Skip to main content

Table 3 Performance metrics of the LACE model and machine learning models based on the testing set with fivefold cross-validation (Mean ± Standard Deviation, Unit: %)

From: Predictive modeling for 14-day unplanned hospital readmission risk by using machine learning algorithms

Model (#Features) Precision Recall F1-Score AUROC AUPRC
LACE (4) 2.97 ± 0.15 68.67 ± 3.86 5.70 ± 0.29 70.58 ± 1.88 34.63 ± 0.00
Logistic Regression: original features (70) 45.76 ± 15.72 4.00 ± 2.00 7.35 ± 3.59 80.46 ± 2.43 10.26 ± 2.23
Logistic Regression: original features (27) 43.62 ± 20.73 5.00 ± 1.05 8.84 ± 2.00 82.88 ± 3.57 11.66 ± 3.54
Random Forest: original features (70) 100.00 ± 0.00 41.33 ± 3.86 58.39 ± 3.79 97.89 ± 0.71 70.15 ± 4.23
Xgboost: original features (70) 93.23 ± 5.35 45.67 ± 3.89 61.25 ± 4.32 97.95 ± 0.52 66.52 ± 2.23
Catboost 1 (C1): original features (70) 93.77 ± 4.05 53.33 ± 5.27 67.80 ± 4.47 99.03 ± 0.07 75.15 ± 1.92
Catboost 2: features in C1 with importance > 0.5 (35) 95.12 ± 2.54 56.00 ± 5.33 70.29 ± 3.84 99.04 ± 0.09 76.11 ± 2.45
Catboost 3: features in C1 with importance > 0.6 (28) 95.09 ± 3.09 55.33 ± 5.31 69.74 ± 3.99 99.08 ± 0.08 76.69 ± 1.85
Catboost 4: features in C1 with importance > 0.8 (21) 94.70 ± 3.52 56.00 ± 6.02 70.10 ± 4.40 99.09 ± 0.08 77.11 ± 1.93
Catboost 5: features in C1 with importance > 0.9 (19) 93.20 ± 1.59 55.33 ± 5.72 69.29 ± 4.76 99.07 ± 0.10 76.80 ± 1.64
Catboost 6: features in C1 with importance > 1.1 (14) 91.46 ± 2.12 56.67 ± 4.47 69.86 ± 3.51 99.00 ± 0.11 76.97 ± 2.90
  1. AUROC = area under the receiver operating characteristic curve; AUPRC = area under the precision–recall curve