Table 3 Comparison of the characteristics of the baseline methods with DeepMPM
From: DeepMPM: a mortality risk prediction model using longitudinal EHR data
Model | RNN architecture | Reverse time training | Attention mechanism | Visit-level attention | Variable-level attention |
|---|---|---|---|---|---|
RNN | \(\surd\) | \(\times\) | \(\times\) | \(\times\) | \(\times\) |
Multi-task Learning | \(\surd\) | \(\times\) | \(\surd\) | \(\times\) | \(\times\) |
LSTM-NN | \(\surd\) | \(\times\) | \(\surd\) | \(\times\) | \(\times\) |
RETAIN | \(\surd\) | \(\surd\) | \(\surd\) | \(\surd\) | \(\surd\) |
Deepcare | \(\surd\) | \(\times\) | \(\times\) | \(\times\) | \(\times\) |
DeepMPM-w/o-\(\beta\) | \(\surd\) | \(\times\) | \(\surd\) | \(\surd\) | \(\times\) |
DeepMPM | \(\surd\) | \(\times\) | \(\surd\) | \(\surd\) | \(\surd\) |