Table 1.

Results of model discrimination and calibration for all task settings on the test data. These correspond to internal validation for PhysioNet 2012 Challenge and external for PhysioNet 2019 Challenge.

	Masking (AUROCa), mean (SD)	Masking (Brier), mean (SD)	No masking (AUROC), mean (SD)	No masking (Brier), mean (SD)
P12b mortality	0.842 (0.82-0.86)	0.093 (0.087-0.100)	0.830 (0.81-0.85)	0.095 (0.088-0.101)
P12 LOSc	0.814 (0.79-0.84)	0.054 (0.049-0.060)	0.737 (0.71-0.77)	0.064 (0.058-0.070)
P19d sepsis-overall	0.907 (0.90-0.92)	0.039 (0.036-0.041)	0.889 (0.88-0.90)	0.045 (0.043-0.048)
P19 sepsis-frequent	0.757 (0.74-0.77)	0.014 (0.013-0.014)	0.766 (0.75-0.78)	0.014 (0.013-0.015)

^aAUROC: area under the curve of the receiver operating characteristic.

^bP12: PhysioNet 2012 Challenge.

^cLOS: length of stay.

^dP19: PhysioNet 2019 Challenge.