Page 20 - AIH-2-2

Artificial Intelligence in Health Machine learning in arthroplasty

Table 1. (Continued)
Author Study population/ Performance (AUROC, Input Function Model type
Training model size accuracy, sensitivity/
specificity, etc.)
Huber et al. 28 30,524 THA VAS AUROC: 0.87 Demographic, medical, Ability to predict Extreme gradient boosting,
THA Oxford AUROC: and outcome scores PROMs from patient multi-step adaptive
0.78 characteristics elastic-net, RF, neural
network, Naïve Bayes,
k-Nearest Neighbors
Klemt et al. 7 4,526 AUROC: >83 Demographic, surgical Ability to predict ANN, SVM, RF, and
variables, outcome scores PROMs from patient elastic-net penalized logistic
characteristics regression
Cost/LOS
Ramkumar 122,334 AUROC: 0.87 for LOS Demographic and Predicts LOS and Naive Bayesian model
et al. 30 AUROC: 0.71 for medical variables payment
payment
Ramkumar 78,335 AUROC: 0.82 for LOS Demographic and Predicts LOS, inpatient Naive Bayesian model
et al. 31 AUROC: 0.83 for medical variables charges, and discharge
Charges disposition
AUROC: 0.79 for
Disposition
Complications
Magnéli et al. 32 1,998 AE within 30 days Demographic, medical Predict adverse events RF, SVM, neural network
Sensitivity: 23%
Specificity: 90%
AEs within 90 days
Sensitivity: 31%
Specificity: 89%
Shah et al. 33 89,986 AUROC: 0.732 Demographic, clinical Predict major Bayesian optimization
complications algorithm
(e.g., infection, venous
thromboembolism,
cardiac complication,
pulmonary
complication)
Harris et al. 34 107,792 Renal complication Demographic, medical Predict death and LASSO
AUROC: 0.78 heart complications
Cardiac complication
AUROC: 0.73
Death AUROC: 0.73
Karhade et al. 36 5,507 AUROC: 0.77 Demographic, medical, Predict post-operative Elastic net-penalized logistic
opioid usage opioid usage in regression, stochastic
patients undergoing gradient boosting, RF, SVM,
THA neural network
Loppini et al. 37 630 AUROC: 0.99 Radiographic imaging Identify prosthetic CNN
Accuracy: 0.97 loosening
Sensitivity: 0.97
Specificity: 0.97
Shah et al. 38 697 Accuracy: 88.3% Radiographic imaging Identify prosthetic CNN
Sensitivity: 70.2% loosening
Specificity: 95.6%
Abbreviations: AE: Adverse events; ANN: Artificial neural network; AUROC: Area under the receiving operating characteristic curve; BMD: Bone
mineral density; CNN: Convolutional neural network; DCNN: Deep convolutional neural network; DXA: Dual-energy X-ray absorptiometry;
HOOS JR: Hip and Knee Disability and Osteoarthritis Outcome Scores for joint replacement; LASSO: Least absolute shrinkage and selection operator;
LOS: Length of stay; MCS: Mental component scores; PCS: Physical component scores; PROMS: Patient-reported outcome measures; RF: Random
forest; SMV: Support vector machine; THA: Total hip arthroplasties; VAS: Visual Analog Scale; VGG: Visual geometry group.

Volume 2 Issue 2 (2025) 14 doi: 10.36922/aih.3278