Skip to main content
. 2021 Jun 21;28(9):1936–1946. doi: 10.1093/jamia/ocab101

Table 3.

Wilcoxon signed rank test for pairwise comparison of performance obtained by ST, SB, and SeqSNR on the full and reduced feature sets.

Task Feature Set Pairwise Comparison P Value for AUPRC P Value for AUROC
AKI Full ST vs SB <.001 <.001
SeqSNR vs SB <.001 <.001
SeqSNR vs ST .002 <.001
Reduced ST vs SB <.001 <.001
SeqSNR vs SB <.001 <.001
SeqSNR vs ST .635 .060
CRRT dialysis Full ST vs SB .003 <.001
SeqSNR vs SB .035 .739
SeqSNR vs ST <.001 <.001
Reduced ST vs SB <.001 .019
SeqSNR vs SB .002 .679
SeqSNR vs ST <.001 .063
Vasoactive medications Full ST vs SB <.001 <.001
SeqSNR vs SB <.001 .888
SeqSNR vs ST <.001 <.001
Reduced ST vs SB .020 <.001
SeqSNR vs SB <.001 <.001
SeqSNR vs ST <.001 <.001
First MV Full ST vs SB .330 <.001
SeqSNR vs SB <.001 .006
SeqSNR vs ST <.001 <.001
Reduced ST vs SB .149 .005
SeqSNR vs SB <.001 .011
SeqSNR vs ST <.001 <.001
Mortality Full ST vs SB .081 <.001
SeqSNR vs SB .019 <.001
SeqSNR vs ST <.001 <.001
Reduced ST vs SB <.001 <.001
SeqSNR vs SB .203 <.001
SeqSNR vs ST <.001 <.001
Remaining LoS Full ST vs SB .021 <.001
SeqSNR vs SB .271 .021
SeqSNR vs ST .025 <.001
Reduced ST vs SB <.001 .186
SeqSNR vs SB <.001 <.001
SeqSNR vs ST <.001 <.001

To adjust for multiple hypothesis testing, we perform false discovery rate correction considering all experiments performed in this study and report the false discovery rate–adjusted P values. The P values marked in bold are statistically significant on the 95% confidence limit (α = 0.05).

AKI: acute kidney injury; AUPRC: area under the precision-recall curve; AUROC: area under the receiver-operating characteristic curve; CRRT: continuous renal replacement therapy; LoS: length of stay; MV: mechanical ventilation; SB: shared bottom; SeqSNR: sequential subnetwork routing; ST: single task.