Transferability and interpretability of the sepsis prediction models in the intensive care unit

Table 1 The results of different models on the MIMIC-III dataset

Model	Preceding hours	Accuracy	AUC	Sensitivity	Specificity
InSight	4	0.57	0.74	0.8	0.54
AISE	4	0.64	0.84	0.85	0.64
MGP-TCN	4	–	Approximately 0.85	–	–
DTW-KNN	4	–	Approximately 0.88	–	–
MLA	0	–	0.88	0.8	0.78
MLA	24	–	0.84	0.8	0.72
DSPA	4	–	0.98	–	–
MGP-AttTCN	4	–	0.75	–	–
NAVOY Sepsis	3	0.81	0.84	0.74	0.83
LightGBM	4	0.91	0.98	0.85	0.97
MLP	4	0.85	0.96	0.73	0.96

AISE Artificial Intelligence Sepsis Expert, MGP-TCN Multi-task Gaussian Process and Temporal Convolutional Networks, DTW-KNN Dynamic Time Warping and K-Nearest Neighbours, MLA Machine Learning Algorithm, DSPA Deep SOFA-Sepsis Prediction Algorithm, MGP-AttTCN Multi-task Gaussian Process and Attention Time Convolutional Network

ISSN: 1472-6947