Table 3.
Performance of the baseline CP rule and rule combinations.
| Rule | Sensitivity (95%CI) |
PPV (95%CI) |
F1-Score (95%CI) |
|---|---|---|---|
| dx≥1 | 0.733 (0.607, 0.889) | 0.851 (0.684, 1.0) | 0.785 (0.667, 0.9) |
| dx≥1 ∣ ICS≥1 ∣ kw1≥1 ∣ kw2≥2 | 0.894 (0.733, 1.0) | 0.657 (0.45, 0.85) | 0.752 (0.571, 0.895) |
| dx≥1 & (ICS≥1 ∣ kw1≥1) | 0.726 (0.607, 0.864) | 0.904 (0.75, 1.0) | 0.803 (0.694, 0.905) |
| dx≥1 & (DX≥2 ∣ BDR≥2 ∣ ICS≥1 ∣ kw1≥1) | 0.728 (0.607, 0.864) | 0.886 (0.737, 1.0) | 0.797 (0.682, 0.9) |
| *dx≥2 | 0.661 (0.567, 0.76) | 0.972 (0.895, 1.0) | 0.785 (0.706, 0.864) |
| *IB≥2 ∣ ICS≥1 ∣ kw1≥1 | 0.709 (0.552, 0.882) | 0.689 (0.45, 0.851) | 0.695 (0.514, 0.842) |
| *dx≥2 & ICS≥1 | 0.635 (0.545, 0.741) | 0.981 (0.9, 1.0) | 0.77 (0.696, 0.851) |
| *CAPriCORN [11] | 0.678 (0.562, 0.8) | 0.918 (0.789, 1.0) | 0.778 (0.68, 0.878) |
| *PheKB [13] | 0.62 (0.5, 0.75) | 0.768 (0.579, 0.947) | 0.684 (0.55, 0.8) |
| *(dx≥2 ∣ dx≥1 & (BDR≥2 ∣ ICS≥1)) & kw1≥1 | 0.682 (0.577, 0.792) | 0.934 (0.8, 1.0) | 0.787 (0.683, 0.87) |
| *dx≥2 ∣ dx≥1 & (ICS≥1 ∣ BDR≥2 & kw1≥1) | 0.684 (0.586, 0.8) | 0.934 (0.8, 1.0) | 0.788 (0.696, 0.884) |
| *(dx≥2 ∣ dx≥1 & (ICS≥1 ∣ BDR≥2 & kw1≥2) | 0.685 (0.581, 0.81) | 0.94 (0.8, 1.0) | 0.791 (0.696, 0.878) |
| *(dx≥2 ∣ dx≥1 & (BDR≥2 ∣ ICS≥1)) & kw1≥1 | 0.691 (0.586, 0.818) | 0.942 (0.8, 1.0) | 0.796 (0.706, 0.889) |
Note: “&” represents “AND”, while “∣” represents “OR”, with “&” being evaluated first.
*
Each visit can count only 1 diagnosis (dx) or 1 medication (ICS or IB) or 1 keyword (kw).
Cells highlighted in yellow indicate the highest rating of the performance metric. Green highlighting shows the CP rule with the best overall balance of performance metrics.