Table 5 The performances on the same test set of DeepMPM trained with different training sets
From: DeepMPM: a mortality risk prediction model using longitudinal EHR data
Training set description | AUC | Precision | Recall | F1-score |
|---|---|---|---|---|
Group I: all of patients were diagnosed with CHF | 0.7593 ± 0.0473 | 0.7533 ± 0.0378 | 0.8419 ± 0.0219 | 0.7853 ± 0.0145 |
Group II: Containing patients weren’t diagnosed with CHF | 0.8239 ± 0.0451 | 0.7984 ± 0.0098 | 0.8428 ± 0.0129 | 0.8127 ± 0.0503 |
Group I: all of patients were diagnosed with diabetes | 0.7468 ± 0.0255 | 0.7244 ± 0.0416 | 0.7184 ± 0.0214 | 0.6913 ± 0.0087 |
Group II: containing patients weren’t diagnosed with diabetes | 0.8014 ± 0.0418 | 0.7516 ± 0.0281 | 0.7773 ± 0.0573 | 0.7562 ± 0.0151 |