Skip to main content

Table 3 Comparison of the performance of each model, by confusion matrix and evaluation measures

From: Development and internal validation of a machine-learning-developed model for predicting 1-year mortality after fragility hip fracture

Models

Testing dataset

Predict deceased

Predict survived

Accuracy [95% CI]

P value

(vs RF model)

Positive predictive value [95% CI]

P value

(vs RF model)

Negative predictive value [95% CI]

P value

(vs RF model)

Specificity [95% CI]

P value

(vs RF model)

Sensitivity [95% CI]

P value

(vs RF model)

AUC [95% CI]

P value

(vs RF model)

Random Forests (RF)

  Deceased

13

6

0.95 [0.91–0.98]

 

0.93 [0.66–1.00]

 

0.96 [0.91–0.98]

 

0.99 [0.96–1.00]

 

0.68 [0.43–0.87]

 

0.99 [0.95–1.00]

 

  Survived

1

128

Gradient Boosting (GB)

  Deceased

11

8

0.93 [0.88–0.97]

0.11

0.85 [0.55–0.98]

0.49

0.94 [0.89–0.97]

0.31

0.98 [0.95–1.00]

0.56

0.58 [0.33–0.80]

0.32

0.98 [0.94–1.00]

0.47

  Survived

2

127

Artificial Neural Network (ANN)

  Deceased

13

6

0.94 [0.89–0.97]

0.5

0.81 [0.54–0.96]

0.32

0.95 [0.90–0.98]

0.95

0.98 [0.93–1.00]

0.32

0.68 [0.43–0.87]

1

0.92 [0.83–1.00]

0.01

  Survived

3

126

Logistic Regression (LR)

  Deceased

7

12

0.91 [0.85–0.95]

0.01

0.88 [0.47–1.00]

0.69

0.91 [0.86–0.95]

0.03

0.99 [0.96–1.00]

1

0.37 [0.16–0.62]

0.03

0.95 [0.88–1.00]

0.02

  Survived

1

128

Naive Bayes (NB)

  Deceased

7

12

0.89 [0.83–0.94]

0.02

0.64 [0.31–0.89]

0.01

0.91 [0.85–0.95]

0.03

0.97 [0.92–0.99]

0.18

0.37 [0.16–0.62]

0.03

0.91 [0.82–1.00]

0.02

  Survived

4

125

Support Vector Machine (SVM)

  Deceased

5

14

0.90 [0.84–0.94]

0.01

0.83 [0.36–1.00]

0.58

0.90 [0.84–0.95]

0.02

0.99 [0.96–1.00]

1

0.26 [0.09–0.51]

0.02

0.94 [0.86–1.00]

0.02

  Survived

1

128

K-Nearest Neighbors (KNN)

  Deceased

5

14

0.90 [0.84–0.94]

0.01

0.83 [0.36–1.00]

0.58

0.90 [0.84–0.95]

0.02

0.99 [0.96–1.00]

1

0.26 [0.09–0.51]

0.02

0.81 [0.68–0.93]

0.01

  Survived

1

128

  1. A P value < 0.05 indicates statistical significance
  2. Abbreviation: CI Confidence interval