2 (IQR) | 7.5 (6–9) | 9 (7–10) | 7 (5–9) | 0.84 (0.79 to 0.90) | 0.65 (0.56 to 0.75) |
| Neutrophil to lymphocyte ratio >9.8 (%) | 64.4 | 55.3 | 27.7 | 0.74 (0.62 to 0.85) | 0.65 (0.57 to 0.73) |
Overall comparison test for observed AUC=0.002.
Individual AUROC comparisons: clinical gestalt vs all scores, p>0.05.
To calculate the relative mean difference, some scores (those not based on 100 points) were converted to a percentage in the following manner: (patient score/maximum possible score)×100.
AUC, area under the curve.