Table 3 Subgroup-Specific ROC of the basic model and modified model in the validation dataset.

From: Low serum calcium promotes traumatic intracerebral hematoma expansion by the response of immune cell: A multicenter retrospective cohort study

Subgroup

Basic model

Modified model

AUC (95% CI)

Sensitivity (%)

Specificity (%)

AUC (95% CI)

Sensitivity (%)

Specificity (%)

Male

0.826 (0.768, 0.884)

67.00

85.37

0.871 (0.821, 0.922)

83.00

76.83

Female

0.823 (0.711, 0.936)

92.59

61.54

0.883 (0.792, 0.974)

88.89

76.92

<65 years

0.830 (0.770, 0.890)

70.93

83.91

0.857 (0.803, 0.912)

86.05

71.26

≥ 65 years

0.923 (0.858, 0.988)

75.61

100.00

0.949 (0.894, 1.000)

95.12

85.71

Mild TBI

0.798 (0.728, 0.868)

63.51

85.33

0.841 (0.778, 0.904)

78.38

76.00

Moderate and Severe TBI

0.899 (0.834, 0.965)

92.45

72.73

0.930 (0.876, 0.984)

88.68

87.88

Baseline tICH volume, < 5 mL

0.805 (0.741, 0.869)

67.05

82.76

0.836 (0.777, 0.896)

72.73

83.91

Baseline tICH volume, ≥ 5 mL

0.974 (0.942, 1.000)

74.36

95.24

0.974 (0.942, 1.000)

94.87

95.24

MLR< 0.89

0.802 (0.737, 0.867)

53.95

93.75

0.847 (0.788, 0.905)

85.53

71.88

MLR≥ 0.89

0.843 (0.739, 0.947)

62.75

100.00

0.907 (0.823, 0.991)

76.47

91.67

Coagulopathy

0.843 (0.765, 0.921)

67.31

95.12

0.878 (0.809, 0.947)

78.85

87.80

Normal coagulation function

0.827 (0.759, 0.895)

72.00

85.07

0.868 (0.811, 0.926)

70.67

89.55

Normocalcemia

0.857 (0.796, 0.917)

69.57

84.88

0.857 (0.796, 0.917)

78.26

81.40

Hypocalcaemia

0.743 (0.663, 0.822)

52.17

90.70

0.814 (0.746, 0.882)

72.46

77.91

SDH (Present, No)

0.788 (0.721, 0.855)

62.82

80.85

0.842 (0.782, 0.902)

67.95

89.36

SDH (Present, Yes)

0.909 (0.824, 0.994)

90.00

78.57

0.914 (0.837, 0.992)

72.00

92.86

  1. MLR, monocyte-to-lymphocyte ratio; CT, computed tomography; SDH, subdural hemorrhage; tICH, traumatic intraparenchymal haematoma; TBI, traumatic brain injury.
  2. Subgroup analyses (e.g., Male, Female, Age) reflect the model’s performance when evaluated on the specified subgroup within the validation dataset. Both the Basic and Modified Models were trained on the full cohort (all patients) and included all variables listed in the Methods section. Subgroup stratification (e.g., sex) was applied only during validation to assess generalizability.
Back to article page