Table 3.
Predictive performance.
| Time point and algorithm | Outcome rate (%) | AUROC (95% CI) | Sensitivity (%) | Specificity (%) | Negative predictive value (%) | |
| Within 7 days of admission |
|
|
|
|
|
|
|
|
LASSOa logistic regression | 0.81 | 0.718 (0.686-0.750) | 65.91 | 64.24 | 99.57 |
|
|
Random forest |
|
0.692 (0.661-0.724) | 66.82 | 62.51 | 99.57 |
|
|
Hendrich II Fall Risk Model | 0.78 | 0.677 (0.658-0.696) | 53.81 | 74.52 | 99.51 |
| During entire hospital stay |
|
|
|
|
|
|
|
|
LASSO logistic regression | 1.36 | 0.726 (0.702-0.750) | 68.29 | 63.43 | 99.31 |
|
|
Random forest | 0.723 (0.698-0.747) | 69.11 | 62.87 | 99.33 | |
|
|
Hendrich II Fall Risk Model | 1.35 | 0.673 (0.659-0.687) | 52.43 | 74.05 | 99.13 |
aLASSO: least absolute shrinkage and selection operator.