Skip to main content

Table 1 Summary of literature predicting total daily patient flows in EDs, highlighting forecasting horizons, best algorithms and the accuracies achieved

From: Forecasting patient flows with pandemic induced concept drift using explainable machine learning

Study

Year

Forecast periods (days)

Features

Best algorithm

MAPE

Boyle et al. [5]

2012

1

4

ARIMA, ES, OLS

7.0%

Xu et al. [51]

2013

1

7

ANN

6.8%–7.3%

Marcilio et al. [29]

2013

7

15

GLM

7.6%

 

30

 

GLM

9.7%

Xu et al. [52]

2016

1

31

ARIMA-LR

6.5%

 

7

 

ARIMA-LR

9.6%

Calegari et al. [7]

2016

1

5

SES

2.9%

 

7

 

SES

10.7%

 

14

 

SES

10.7%

 

21

 

SES

11.4%

 

30

 

SES

11.7%

Navares et al. [32]

2018

1

13

ARIMA

8.1%–12.3%

Whitt and Zhang [49]

2019

1

57

SARIMAX

8.4%

Rocha and Rodrigues [39]

2021

1

15

LSTM

4.2%

Sudarshan et al. [42]

2021

3

10

CNN

9.2%

 

7

 

LSTM

8.9%

Vollmer et al. [47]

2021

1

30

Stacking

6.8%–8.6%

Harrou et al. [21]

2022

1

7

DBN

4.1%

Zhang et al. [53]

2022

1

29

SVR

8.8%

Petsis et al. [37]

2022

1

38

XGBoost

6.5%

 

2

 

XGBoost

6.9%