DOC Validator · Literature Admin · Risk Model Dashboard

All Models Dashboard

Comprehensive overview of all DOC prediction models - performance metrics, validation status, and clinical applications.

TKR Surgery Risk Model Production Externally Validated

0.92
AUC
Discrimination
0.037
Brier Score
Calibrated
6,954
Training N
OAI + Khatib
57%
Brier Improvement
vs Original
586
Literature Articles
For Calibration
Purpose: Predicts probability of needing total knee replacement (TKR) within 4 years.
Algorithm: Random Forest Classifier with literature-informed Platt scaling calibration.
Validation: Internal (OAI test set) + External (Khatib cohort, Bergman registry).
Age Sex BMI WOMAC Score KL Grade Family History Walking Time

WOMAC Outcome Model Production Externally Validated

0.69
R²
Variance Explained
±10
MAE
WOMAC Points
2,528
Training N
OAI + Khatib
0-96
Scale
WOMAC Points
≥30
Success Threshold
Points Improvement
Purpose: Predicts expected WOMAC improvement after TKR surgery.
Algorithm: Random Forest Regressor with optional literature calibration.
Validation: Internal + External (Khatib surgical cohort).
Age Sex BMI Baseline WOMAC KL Grade Cohort

NRS/VAS Outcome Model Production Externally Validated

0.65
R²
Variance Explained
±2
MAE
NRS Points
4,426
Training N
Bergman Registry
0-10
Scale
NRS/VAS Points
≥3
Success Threshold
Points Improvement
Purpose: Predicts expected NRS pain improvement after TKR surgery. Native NRS model - no conversion needed.
Algorithm: Random Forest Regressor trained on Dutch Bergman registry.
Best For: European clinics using NRS/VAS pain scales (0-10).
Age Sex BMI NRS Activity NRS Rest EQ-5D VAS OKS ASA Score

Conservative Treatment Models Exploratory Lower Confidence

Important: These models have modest predictive accuracy (R² ~0.15) compared to surgical models (R² 0.65-0.69). Based on observational, non-randomized data. Use for shared decision-making context only, not precise predictions. PROBAST: Moderate Risk due to potential selection bias and lack of external validation.

Surgery Avoidance Model (Not currently displayed - reflects OAI base rate)

0.887
AUC
Discrimination
0.031
Brier Score
Calibration
4,625
Training N
OAI Non-Surgical

Treatment Response Models

0.15
R² (NSAID)
WOMAC Change
0.15
R² (Steroid)
WOMAC Change
0.15
R² (HA)
WOMAC Change
7.1
MAE
WOMAC Points
Purpose: Predicts probability of avoiding surgery with conservative treatment, and expected symptom improvement from NSAIDs, steroid injections, and HA injections.
Algorithm: Logistic Regression (surgery avoidance) + Random Forest Regressors (treatment response).
Limitations: Observational data only - treatment assignment not randomized. Modest R² typical for pain-change prediction. Best used for shared decision-making support.
Age Sex BMI Baseline WOMAC KL Grade Treatment Visits

PROBAST Compliance Summary

Model Participants Predictors Outcome Analysis Overall
TKR Risk Model Low Low Low Low Low Risk
WOMAC Outcome Low Low Low Low Low Risk
NRS Outcome Low Low Low Low Low Risk
Surgery Avoidance Low Low Low Moderate Moderate
Treatment Response Low Moderate Low Moderate Moderate
Last updated: -- · Detailed Risk Model Metrics