From: Prediction models for amputation after diabetic foot: systematic review and critical appraisal
First author year | Candidate variables | Missing data | Variable selection methods None | Type of validation | Model evaluation | Calibration method | |||
---|---|---|---|---|---|---|---|---|---|
No | Continuous variables processing method | EPV | No | Processing methods | |||||
Chen 2023 | 14 | Remain unaltered | 1.786 | – | Missing were excluded, analysis with complete data | CR | Internal | Random split validation | None |
Li 2023 | 31 | Remain unaltered | 0.484 | – | Missing were excluded, analysis with complete data | VIMP | Internal | Bootstrap | H–L test |
Yang 2023 | 32 | Remain unaltered | 6.656 | < 10% of quantitative data | Missing values > 40% were deleted, Median was used for missing quantitative data | FSR, Pruning algorithm | Internal | Tenfold cross validation | H–L test |
Stefanopoulos 2022 | 36 | Some converted to categorical variables | 53.556 | – | Missing were excluded, analysis with complete data | Lasso regression | Internal | Random split validation | None |
Wang 2022 | 21 | All converted to categorical variables | 3.571 | – | Missing were excluded, analysis with complete data | LR | Internal | Tenfold cross validation | None |
Xie 2022 | 37 | Remain unaltered | 1.919 0.270 | – | Model automatically handle | None | Internal | Fivefold cross validation | H–L test, calibration curve |
Du 2021 | 31 | Some converted to categorical variables | 0.194 | – | – | None | Internal | Threefold cross validation | None |
Li 2021 | 44 | Z-score standardization | 2.682 | 70 | Delete of features with > 60% missing values. The rest were imputed using median, mean, mode, fixed value, or KNN | RF-RFE, mRMR, JMI, JMIM, original | Internal | Fivefold cross validation | None |
Peng 2021 | 21 | Remain unaltered | 2.762 | – | Sever missing values were excluded | FSR | Internal | Bootstrap | H–L test, Calibration curve |
Hüsers 2020 | 7 | Remain unaltered | 10.714 4.143 | LTFU:16 | Analysis with complete data | – | None | None | None |
Lin 2020 | 33 | Remain unaltered | – | – | Analysis with complete data | CR | Internal | Random split validation | None |
Vera-Cruz 2020 | NA | Some converted to categorical variables | NA | None | – | NA | External | Spatial validation | None |
Chetpet 2018 | 13 | All converted to categorical variables | 3.384 | LTFU:21 | Analysis with complete data | – | None | None | None |
Chen 2018 | 33 | Some converted to categorical variables | 1.121 | – | – | CR | Internal | Random split validation | None |
Jeon 2017 | NA | Remain unaltered | NA | LTFU:21 | Analysis with complete data | NA | External | Spatial validation | None |
Kasbekar 2017 | 17 | Remain unaltered | 4.882 | – | Missing were excluded, analysis with complete data | – | Internal | Random split validation | None |
Monteiro-Soares 2015 | NA | Remain unaltered | NA | LTFU: 9 Miss: 223 | Missing were excluded, analysis with complete data | NA | External | Spatial validation | None |
Pickwell 2015 | 20 | Some converted to categorical variables | 7.950 5.150 | – | – | FSR | None | None | None |
Lipsky 2011 | 33 | Some converted to categorical variables | 19.606 | – | – | SR | Internal | Random split validation | H–L test |
Barberan 2010 | 19 | Remain unaltered | 1.368 | – | – | UA | None | None | None |