Table 6.
Feature Importance Given by the Top 3 Classifiers, Evaluated Over 100 Train-Test Data Splits for Model III.a
| Importance (classifiers) | Rank | |||||
|---|---|---|---|---|---|---|
| Features | KNN | NB | LC | KNN | NB | LC |
| Age | 0.57 (0.03) | 0.57 (0.03) | 0.49 (0.37) | 4 | 4 | 8 |
| Gender | 0.54 (0.03) | 0.54 (0.02) | 0.97 (0.53) | 5 | 5 | 6 |
| Admi_weight | 0.53 (0.02) | 0.53 (0.02) | 1.45 (0.58) | 6 | 6 | 4 |
| Admi_APACHE | 0.76 (0.03) | 0.76 (0.03) | 3.61 (0.25) | 1b | 1b | 1b |
| Tot_orders_24h | 0.60 (0.03) | 0.60 (0.03) | 0.49 (0.44) | 3b | 3b | 7 |
| Tot_medica_24h | 0.53 (0.02) | 0.53 (0.02) | 1.26 (0.57) | 7 | 7 | 5 |
| Mecha_ventilation | 0.53 (0.02) | 0.53 (0.02) | 2.25 (0.41) | 8 | 8 | 2b |
| MRC_score_24h | 0.65 (0.03) | 0.65 (0.03) | 1.82 (0.42) | 2b | 2b | 3b |
Abbreviations: 24h, 24 hours; Admi, admission; KNN, k-nearest neighbor; LC, logistic classifier; Mecha, mechanical; MRC-ICU, Medication Regimen Complexity–intensive care unit; NB, naive Bayes; Tot_medica, total medication.
a
Importance values are presented as raw values as mean (SD).
b
Top 3 features (1, 2, and 3, respectively).