Table 3 |
||||||||
|
Accuracies and Predictive Values of Pooled Conventional versus A-ECG Criteria in the Test Set |
||||||||
|
Disease (N = 208) |
Healthy (N = 107) |
|||||||
|
TP |
FN |
TN |
FP |
Sensitivity(CLs) |
Specificity(CLs) |
Accuracy(CLs) |
||
|
|
||||||||
|
Conventional ECG status |
||||||||
|
Abnormal (optimized pooled criteria) |
163 |
45 |
91 |
16 |
78%(72-84%) |
85%(77-91%) |
81%(76-85%) |
|
|
Best individual parameter status* |
||||||||
|
Abnormal QTVI in lead II (>-1.64 units) |
161 |
47 |
87 |
20 |
77%(71-83%) |
81%(73-88%) |
79%(74-83%) |
|
|
Best 1° A-ECG scores status* "Full Disclosure" (5-min) A-ECG: |
||||||||
|
Abnormal 9-parameter score |
185 |
23 |
98 |
9 |
89%(84-93%)‡ |
92%(85-96%) |
90%(86-93%)† |
|
|
Abnormal 7-parameter score |
194 |
16 |
101 |
6 |
92%(88-96%)‡ |
94%(88-98%)† |
93%(90-96%)† |
|
|
"Snapshot" (10-sec) A-ECG: |
||||||||
|
Abnormal 7-parameter score |
192 |
16 |
92 |
15 |
92%(88-96%)‡ |
86%(78-92%) |
90%(86-93%)† |
|
|
Disease +LVSD (N = 41) |
Disease no LVSD (N = 44) |
|||||||
|
TP |
FN |
TN |
FP |
PPV (CLs) |
NPV (CLs) |
+LR |
-LR |
|
|
Conventional ECG status |
||||||||
|
Abnormal (optimized pooled criteria) |
39 |
2 |
10 |
34 |
53%(41-65%) |
83%(51-98%) |
1.23 |
0.21 |
|
Best individual parameter status* |
||||||||
|
Abnormal Z integral (>12.4 mV*ms) |
28 |
13 |
31 |
13 |
68%(52-82%) |
70%(55-83%) |
2.31 |
0.45 |
|
Best 2° A-ECG scores status* |
||||||||
|
Abnormal 5-parameter score |
34 |
7 |
41 |
3 |
92%(78-98%)‡ |
85%(72-94%) |
12.16 |
0.18 |
|
|
||||||||
|
A-ECG, advanced ECG; LVSD, left ventricular systolic dysfunction; TP and FP, true and false positives; TN and FN, true and false negatives; CLs, exact 95% binomial confidence limits; PPV and NPV, positive and negative predictive values; +LR and -LR, positive and negative likelihood ratios; QTVI, the QT interval variability index; Z integral, the total integral of the Z-lead QRS complex above 5 Hz. *The best individual parameters and their cut-off values and the best A-ECG scores are all carried over from the training set and are not necessarily fully optimized with respect to the test set. †P < 0.05 and ‡P < 0.0001 versus the optimized pooled conventional ECG criteria. |
||||||||
|
Schlegel et al. BMC Cardiovascular Disorders 2010 10:28 doi:10.1186/1471-2261-10-28 |
||||||||