Table 2.
Importance of variables in development data for each prediction model
| Variable importance rank | Logistic regression (deviance difference) | Random forest (mean decreased Gini) | Deep learning (difference in AUC) |
|---|---|---|---|
| 1 | Age | Age | Age |
| (2450) | (2135) | (0.069) | |
| 2 | T-wave axis | T-wave axis | R-wave axis |
| (521) | (1712) | (0.059) | |
| 3 | Height | QT interval | T-wave axis |
| (249) | (1270) | (0.022) | |
| 4 | Weight | Height | Weight |
| (147) | (1148) | (0.021) | |
| 5 | Presence of atrial fibrillation or flutter | R-wave axis | QRS duration |
| (144) | (1120) | (0.019) | |
| 6 | P-wave axis | P-wave axis | Height |
| (56) | (1043) | (0.017) | |
| 7 | QT interval | Weight | QT interval |
| (44) | (1009) | (0.016) | |
| 8 | QRS duration | QRS duration | P-wave axis |
| (32) | (997) | (0.015) | |
| 9 | PR interval | PR interval | PR interval |
| (24) | (988) | (0.014) | |
| 10 | R-wave axis | Heart rate | Heart rate |
| (15) | (893) | (0.012) | |
| 11 | Heart rate | Sex | Sex |
| (10) | (144) | (0.011) | |
| 12 | Sex | Presence of atrial fibrillation and flutter (132) | Presence of atrial fibrillation and flutter (0.005) |
| (1) |
AUC, area under the receiver operating characteristic curve.