Diagnosing AF with and without clinician over-read of unclassified readings, against ECG as gold standard
KardiaBand | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | PLR | NLR | Accuracy (%) | Kappa | 95% CI | P value |
Consecutive unclassified excluded | 94.4 | 90.7 | 72.3 | 98.4 | 10.2 | 0.06 | 91.5 | 0.76 | 0.67 to 0.88 | <0.001 |
Consecutive unclassified marked as incorrect | 94.4 | 81.9 | 54.8 | 98.4 | 5.2 | 0.07 | 84.3 | 0.60 | 0.47 to 0.72 | <0.001 |
KB +appropriate unclassified diagnosis | 95.4 | 81.9 | 59.4 | 98.5 | 5.2 | 0.06 | 84.9 | 0.71 | 0.62 to 0.77 | <0.001 |
KB +EP1 review (only unclassified) | 94.4 | 91.0 | 70.8 | 98.6 | 10.5 | 0.06 | 91.7 | 0.76 | 0.65 to 0.87 | <0.001 |
KB +EP2 review (only unclassified) | 94.4 | 90.4 | 69.4 | 98.6 | 9.8 | 0.06 | 91.2 | 0.74 | 0.63 to 0.86 | <0.001 |
Detailed analysis of KB accuracy for diagnosing AF with and without clinician interpretation of unclassified tracings, against cardiologist ECG diagnosis as gold standard. When the initial iECG gave an unclassified reading, a repeat iECG was taken and used as the final diagnosis. An appropriate unclassified diagnosis was determined if the tracing was due to a non-AF arrhythmia. P values are derived for the kappa coefficient.
AF, atrial fibrillation; EP, electrophysiologist;iECG, intelligent ECG; KB, KardiaBand; NLR, negative likelihood ratio; NPV, negative predictive value; PLR, positive likelihood ratio; PPV, positive predictive value.