Skip to main content

Table 4 Model performance by selection coverages and reduction rate

From: Confidence-based laboratory test reduction recommendation algorithm

Target Coverage

Model Coverage

Reduction Rate

Normality

Stability

Prev

AUC

Acc

Prec

AUPRC

Prev

AUC

Acc

Prec

AUPRC

0.75

77.92%

3.08%

10.09%

95.92%

95.16%

78.28%

81.52%

99.70%

96.91%

99.70%

99.70%

99.99%

0.80

81.64%

4.94%

12.24%

95.34%

94.49%

82.33%

80.78%

98.99%

95.99%

99.01%

99.02%

99.96%

0.85

83.80%

9.91%

15.25%

95.89%

93.17%

74.38%

80.05%

97.40%

95.94%

97.41%

97.46%

99.89%

0.90

89.84%

11.71%

17.28%

93.89%

89.47%

64.89%

71.12%

93.94%

96.06%

94.84%

97.21%

99.74%

0.95

96.91%

16.89%

19.07%

92.24%

86.76%

60.99%

67.56%

92.25%

95.19%

94.22%

95.77%

99.58%

1.0

99.99%

17.48%

19.51%

91.10%

85.64%

59.18%

63.43%

90.82%

94.94%

93.43%

95.36%

99.47%

  1. The model coverage rate refers to the actual proportion of Hgb samples considered by the model; the reduction rate refers to the proportion of Hgb samples recommended for reduction among the entire Hgb test dataset; Prev indicates prevalence, which denotes the proportion of normal/stable Hgbs in selective candidates; AUC indicates area under the ROC curve; Acc indicates accuracy; Prec indicates precision; AUPRC indicates the precision-recall curve