Skip to main content

Table 1 Characteristics of included studies

From: Accuracy of machine learning in identifying candidates for total knee arthroplasty (TKA) surgery: a systematic review and meta-analysis

No.

Study (years of publication)

Country

Study design

Follow time

Source of data

The quantity of TKA

Total quantity

Total quantity of training sets

Method of generating validation set

Total quantity of validation sets

Methods for variable screening/feature selection

Types of models

1

Sharmala Thuraisingam [30]

Australia

Cohort

5y

MedicineInsight

15979(P)

201462(P)

201462(P)

Bootstrap

NA

Literature review, Delphi process

Fine–Gray

2

Ahmad Almhdie-Imjabbar [31]

France

Cohort

9y

OAI

375(P) /291(P)

4382(P)/4296(P)

4382(P)/4296(P)

Cross-validation

NA

Literature review

LR

3

Kevin Leung [25]

USA

Cohort

9y

OAI

364(P)

728(P)

728(P)

Cross-validation

NA

Univariable and multivariable analysis

DL

4

David J. Houserman [19]

USA

Case–control

NA

A high-volume joint replacement practice

NA

NA

NA

Random sampling

812(P)

NA

DL

5

Afshin Jamshidi [32]

Canada

Cohort

9y

OAI

413(K)

7589(K)

6071(K)

Random sampling

1518(K)

Lasso’s Cox

Cox, DL, RF, SVM, LR

6

Aniket A. Tolpadi [26]

USA

Cohort

5y

OAI

NA

35482(I)

23126(I)

Random sampling

5241(I)

Literature review, multivariable analysis

DL

7

Aleksei Tiulpin [32]

Finland

Cohort

7y

MUST

30(P)

557(P)

557(P)

Cross-validation

NA

Literature review, univariable, and multivariable analysis

LR

8

Dahai Yu [28]

UK

Cohort

10y

CPRD and Multicenter

NA

NA

416030(P)

Cross-validation

NA

Literature review, Record-Wide Association Study, and panel consensus

Cox

9

Khadija Mahmoud [29]

UK

Cohort

5y

OAI and MOST

297(P)

6291(P)

3234(P)

Internal validation: random sampling

External validation: MOST

3057(P)

Literature review, expert knowledge

LR, LASSO, Ridge, DT, RF, GBM

10

Stephan Heisinger [20]

Austria

Cohort

4y

OAI

165(P)

165(P)

83(P)

Random sampling

82(P)

Variables of interest were publicly available

ANN

11

Haresh Rengaraj Rajamohan [27]

USA

Cohort

9y

OAI and MOST

650(P)

5307(P)

706(P)

Cross-validation

4601(P)

Univariable and multivariable analysis

ANN, DL, Ensemble model

12

Yang Li [34]

China

Cohort

11y

OAI

273(P)

4796(P)

188(P)

Random sampling

81(P)

NA

LASSO

13

Hongzhi Liu [35]

China

Cohort

5y

A high‐volume joint replacement practice

537(K)

1779(K)

1156(K)

Random sampling

623(K)

Univariate cox, multi-variate cox, and Lasso’s Cox

DL

  1. USA United States of America, UK United Kingdom, NA not available, OAI, Osteoarthritis Initiative: A Knee Health Study, MUST Musculoskeletal pain in Ullensaker Study, CPRD Clinical Practice Research Datalink, MOST Multicenter Osteoarthritis Study, P patient, K knee, I image, Fine–Gray competing risk regression model, LR logistic regression model, DL deep learning model, Cox Cox proportional-hazards model, SVM support vector machine model, RF random forest model, DT decision tree model, GBM gradient boosting machine model ANN, artificial neural network model