Table 2 Performance of pediatric emergency prediction using natural language processing techniques and topic models compared to KTAS.

From: A pediatric emergency prediction model using natural language process in the pediatric emergency department

Variable

KTAS

Logistic regression

XGBoost

Gradient boosting

Random forest

KM-BERT

KM-BERT with MLM*

AUROC

0.610 ± 0.002

0.709 ± 0.003

0.723 ± 0.003

0.726 ± 0.003

0.691 ± 0.003

0.800 ± 0.003

0.849 ± 0.003

AUPRC

0.679 ± 0.003

0.780 ± 0.003

0.801 ± 0.003

0.802 ± 0.004

0.761 ± 0.004

0.861 ± 0.003

0.896 ± 0.003

Recall

0.733 ± 0.003

0.655 ± 0.004

0.654 ± 0.003

0.662 ± 0.004

0.650 ± 0.003

0.748 ± 0.004†

0.748 ± 0.004†

Precision

0.685 ± 0.003

0.754 ± 0.004

0.758 ± 0.004

0.757 ± 0.005

0.738 ± 0.004

0.792 ± 0.003

0.842 ± 0.002

F1-score

0.708 ± 0.002

0.701 ± 0.004

0.702 ± 0.003

0.706 ± 0.004

0.691 ± 0.003

0.769 ± 0.003

0.792 ± 0.003

Accuracy

0.628 ± 0.002

0.657 ± 0.003

0.660 ± 0.003

0.662 ± 0.004

0.644 ± 0.003

0.724 ± 0.003

0.760 ± 0.003

Brier

0.271 ± 0.001

0.209 ± 0.001

0.203 ± 0.001

0.202 ± 0.001

0.216 ± 0.001

0.179 ± 0.002

0.156 ± 0.002

  1. KTAS, Korean Triage and Acuity Scale; KM-BERT, Korean medical bidirectional encoder representations from transformers; MLM, masked language modeling; AUROC, area under the receiver operating characteristics; AUPRC, area under the precision-recall curve.
  2. Bold values indicate the best-performing model across metrics.
  3. *Indicates overall best-performing model; †Indicates tied best performance for Recall.
Back to article page