Table 3.
Boosted Regression Tree Model Performance
| Model Performance | |||
|---|---|---|---|
| Performance Measures | Full Modela With Bootstrap Median, % (95% CI) | Simplified Modelb With Bootstrap Median, % (95% CI) | Minimal Modelc With Bootstrap Median, % (95% CI) |
| Sensitivity | 70.5 (56.0–76.5) | 69.7 (52.5–75.6) | 69.7 (51.7–75.7) |
| Specificity | 70.5 (56.8–75.9) | 69.8 (54.1–75.6) | 69.8 (51.2–75.6) |
| PCC | 70.5 (56.6–75.9) | 69.9 (53.7–75.5) | 69.7 (51.0–75.4) |
| AUC | 76.7 (61.0–82.7) | 76.0 (56.8–82.1) | 75.7 (56.1–82.1) |
Medians and 95% CIs are reported, based on 1000 bootstrap realizations using confirmed, probable, and suspected cases.Abbreviations: AUC, area under the receiver operating characteristic curve; BRT, boosted regression tree; CI, confidence interval; PCC, percentage correctly classified.
aThe full BRT model used all 43 candidate predictors.
bThe simplified BRT model used the 24 predictors retained after model simplification.
cThe minimal BRT model used the 10 most important predictors.