DOC Clinical Decision Support

Digital Osteoarthritis Counseling System

Clinical intelligence suite

Knee osteoarthritis · Choose models below, then analyze: expected arthroplasty outcomes and conservative treatment estimates.

1
TKR risk model Random Forest · AUC 0.92
Choose model

Select which risk-model variant runs this analysis. Demo results below focus on expected surgical improvement and conservative options.

2
Post-TKR symptom improvement WOMAC pooled · OAI + Khatib

Validated on OAI, Khatib, and Bergman cohorts.

3
Conservative treatment estimates Always included
NSAID response Corticosteroid HA injection Surgery avoidance Nutrition (NRS)

Exploratory · Shared decision-making context · Run automatically with each analysis.

Evidence layer: PROBAST-gated literature mining · Continuous learning pipeline · Does not change core model weights without review.
PROBAST Low Risk N=6,954 patients No PHI stored

Patient data privacy

Designed for clinical use — we do not build a patient database from your entries.

Optional usage analytics below is off by default and never includes clinical values (no age, WOMAC, BMI, etc.).

Enter Patient Data

Enter patient information to analyze surgical risk and expected outcomes. Required: Age, Sex, BMI, at least one KL grade ≥1. Pain scores are optional but improve accuracy.

Quick demo:
Not required. Do not enter names or hospital MRN.
Enter pain scores to improve prediction accuracy, or leave blank to skip.
Higher score = more symptoms
Higher score = more symptoms
For counseling display only; does not change model weights.
Counseling factors (optional)

Literature-based flags only — does not change model predictions or conservative treatment numbers.

Time (in minutes) to walk 0.4 kilometers. Typical range: 2-10 minutes.
This field is optional but may improve prediction accuracy.
Note: Has the patient had total knee replacement on the opposite knee? Patients who already had TKR on one knee are more likely to eventually need TKR on the other knee.
This field is collected for future use and is not currently used in predictions.

Model choices (risk + outcome) are set in the Clinical intelligence suite above.

Clinical inputs are processed once and not retained on our servers.

Step 1: Basic Information

Quick demo:
Model configuration
TKR risk model — choose model

Step 2: Pain Assessment

Model works without pain scores but predictions may be less precise

Step 3: Assessment Details

Higher score = more symptoms
Higher score = more symptoms
Counseling factors (optional)

Literature flags only — does not change predictions.

Time (in minutes) to walk 0.4 kilometers. Typical range: 2-10 minutes.
Note: Has the patient had total knee replacement on the opposite knee?

Results