Skip to main content

Table 1 Characteristics of the selected articles

From: Machine learning models for predicting unscheduled return visits to an emergency department: a scoping review

Study, Year

Country

Research scale

Subject group

Number

Record period

ML model

Predictive power

Validation

72-hour URVs

Lee et al. 2012

US

Single center

Pediatric patients

66,861

2009

DAMIP model

AUC 0.85

AUC 0.831

Hu et al. 2017

Taiwan

Nation-based

Pediatric patients

125,940

1998 to 2009

NaĂŻve Bayes model

AUC 0.644

x

Simple Cart model

AUC 0.721

x

RF model

AUC 0.723

x

LR model

AUC 0.718

x

Pellerin et al. 2018

US

Multicenter (4)

All patients

21,141

2014

LR model

AUC 0.74

AUC 0.7279

Fernades et al. 2019

Portugal

Single center

Adult patients

511,301

2012 to 2016

LR model

AUC 0.842

x

SVM model

AUC 0.791

x

Meng et al. 2019

Singapore

Single center

All patients

328,733

2011 to 2013

LR model

Accuracy 0.67

AUC 0.66

DAMIP model

Accuracy 0.728

AUC 0.728

Hong et al. 2019

US

Multicenter (2)

Adult patients

330,631

March 2013 to July 2017

LR model

72 h-URV: AUC 0.69

AUC 0.692

9d-URV: AUC 0.71

AUC 0.708

XGB model

72 h-URV: AUC 0.73

AUC 0.717

9d-URV: AUC 0.73

AUC 0.727

Chen et al. 2021

Taiwan

Nation-based

Patients older than 65

49,252

1996 to 2010

Decision tree-based model

AUC 0.768

x

Chimel et al., 2021

UK

Single center

Adult patients

44,294

April 2019 to April 2020

XGB model

AUC 0.761

AUC 0.747

Hsu et al. 2022

Taiwan

Multicenter (2)

Patients with abdominal pain

290,914

2018 to 2019

LR model

AUC 0.73

x

RF model

AUC 0.71

x

XGB model

AUC 0.74

x

Reduced VC model

AUC 0.72

x

Xie et al. 2022

Singapore

Nation-based

All patients

216,877

2011 to 2019

LR model

AUC 0.683

x

RF model

AUC 0.666

x

XGB model

AUC 0.700

x

AutoScore model

AUC 0.673

x

MLP model

AUC 0.696

x

Med2Vec model

AUC 0.673

x

LSTM model

AUC 0.694

x

30-day URVs

Hao et al. 2014

US

Nation-based

All patients

293,461

2012

Decision tree-based model

AUC 0.71

x

Suffoletto et al. 2016

US

Multicenter (2)

Patients older than 65

202

2015

LR model

AUC 0.69

x

Poole et al. 2016

US

County-based

All patients

1,125,118

2010 to 2014

LR model

AUC: 0.77

x

Lasso model

AUC 0.84

x

RF model

AUC 0.98

x

Fowler et al. 2017

US

Single center

All patients

91,297

2010 to 2015

XGB model

AUC 0.75

x

  1. URVs: unscheduled return visits; ML: machine learning; AUC: area under the curve; DAMIP: Discriminant Analysis Via Mixed Integer Programming; RF: Random Forest; LR: Logistic Regression; SVM: Support Vector Machine; XGB: Extreme Gradient Boosting; VC: voting classifier; MLP: Multilayer perceptron; LSTM: Long short-term memory; US: United States; UK: United Kingdom