Skip to main content

Table 5 Model performance using the Braden scores

From: Modeling and prediction of pressure injury in hospitalized patients using artificial intelligence

AI technique Original (row, imbalanced) data Post-processed (rebalanced) cohorts
Model Model
LASSO linear regularized model Performance of testing data Model performance of testing data
  Confusion matrix Result false Result true Confusion matrix Result false Result true
  Prediction false 4759 172 Prediction False 4202 38
  Prediction true 14 68 Prediction True 571 202
  Accuracy 96.29%   Accuracy 87.85%  
  95% CI 95.73% 96.80% 95% CI 86.92% 89.74%
  Kappa 0.4079   Kappa 0.3514  
  Sensitivity 0.28333   Sensitivity 0.84167  
  Specificity 0.99707   Specificity 0.88037  
  Prevalence 0.01636   Prevalence 0.15420  
Random forest
Confusion matrix Result false Result true Confusion matrix Result false Result true
Neural networks Prediction false 6120 236 Prediction false 6058 225
  Prediction true 99 110 Prediction true 161 121
  Accuracy 94.897%   Accuracy 94.120%  
  Precision 0.5263   Precision 0.4291  
  AUC 0.7252   AUC 0.7134  
  Recall 0.3179   Recall 0.3497