Identification of avoidable patients at triage in a Paediatric Emergency Department: a decision support system using predictive analytics

Table 1 List of classifications models used to predict if a visit to the Paediatrics emergency department is avoidable and should be directed the fast-track area

Model	Type	Interpretability	Implementation	R Package
Split made by triage levels¹	Rules	Glass box	Paper based	Base R
Simple classification tree²	Tree based	Glass box	Paper based	rpart
Logistic regression	Regression	Glass box	Integrated in HIS	glm
Naive bayes	Bayesian	Glass box	Integrated in HIS	klaR
Complex classification tree	Tree based	Glass box	Integrated in HIS	rpart
Random forest	Tree based ensemble	Black box	Integrated in HIS	ranger
XGboost [21]	Tree based ensemble	Black box	Integrated in HIS	xgboost
Tabnet [22]	Deep learning	Black box	Integrated in HIS	torch
Tensor Flow through Keras	Deep learning	Black box	Integrated in HIS	keras

¹Visits triaged Red/Orange/yellow are directed to regular emergency department patient Flow and Green/Blue visits are directed to fast track. HIS – hospital’s information system
²Tree depth, complexity, and the use of variables with many categories was taken into consideration to keep the tree simple and therefore applicable in the fast-paced context of triage

ISSN: 1471-227X