Skip to main content

Table 2 In-hospital mortality prediction on MIMIC-III. U, T, S represents unstructured data, temporal signals, and static information respectively

From: Combining structured and unstructured data for predictive models: a deep learning approach

  Model Model inputs F1 AUROC AUPRC P value
Baseline models LR \(T + S\) 0.341 (0.325, 0.357) 0.805 (0.799, 0.811) 0.188 (0.173, 0.203) 1
LR U 0.373 (0.358, 0.388) 0.825 (0.817, 0.833) 0.210 (0.200, 0.220) < 0.001
LR \(U + T + S\) 0.395 (0.380, 0.410) 0.862 (0.859, 0.865) 0.230 (0.217, 0.243) < 0.001
RF \(T + S\) 0.349 (0.325, 0.373) 0.735 (0.720, 0.750) 0.181 (0.157, 0.205) < 0.001
RF U 0.255 (0.236, 0.274) 0.665 (0.657, 0.673) 0.134 (0.126, 0.142) < 0.001
RF \(U + T + S\) 0.349 (0.331, 0.367) 0.735 (0.724, 0.746) 0.181 (0.163, 0.199) < 0.001
Deep models Fusion-CNN \(T + S\) 0.346 (0.330, 0.362) 0.827 (0.823, 0.831) 0.194 (0.184, 0.204) < 0.001
Fusion-CNN U 0.358 (0.341, 0.375) 0.826 (0.825, 0.827) 0.201 (0.198, 0.204) < 0.001
Fusion-CNN \(U + T + S\) 0.398 (0.378, 0.418) 0.870 (0.866, 0.874) 0.233 (0.220, 0.246) < 0.001
Fusion-LSTM \(T + S\) 0.374 (0.365, 0.383) 0.837 (0.834, 0.840) 0.211 (0.207, 0.215) < 0.001
Fusion-LSTM U 0.372 (0.352, 0.392) 0.828 (0.824, 0.832) 0.209 (0.207, 0.211) < 0.001
Fusion-LSTM \(U + T + S\) 0.424 (0.419, 0.429) 0.871 (0.868, 0.874) 0.250 (0.241, 0.259) < 0.001
  1. The bold in the table is maximum values of that evaluation metrics