Table 4.
Multivariable survival model using categorical variables for predicting mortality
| Variablesb | Log-HR (β) | SE | p value | Assigned pointsa |
|---|---|---|---|---|
| Age | ||||
| 50 to 65 years | 0.58 | 0.26 | 0.02 | 6 |
| 65 to 80 years | 1.15 | 0.25 | < 0.001 | 12 |
| ≥ 80 years | 1.96 | 0.27 | < 0.001 | 20 |
| Female | 0.14 | 0.13 | 0.27 | 1 |
| Race | ||||
| Black/African American | − 0.36 | 0.13 | 0.01 | − 4 |
| Other | − 0.52 | 0.25 | 0.04 | − 5 |
| BMI ≥ 30 kg/m2 | 0.14 | 0.13 | 0.26 | 1 |
| Comorbidity | ||||
| Coronary artery disease | 0.27 | 0.17 | 0.10 | 3 |
| Diabetes Mellitus | 0.09 | 0.13 | 0.49 | 1 |
| Hypertension | − 0.12 | 0.14 | 0.38 | − 1 |
| Respiratory rate ≥ 24 breaths per minute | 0.50 | 0.12 | < 0.001 | 5 |
| Blood oxygen saturation | ||||
| ≤ 88% | 0.62 | 0.16 | < 0.001 | 6 |
| 88 + to 94− % | 0.29 | 0.17 | 0.09 | 3 |
| White blood cell count | ||||
| < 4 × 109/L | 0.07 | 0.21 | 0.74 | 1 |
| > 10 × 109/L | − 0.24 | 0.16 | 0.13 | − 2 |
| Hemoglobin ≤ 11 g/dL | 0.34 | 0.15 | 0.03 | 3 |
| ALT > 40 U/L | 0.36 | 0.13 | 0.01 | 4 |
| Creatinine > 1.33 mg/dL | 0.30 | 0.14 | 0.03 | 3 |
| D-dimer | ||||
| 500–1000 ng/mL FEU | 0.07 | 0.26 | 0.78 | 1 |
| > 1000 ng/mL FEU | 0.10 | 0.27 | 0.71 | 1 |
| Procalcitonin | ||||
| 0.1–0.25 ng/mL | 0.22 | 0.23 | 0.33 | 2 |
| 0.25–0.5 ng/mL | 0.35 | 0.25 | 0.16 | 4 |
| > 0.5 ng/mL | 0.79 | 0.23 | < 0.001 | 8 |
| Lactic acid ≥ 2 mmol/L | 0.46 | 0.15 | 0.002 | 5 |
ALT alanine aminotransferase, HR hazard ratio, SE standard error
aEstimates of regression coefficient (β) were pooled from survival models fitted on 20 imputed datasets. Points associated with the presence of a given level of a risk factor (variable) were assigned by multiplying β by 10 and rounding to the nearest integer. The reference level was assigned zero point. The theoretical ranges of total score were − 8 and 62
bUsing bootstrap cross-validation with 1000 bootstrap samples per imputed dataset, the estimate of the concordance index (c-statistics) was 0.78 (95% CI 0.75–0.81)