Table 2.
Performance measure of our prediction models developed for the HF cohort
| Model | AUC | AUPRC | Accuracy (%) | Sensitivity (%) | Specificity (%) |
|---|---|---|---|---|---|
| LACE | 0.551 | 0.448 | 59.85 | 45.54 | 64.80 |
| Logistic regression | 0.576 | 0.455 | 62.26 | 48.37 | 66.85 |
| Random forests | 0.501 | 0.319 | 76.39 | 0.52 | 99.75 |
| Weighted random forests | 0.548 | 0.386 | 76.22 | 21.71 | 88.07 |
| Decision trees | 0.520 | 0.367 | 66.97 | 22.84 | 81.22 |
| Weighted decision trees | 0.528 | 0.379 | 64.18 | 31.44 | 74.16 |
| Support‐vector machines | 0.528 | 0.367 | 71.80 | 16.03 | 89.76 |
| Weighted support‐vector machines | 0.535 | 0.377 | 65.36 | 31.39 | 75.78 |
| Multilayer perceptron | 0.628 | 0.461 | 64.93 | 48.42 | 70.01 |
AUC, area under the receiver operating characteristic curve; AUPRC: area under the precision–recall curve; LACE, length of stay (L), acuity of admission (A), Charlson Comorbidity Index score (C), and number of emergency visits in the last 6 months (E).