Table 5.
Variables included in the DRF model after recursive feature elimination and their importance
| Variable | Relative importance | Scaled importance (0–1) |
|---|---|---|
| Posttreatment bicaudate index | 23.29 | 1.00 |
| EVD | 19.87 | 0.85 |
| Days with EVD | 16.20 | 0.70 |
| NIHSS on admission | 15.01 | 0.64 |
| Fisher | 11.21 | 0.48 |
| IVH sum score | 9.76 | 0.42 |
| Other Infections | 7.82 | 0.34 |
| IVH | 5.70 | 0.24 |
| WFNS | 4.80 | 0.21 |
| Age at SAH | 4.38 | 0.19 |
| mRS on admission | 3.41 | 0.15 |
| DCI | 2.85 | 0.12 |
| Aneurysm location | 2.15 | 0.09 |
| Hunt-Hess | 1.33 | 0.06 |
| KPS on admission | 0.93 | 0.04 |
| NIHSS motor on admission | 0.67 | 0.03 |
| Treatment timing from symptoms onset | 0.54 | 0.02 |
| Fever onset | 0.47 | 0.02 |
| Post-intervention ICU | 0.40 | 0.02 |
| ICH | 0.40 | 0.02 |
| ASA SCORE | 0.16 | 0.01 |
Values are determined according to how much the squared error over all trees improves after the single variables is selected for splitting on a decision tree