Skip to main content

Table 4 Models with the best performance among all tested models

From: Developing a mortality risk prediction model using data of 3663 hospitalized COVID-19 patients: a retrospective cohort study in an Egyptian University Hospital

Models

Number (death/total)

AUC (95%CI)

Sensitivity (%)

Specificity (%)

PPV (%)

NPV (%)

Accuracy (%)

HL P-value

Overall

Balanced

Basic and biomarkers’ models

Basic Model

972/3663

0.832 (0.816–0.847)

55.0

90.6

67.8

84.8

81.1

72.8

0.010

INR

272/668

0.842 (0.812–0.873)

75.7

79.8

72.0

82.7

78.1

77.8

0.982

Creatinine

658/1980

0.814 (0.794–0.834)

56.2

87.8

69.7

80.1

77.3

72.0

0.583

TLC

745/2429

0.818 (0.800–0.837)

54.1

88.8

68.2

81.4

78.2

71.5

0.597

PLT

724/2340

0.820 (0.801–0.838)

53.2

88.6

67.7

80.9

77.6

70.9

0.549

HB

615/1813

0.803 (0.782–0.825)

53.7

87.9

69.5

78.7

76.3

70.8

0.093

LDH

326/1026

0.815 (0.787–0.843)

50.6

89.4

69.0

79.5

77.1

70.0

0.783

Troponin

112/285

0.747 (0.689–0.805)

54.5

84.4

69.3

74.1

72.6

69.5

0.671

CRP

538/1872

0.799 (0.777–0.821)

47.6

90.8

67.5

81.1

78.4

69.2

0.575

CK-Total

205/615

0.739 (0.697–0.781)

42.0

87.6

62.8

75.1

72.4

64.8

0.597

CK-MB

157/435

0.729 (0.680–0.777)

45.2

84.2

61.7

73.1

70.1

64.7

0.975

Extra models

Smoking

972/3663

0.832 (0.817–0.847

55.0

90.6

67.9

84.8

81.2

72.8

0.029

Obesity

972/3663

0.830 (0.814–0.845)

55.3

90.5

67.8

84.9

81.2

72.9

0.005

Haematological disease

972/3663

0.829 (0.814–0.845)

55.3

90.4

67.5

84.8

81.1

72.9

0.0004

Malignancy

972/3663

0.831 (0.816–0.846)

56.3

90.3

67.7

85.1

81.3

73.3

0.004

  1. Basic Model included: Age, Comorbidity presence, and Condition on admission. Each of the biomarker’s models and Smoking model included variables of basic model + the variable of the corresponding biomarker or smoking status. Obesity, Haematological, and Malignancy models included Age, Condition on admission, and the corresponding comorbidity variable
  2. Binary logistic regression was used to calculate the model parameters then, death predicted probability was calculated for each model
  3. AUC Area under the ROC curve, PPV Positive predictive value, NPV Negative predictive value, HL Hosmer–Lemeshow test, INR International normalized ratio, TLC Total leucocytic count, PLT Platelet count, HB Haemoglobin, LDH Lactate dehydrogenase, CRP C-reactive protein, CK Creatine kinase. CK-MB Creatine kinase-myoglobin binding