Skip to main content

Table 2 Prediction results of mortality, readmission, sepsis, heart failure and next diagnoses

From: Interpretable time-aware and co-occurrence-aware network for medical prediction

Method Mortality Readmission Sepsis Heart Failure
ROC-AUC PR-AUC ROC-AUC PR-AUC ROC-AUC PR-AUC ROC-AUC PR-AUC
(a) Results of binary classification prediction
GRU 0.7902 0.7400 0.7023 0.6713 0.6202 0.6063 0.6525 0.6187
Med2Vec 0.8025 0.7950 0.7125 0.6833 0.8211 0.7943 0.7225 0.7101
Dipole 0.8133 0.8103 0.7341 0.7243 0.8001 0.7823 0.7067 0.6923
TLSTM 0.7893 0.7392 0.7256 0.7023 0.6432 0.6189 0.7432 0.6033
TCoN 0.8224 0.8134 0.7403 0.7278 0.8433 0.8233 0.7698 0.7313
  Accuracy@5 Accuracy@15 Accuracy@25 Accuracy@35
(b) Multi-classification result of next diagnoses prediction
GRU 0.7723 0.6298 0.5801 0.4523
Med2Vec 0.8025 0.7061 0.6250 0.5025
Dipole 0.8043 0.6514 0.6012 0.5044
TLSTM 0.7833 0.6367 0.5814 0.4515
TCoN 0.8398 0.7223 0.6577 0.5113