Skip to main content

View full-text article in PMC

. 2024 Mar 8;8:e45202. doi: 10.2196/45202

Table 4.

Performances of models with Medical Information Mart for Intensive Care–IV data set.

Model	MIMIC^a, RMSE^b (95% CI)	MIMIC, MAE^c (95% CI)	MIMIC, R² (95% CI)
PPK^d	22.22 (18.88 to 25.72)	12.13 (11.42 to 12.87)	−5.93 (−8.52 to –3.99)
XGBoost^e	8.93 (8.60 to 9.23)	6.93 (6.69 to 7.15)	–0.12 (−0.21 to 0.04)
TabNet	8.91 (8.53 to 9.25)	6.69 (6.45 to 6.93)	–0.11 (−0.17 to –0.06)
300-layer MLP^f	10.65 (8.88 to 12.72)	6.80 (6.48 to 7.13)	–0.59 (–1.30 to –0.12)
JointMLP^g (proposed)	8.53 (8.17 to 8.88)	6.40 (6.17 to 6.64)	–0.02 (−0.08 to 0.03)

^aMIMIC: Medical Information Mart for Intensive Care.

^bRMSE: root mean squared error.

^cMAE: mean absolute error.

^dPPK: population pharmacokinetic.

^eXGBoost: extreme gradient boosting.

^fMLP: multilayer perceptron.

^gJointMLP: joint multilayer perceptron.