Table 4.
Accuracy of different algorithms to detect POAF
| Sensitivity % (95% CI) | Specificity % (95% CI) | PPV % (95% CI) | NPV % (95% CI) | |
|---|---|---|---|---|
| Algorithm 1a | 70.4 (65.1–75.3) | 84.4 (81.3–87.1) | 69.1 (63.8–74.0) | 85.1 (82.2–87.8) |
| Algorithm 2b | 69.4 (64.1–74.4) | 85.7 (82.8–88.3) | 70.8 (65.4–75.7) | 85.0 (82.0–87.6) |
| Algorithm 3 | 70.4 (65.1–75.3) | 85.4 (82.5–88.1) | 70.6 (65.3–75.5) | 85.3 (82.4–87.9) |
| Algorithm 4 | 69.4 (64.1–74.4) | 86.7 (83.8–89.2) | 72.1 (66.8–77.0) | 85.1 (82.2–87.7) |
| Algorithm 5 | 70.4 (65.1–75.3) | 86.0 (83.1–88.6) | 71.5 (66.2–76.4) | 85.4 (82.5–88.0) |
| Algorithm 6 | 69.4 (64.1–74.4) | 87.3 (84.5–89.7) | 73.1 (67.8–77.9) | 85.2 (82.3–87.8) |
Abbreviations: AF Atrial fibrillation, NPV Negative predictive value, PPV Positive predictive value, 95% CI 95% confidence interval
aAlgorithm 1, 3, and 5 all included ICD codes I48.0, I48.1 and I48.90 to identify possible cases of POAF. They differed on the look-back window used to exclude patients with a history of AF: 1 year, 3 years, and 6 years for Algorithm 1, 3, and 5, respectively.
bAlgorithm 2, 4, and 6 all included ICD codes I48.0, I48.1 and I48.90 to identify possible cases of POAF. They differed on the look-back window used to exclude patients with a history of AF: 1 year, 3 years, and 6 years for Algorithm 2, 4, and 6. In these algorithms, all patients who received a maze procedure at the time of their cardiac surgery were considered POAF-negatives.