Skip to main content

Table 2 Performance of neural network-based model for initial screening abnormal glucose tolerance

From: Using neural network as a screening and educational tool for abnormal glucose tolerance in the community

Group

Sensitivity (%)

Specificity (%)

Accuracy (%)

PNFT/TPR (%)

Cutoff point 1: 2 hPG# = 7.2 mmol/L

   

Test

90.0 (78.6 to 95.7)

47.7 (44.5 to 50.9)

49.8 (46.7 to 52.9)

54.2 (51.1 to 57.3)/5.0

Validation

90.0 (74.4 to 96.5)

50.4 (46.3 to 54.4)

52.3 (48.3 to 56.3)

51.7 (47.7 to 55.7)/5.0

Training

93.3 (88.2 to 96.3)

68.0 (54.2 to 79.2)

87.0 (81.6 to 91.0)

78.0 (71.8 to 83.2)/75.0

Cutoff point 2: 2 hPG = 7.6 mmol/L

   

Test

88.0(76.2 to 94.4)

51.6 (48.4 to 54.8)

53.4 (50.3 to 56.5)

50.4 (47.3 to 53.5)/5.0

Validation

90.0(74.4 to 96.5)

53.9 (49.8 to 57.9)

53.9 (51.7 to 59.6)

48.3 (44.4 to 52.3)/5.0

Training

90.7(85.0 to 94.4)

76.0 (62.6 to 85.7)

87.0 (81.6 to 91.0)

74.0 (67.5 to 79.6)/75.0

Operating cutoff point: 2 hPG = 7.8 mmol/L

   

Test

88.0(76.2 to 94.4)

54.3 (51.1 to 57.5)

56.0 (52.9 to 59.1)

47.8 (44.7 to 50.9)/5.0

Validation

86.7(70.3 to 94.7)

56.3 (52.2 to 60.3)

57.8 (53.8 to 61.7)

45.8 (41.9 to 49.8)/5.0

Training

86.7(80.3 to 91.2)

82.0 (69.2 to 90.2)

85.5 (80.0 to 89.7)

69.5 (62.8 to 75.5)/75.0

  1. Numbers in parentheses are 95% confidence intervals
  2. *: PNFT: Percentage that needed further test (including true positive and false positive)
  3. TPR: true positive rate in the studied group
  4. #: 2 hPG: 2-hour plasma glucose after oral glucose tolerance test