Skip to main content

Table 3 Overview of methods for model validation across studies (N = 43)

From: Application of machine learning in predicting hospital readmissions: a scoping review of the literature

Type of validation

Number of studies (Percent)

Featuring studies

Internal validation

33 (77%)

 

Training/testing split

21 (49%)

[51, 53, 55, 56, 58, 60, 65,66,67, 70, 73, 74, 76, 78, 79, 82,83,84,85,86,87]

Resampling

12 (28%)

 

 k-fold cross-validation

8

[49, 54, 57, 61, 68, 69, 77, 81]

 Stratified k-fold cross-validation

2

[61, 80]

 Repeated k-fold cross-validation

1

[48]

 Repeated random test-train splits

1

[50]

External validation

4 (9%)

 

 Split by time

3

[48, 59, 63]

 Separate dataset

1

[57]

No Validation

6 (14%)

[46, 47, 52, 62, 72, 88]