Clinical Prediction on MIMIC-III
94.1AUROCFeatEHR-LLM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| FeatEHR-LLMFeatures=Baseline + FeatEHR-LLM2026.04 | 94.1 | — | — | — | |
| TFFeatures=Baseline + TF2026.04 | 92.6 | — | — | — | |
| OpenFEFeatures=Baseline + OpenFE2026.04 | 91.2 | — | — | — | |
| BaselineFeatures=BL2026.04 | 91 | — | — | — | |
| FeatLLMFeatures=Baseline + FeatLLM2026.04 | 90.4 | — | — | — | |
| CAAFEFeatures=Baseline + CAAFE2026.04 | 90.4 | — | — | — | |
| TreeText-CTS2026.05 | 85.79 | — | 50.11 | 49.58 | |
| ODE-RNNtime per epoch=16.52021.01 | 85.61 | — | — | — | |
| L-ODE-ODEtime per epoch=222021.01 | 85.59 | — | — | — | |
| mTAND-Fulltime per epoch=0.22021.01 | 85.44 | — | — | — | |
| GRU-D2026.05 | 85.41 | — | 50.08 | 48.84 | |
| GRU-DModel Category=Proposed2016.06 | 85.27 | — | — | — | |
| RAM-EHRFull Model=true2024.02 | 84.97 | 81.59 | 74.64 | 48.19 | |
| SeFTtime per epoch=0.52021.01 | 84.85 | — | — | — | |
| DuETT2026.05 | 84.6 | — | 45.76 | 48.2 | |
| STraTS2026.05 | 84.59 | — | 47.05 | 47.16 | |
| Phased-LSTMtime per epoch=0.32021.01 | 84.29 | — | — | — | |
| KEDGN2026.05 | 84.25 | — | 47.34 | 48.19 | |
| mTAND-Enctime per epoch=0.12021.01 | 84.19 | — | — | — | |
| mTAND Transformer2026.05 | 83.99 | — | 46.03 | 47.53 | |
| DecodeInterface=Faithful2026.05 | 83.94 | — | 42.92 | 20.77 | |
| RNN Decaytime per epoch=0.72021.01 | 83.92 | — | — | — | |
| IP-Netstime per epoch=1.32021.01 | 83.9 | — | — | — | |
| RAM-EHRMode=Augmented Model g_theta Only2024.02 | 83.8 | 81.04 | 73.41 | 46.83 | |
| GRU-simpleModel Category=RNN2016.06 | 83.8 | — | — | — | |
| RNN-Delta-ttime per epoch=0.52021.01 | 83.64 | — | — | — | |
| RAM-EHRRetrieval=false2024.02 | 83.29 | 80.68 | 72.95 | 44.65 | |
| RF-simpleModel Category=Non-RNN2016.06 | 82.94 | — | — | — | |
| RF-forwardModel Category=Non-RNN2016.06 | 82.93 | — | — | — | |
| RNN GRU-Dtime per epoch=0.72021.01 | 82.7 | — | — | — | |
| BEEP2024.02 | 82.67 | 79.9 | 71.58 | 44.15 | |
| GRU-forwardModel Category=RNN2016.06 | 82.52 | — | — | — | |
| RNN-Imputetime per epoch=0.52021.01 | 82.49 | — | — | — | |
| LOCF LSTM2026.05 | 82.4 | — | 48.11 | 47.06 | |
| LOCF Transformer2026.05 | 82.3 | — | 44.65 | 46.41 | |
| GraphCare2024.02 | 82.26 | 80.11 | 71.19 | 44.33 | |
| RAM-EHRLLM Summarization=false2024.02 | 82.14 | 80.08 | 71.35 | 41.49 | |
| CORE2024.02 | 82.05 | 79.63 | 70.79 | 43.76 | |
| GRU-meanModel Category=RNN2016.06 | 81.92 | — | — | — | |
| SVM-simpleModel Category=Non-RNN2016.06 | 81.46 | — | — | — | |
| SEFT2026.05 | 81.46 | — | 40.34 | 43.84 | |
| LSTM-meanModel Category=RNN2016.06 | 81.42 | — | — | — | |
| HyGT2024.02 | 81.09 | 78.07 | 68.08 | 44.93 | |
| SeqCare2024.02 | 80.98 | 77.44 | 68.53 | 39.16 | |
| TimeCPInterface=Adapted2026.05 | 80.65 | — | 41.1 | 44.42 | |
| Transformer2024.02 | 80.61 | 76.18 | 67.12 | 42.75 | |
| Record2VecInterface=Adapted2026.05 | 80.4 | — | 36.92 | 41.67 | |
| MedRetriever2024.02 | 80.14 | 77.15 | 68.45 | 39.29 | |
| DecodeInterface=Adapted2026.05 | 79.33 | — | 36.29 | 42.23 | |
| SVM-forwardModel Category=Non-RNN2016.06 | 79.08 | — | — | — | |
| GCT2024.02 | 78.62 | 77.2 | 64.87 | 37.57 | |
| WRDPInterface=Adapted2026.05 | 78.43 | — | 35.8 | 40.46 | |
| Record2VecInterface=Faithful2026.05 | 78.39 | — | 37.38 | 40.4 | |
| L-ODE-RNNtime per epoch=6.72021.01 | 77.34 | — | — | — | |
| LR-simpleModel Category=Non-RNN2016.06 | 77.15 | — | — | — | |
| LR-forwardModel Category=Non-RNN2016.06 | 75.89 | — | — | — | |
| WRDPInterface=Faithful2026.05 | 75.43 | — | 29.58 | 35.79 | |
| TimeCPInterface=Faithful2026.05 | 61.2 | — | 14.53 | 26.41 | |
| RNN-VAEtime per epoch=22021.01 | 51.75 | — | — | — | |
| Basetraining_mode=trained on source data2025.06 | — | — | 68.83 | 57.31 | |
| CondAdvtraining_mode=trained on source data2025.06 | — | — | 76.59 | 64.42 | |
| DANNtraining_mode=trained on source data2025.06 | — | — | 75.18 | 60.55 | |
| IRMtraining_mode=trained on source data2025.06 | — | — | 69.47 | 62.2 | |
| ManyDGtraining_mode=trained on source data2025.06 | — | — | 77.29 | 63.7 | |
| MLDGtraining_mode=trained on source data2025.06 | — | — | 74.76 | 58.85 | |
| Oracletraining_mode=trained on target data2025.06 | — | — | 80.43 | 67.41 | |
| UDONCAREtraining_mode=trained on source data2025.06 | — | — | 78.16 | 66.7 |