Table 3 Model Complete Set
Author (year) | Feature Selection | AUC | Calibration Intercept | Calibration Slope | Brier Score | Concordance (95 CI) | Calibration Error |
|---|---|---|---|---|---|---|---|
Martin (2023) | Age at surgery Yrs. injury to surgery KOOS QOL Graft: hamstring Age at injury Femur fix: susp/cort. Graft: QT/BQT KOOS Sport Men. injury: none Activity: pivoting Graft: other Fix. comb: susp/interference Surgery on same knee KOOS All low | NR | NR | NR | NR | 1 year: Cox Lasso 0.59 (0.56–0.61) RSF: 0.67 (0.64–0.69) GB: 0.67 (0.65–0.70) SL: 0.67 (0.65–0.69) 2 year: Cox Lasso 0.58 (0.56–0.61) RSF: 0.67 (0.64–0.69) GB: 0.67 (0.64–0.69) SL: 0.67 (0.64–0.69) 5 year: Cox Lasso 0.58 (0.56–0.61) RSF: 0.67 (0.65–0.69) GB: 0.67 (0.64–0.69) SL: 0.67 (0.64–0.69) | 1 year: Cox Lasso 7.19, n.s RSF: 5.54, n.s GB: 7.48, n.s SL: 8.67, p = 0.034 2 year: Cox Lasso 8.17, p = 0.043 RSF: 6.42, n.s GB: 4.53, n.s SL: 4.10, n.s 5 year: Cox Lasso: 11.37, p = 0.01 RSF: 9.27, p = 0.026 GB: 11.07, p = 0.011 SL: 11.82, p = 0.008 |
Martin (2022a) | Patient age at primary surgery KOOS QoL score at primary surgery Graft choice Femur fixation method Years between injury and ACLR | NR | NR | NR | NR | 1 year: Cox Lasso: 0.678 2 years: Cox Lasso: 0.676 5 years: Cox Lasso : 0.678 | 1 year: Cox Lasso: 22.24, p < 0.001 2 years: Cox Lasso: 11.82, p = 0.008 5 years: Cox Lasso : 13.98, p = 0.003 |
Martin (2022b) | Age at surgery Fixation combination Tibia fixation Femur fixation BMI KOOS Sport at surgery KOOS QOL at surgery Years from injury to surgery Age at injury Hospital type Further injury Meniscus injury Injured side | NR | NR | NR | NR | 1 year Cox Lasso: 0.686 Random Forest: 0.672 GAM 0.687 GBM 0.669 2 year Cox Lasso 0.684 Random Forest: 0.670 GAM 0.685 GBM: 0.666 5 year: Cox Lasso: 0.683 Random Forest: 0.670 GAM: 0.684 GBM: 0.665 | 1 year Cox Lasso: 4.89, n.s Random Forest: 3.12, n.s GAM 4.79, n.s GBM 4.98, n.s 2 year Cox Lasso 11.35, p = 0.01 Random forest: 11.66, p = 0.009 GAM 11.19, p = 0.011 GBM: 3.76, n.s 5 year: Cox Lasso: 6.19, n.s Random Forest: 3.71, n.s GAM: 6.98, n.s GBM: 0.38, n.s |
Johnson (2023) | Age Sex BMI Occupation Sport participation Injury mechanism Occurrence of reoperation after ACLR | MLPClassifier: AUC = 0.61 GaussianNB: AUC = 0.58 LogisticRegression: AUC = 0.70 KNeighborsClassifier: AUC = 0.68 BaggingClassifier: AUC = 0.75 RandomForestClassifier: AUC = 0.76 AdaBoostClassifier: AUC = 0.73 GradientBoostingClassifier: AUC = 0.75 XGBClassifier: AUC = 0.77 | NR | NR | NR | NR | NR |
Lopez (2023) | Sex Race BMI (Calculated From The Recorded Height And Weight) American Society Of Anesthesiologists (ASA) Classification History Of Smoking Diabetes Hypertension Requiring Medication Wound Infection Use Of Steroids For A Chronic Condition Bleeding disorders were Abstracted Perioperative Data Such As Anesthesia Type (General, Spinal, IV Sedation, Regional, Other) Surgery Setting (Inpatient Vs Outpatient) Operative Time (Prolonged Operative Time Defined As > 120 min) | ANN: Reoperation: 0.842 ACLR-related Readmission: 0.872 Logistic Regression: Reoperation: 0.601 ACLR-related Readmission: 0.606 | NR | NR | NR | NR | NR |
Ye (2022) | Age Sex BMI Time From Injury To Surgery Participation In Competitive Sports Preoperative Lysholm And IKDC Scores Posterior Tibial Slope High-Grade Knee Laxity Graft Diameters Of Anteromedial And Posterolateral Bundles Medial And Lateral Meniscal Resection Follow-Up Period Meniscal Reinjury After ACLR | Graft Failure: XGBoost (excellent): AUC = 0.944 (0.001), Accuracy = 0.986 (0.012) Residual Laxity: Random Forest (excellent): AUC = 0.920 (0.014), Accuracy = 0.914 (0.024) | NR | NR | NR | NR | NR |
Martin (2024) | Patient Age At Primary Surgery Knee Injury And Osteoarthritis Outcome Score Quality Of Life Subscale (KOOS-QOL) Score At Primary Surgery Graft Choice Femur Fixation Method Time Between Injury And ACLR | NR | NR | NR | NR | 1 year: Original Norwegian Algorithm Performance: 0.686 (0.652–0.721) STABILITY data: HT = HT, HT + LET = BPTB: 0.713 (0.634–0.791) HT = HT, HT + LET = Unknown: 0.609 (0.528–0.691) All patients = HT: 0.674 (0.597–07.51) 2 year: Original Norwegian Algorithm Performance: 0.684 (0.650–0.718) STABILITY data: HT = HT, HT + LET = BPTB: 0.713 (0.637–0.789) HT = HT, HT + LET = Unknown = 0.608 (0.530–0.688) All patients = HT: 0.673 (0.598–0.747) | 1 year: Original Norwegian Algorithm Performance: 4.9 n.s. STABILITY data: HT = HT, HT + LET = BPTB: 2.6 n.s. HT = HT, HT + LET = Unknown: 10.6 p < 0.01 All patients = LT: 8.7 p < 0.01 2 year: Original Norwegian Algorithm Performance: 11.3 p = 0.01 STABILITY data: HT = HT, HT + LET = BPTB: 11.7 p < 0.01 HT = HT, HT + LET = Unknown: 8.9 p < 0.01 All patients = LT: 10.2 p < 0.01 |
Jurgensmeier (2023) | Age Sex Body mass index Sport participation Diagnosis of hypermobility or malalignment Radiographic findings Management | SVM: Apparent 0.782 (0.779–0.785), Internal Validation 0.738 (0.736–0.739) Random Forest: Apparent 0.997 (0.994–0.999), Internal Validation 0.790 (0.785–0.795) XGBoost: Apparent 0.814 (0.813–0.816), Internal Validation 0.758 (0.755–0.761) Elastic Net: Apparent 0.673 (0.61–0.736), Internal Validation 0.646 (0.643–0.648) | SVM: 0.0161 (− 0.0173 − 0.0149) Random Forest: 0.006 (0.005–0.0071) XGBoost: 0.007 (0.0055–0.0077) Elastic Net: 0.0165 (0.0152–0.0178) | SVM: 1.091 (1.086–1.096) Random Forest: 0.9608 (0.9562–0.9654) XGBoost: 0.9569 (0.9522–0.9616) Elastic Net: 0.8926 (0.8861–0.8992) | SVM: 0.14 (0.13–0.15) Random Forest: 0.10 (0.09–0.12) XGBoost: 0.12 (0.11–0.14) Elastic Net: 0.18 (0.17–0.20) | NR | NR |
Lu (2022) | Age Sex Body mass index Activity level Occupation Comorbid diagnosis Radiographic findings Management | ACLR: 0.80 (0.76–0.83) Non-op: 0.66 (0.58–0.74) | ACLR: 0.0051 (− 0.014–0.024) Non-op: 0.0048 (− 0.059–0.069) | ACLR: 0.97 (0.89–1.05) Non-op: 0.97 (0.65–1.30) | ACLR: 0.106 (0.029–0.183 Non-op: 0.111 (0.034–0.188) | NR | NR |