Transparent AI for Orthopedic Surgery Prediction

Per-member clinical predictions with SHAP explanations, built for InterQual/MCG workflows and HL7 FHIR PAS.

93%
Model AUC
Clinical prediction accuracy for MCID achievement at 12 months
FHIR
PAS Ready
Built for Da Vinci PAS workflows and ePA compliance by 2027
SHAP
Explainable
Transparent AI with per-member clinical explanations
Current Challenges

The Perfect Storm: Four Converging Healthcare Forces

The U.S. healthcare system is experiencing unprecedented regulatory convergence that's transforming how knee replacement decisions are made, measured, and governed. Four critical forces are creating both urgent challenges and massive opportunities for transparent, evidence-based prior authorization.

The Accountability Challenge

25% APU Reduction Penalty For non-compliance with PROMs collection
2027 Public Reporting CMS will publicly report improvement rates
50% Data Collection Minimum patient coverage required

Medicare's mandatory PROMs collection for TKA/THA shifts focus from volume to outcomes, with severe financial penalties for hospitals failing to demonstrate patient improvement.

Source: CMS Inpatient THA/TKA PRO-PM Requirements

The Subjectivity Problem

High Physician Variation Significant disagreement on TKA indications
Elusive Surgical Readiness Guidelines remain unclear and subjective
Conflict Patient Decisions Uncertainty about surgical outcomes

Despite being one of the most common surgeries, TKA decisions remain "highly subjective and discretionary" with significant variation between orthopedic surgeons, rheumatologists, and primary care providers.

Source: BMC Health Services Research, Clinical Orthopaedics

The Transparency Requirement

6+ States with AI Laws Requiring physician oversight of AI decisions
Individual Data Requirement AI must use patient-specific, not group data
Explainable AI Mandate Transparent, reviewable decision-making

New state laws prohibit AI from making final prior authorization decisions without physician review, requiring transparent, explainable AI that bases decisions on individual patient data rather than population averages.

Source: Cooley LLP State AI Regulation Analysis 2025

The Evidence Gap Problem

40% TKAs Avoidable With structured conservative care
Weak Head-to-Head Evidence Conservative vs. surgical comparisons
Default TKA Becomes Default Choice due to evidence uncertainty

While structured conservative care (education, exercise, weight management) can avoid up to 40% of TKAs, the lack of robust head-to-head comparison evidence makes surgery the default choice. This evidence gap prevents optimal patient-treatment matching.

Source: Arthritis Care & Research, Budget Impact Analysis
CareFuse Solution

Clinical Decision Support

Predict whether a member will achieve MCID (ΔWOMAC ≥ 10) at 12 months with transparent, explainable AI built for UM workflows.

Patient Profile

Feature Value
Age 74
Sex F
Body Mass Index (BMI) Normal
Baseline Pain & Function 15.6
Baseline Pain 4.0
Baseline Stiffness 1.0
Baseline Disability 10.6
Lifestyle Modification None

About TKA Surgery

Total knee arthroplasty (TKA) is an invasive surgery with risks and a lengthy recovery period. This software helps identify potentially ineffective surgeries before they occur to protect patient well-being.

TKA Success

Prediction: Low Benefit Likelihood
Predicted Probability: 37%
Clinically significant benefit predicted
Summary: The model estimates a 37% chance of success, with a cutoff of 21%. Your prediction is driven up by Body Mass Index, but somewhat offset by Baseline Disability, Baseline Stiffness, Baseline Pain & Function.
Top Influencers
Baseline Disability 18% hurts
Baseline Stiffness 18% hurts
Baseline Pain & Function 15% hurts
Body Mass Index (BMI) 6% helps

Conservative Success

Predicted Probability: 60%
Clinically significant benefit predicted
Summary: The model estimates a 60% chance of success, with a cutoff of 50%. Your prediction is driven up by Baseline Pain, Baseline Pain & Function, but somewhat offset by Lifestyle Modification.
Top Influencers
Baseline Pain 23% helps
Baseline Pain & Function 18% helps
Baseline Disability 16% helps
Baseline Stiffness 12% helps

Counterfactual Analysis

Propensity Score Weighting: Our model generates counterfactual predictions for conservative care using propensity-score weighting, allowing direct comparison between surgical and non-surgical treatment outcomes for informed decision-making.

Main contributing factors List of features that most influence the predicted outcome.

Decision Support

Not auto-denial—configurable thresholds aligned with plan policy. Human-in-the-loop review required for all decisions.

Workflow Integration

Built for InterQual/MCG workflows with HL7 FHIR PAS compatibility. Seamless integration with existing UM systems.

Complete Documentation

Predicted probability, risk narrative, literature snippets, and exportable FHIR bundles for comprehensive documentation.

Clinical Evidence

Validation & Performance

At CareFuse, we've built a calibrated, explainable logistic-regression model (OAI: demographics + WOMAC) that estimates a TKA patient's probability of achieving MCID and generates a counterfactual for conservative care using propensity-score weighting. Across 50 multi-seed holdout splits it achieves AUC ≈ 0.93, with transparent calibration/thresholding and SHAP explanations suited for clinical review.

How we generate a prediction

This tool can generate a personalized prediction using information the patient can readily provide (basic demographics and medical history) together with responses to an internationally validated arthritis questionnaire . No imaging or specialized tests are required for the demo.

Validation Highlights

Up to 97%
Accuracy
AUC 0.93
TKA pathway
AUC 0.87
Conservative pathway
Sensitivity
to non-MCID procedures

Internal validation. Performance depends on thresholds and population. For demonstration only; not a coverage determination.

See methods

Methods

Our model uses logistic regression with demographics and WOMAC scores from the OAI dataset. We completed a live pilot with Hapvida (15.7M policyholders) to demonstrate feasibility and workflow integration. In the U.S., our pilot would train a new model on your de-identified data using the same pipeline and deploy a secure, metered API with audit logging.

CareFuse is represented by Gunderson Dettmer, and our product is built under the FDA non-device CDS exemption.

Platform & Integration

Authorization Plugin

Seamless integration with existing UM workflows through Da Vinci PAS and FHIR R4 compliance.

EHR/Portal
→
Da Vinci PAS
→
CareFuse
→
Payer System

CareFuse creates clinical attachments and rationale within the standard PAS workflow

Versioned Model Card

Complete model documentation with version control and performance tracking

Audit Logs

Comprehensive audit trail for all predictions and decisions

Data Provenance

Full traceability of data sources and processing steps

Drift & Bias Monitoring

Continuous monitoring for model performance and fairness

Role-Based Access

Granular permissions and access controls for different user roles

FHIR R4 Compliance

Generate FHIR R4 resources conformant with Da Vinci PAS (FHIR→X12 278)

Compliance & Governance

Regulatory Alignment

Built for healthcare compliance with comprehensive governance and audit capabilities.

HIPAA/BAA Compliance

  • Signed Business Associate Agreements
  • At-rest and in-transit encryption
  • Least-privilege access controls
  • Comprehensive audit logging
  • Breach response procedures

NAIC AI Bulletin Alignment

  • AIS Program support documentation
  • Detailed model cards and validation reports
  • Subgroup and fairness testing
  • Continuous monitoring plans
  • Vendor due-diligence packages

California AB-3030

For member-facing clinical communications to California residents:

  • Required AI disclaimers and human contact instructions
  • "Reviewed by licensed clinician" toggle to suppress disclaimers
  • Compliance with GenAI patient communication requirements

Colorado Readiness

Prepared for Colorado SB21-169 and SB24-205 requirements for plans operating in Colorado.

HIPAA Compliant
SOC 2 Ready
ISO 27001 Ready
Impact & ROI

Financial Impact Calculator

Estimate the financial impact of CareFuse on your utilization management operations.

85%
25%
38
Avoided Surgeries/Year
$950,000
Annual Medical Savings
$75,000
Admin Savings
$1,025,000
Net Annual ROI

Ready to Transform Your Prior Authorization Process?

See how CareFuse can improve your UM operations with transparent, evidence-based decision support.

30-Minute Demo

Personalized demonstration of CareFuse capabilities

ROI Analysis

Specific impact assessment for your operation

Pilot in 30 Days

Quick implementation with measurable outcomes

Pilot Deliverables

Prior Authorization Pipeline Implementation
Avoided surgeries quantification
Saved costs quantification

Your data is protected under HIPAA/BAA agreements

Strategic Vision

Expanding Clinical Decision Support

CareFuse: comprehensive platform for evidence-based utilization management across orthopedic procedures.

Hip & Knee Arthroplasty

Available Now

Validated model with 93% AUC, preventing unnecessary procedures with transparent AI

Shoulder Arthroplasty

Q2 2025

Expanding predictive technology to shoulder replacement procedures

Spinal Fusion

2026

Intelligent prediction for spinal arthrodesis procedures

Scientific References

Clinical & Regulatory Citations

1

CMS Interoperability & Prior Authorization Final Rule (CMS-0057-F): fact sheet & PDF.

CMS Fact Sheet (PDF)
2

HL7 Da Vinci PAS Implementation Guide.

Da Vinci PAS (FHIR IG)
3

NAIC Model AI Bulletin (AI governance expectations).

NAIC AI Model Bulletin (PDF)
4

California AB-3030 bill text / Medical Board summary.

Medical Board of California - GenAI Notification
5

AJRR annual report (context on U.S. hip/knee volumes).

AJRR Annual Report (AAOS)
6

Papakostidis C, Giannoudis PV, Watson JT, et al. Serious adverse events and 30‑day hospital readmission following elective total knee arthroplasty: a systematic review and meta‑analysis. J Orthop Surg Res. 2021;16:236.

PubMed (PMID 33789702) PMC8011390
7

PJI incidence within 12 months (recent systematic reviews and registry analyses).

PMC10618849
8

Clement ND, et al. WOMAC MCID ≈ 10 at 12 months post-TKA.

PubMed (PMID 30179956) PMC6259858
9

AAOS clinical news / costs context.

AAOSNow Clinical (June 2025)

MCID (Minimal Clinically Important Difference)

The smallest change in a treatment outcome that a patient would identify as important. For WOMAC scores, ΔWOMAC ≥ 10 points represents meaningful clinical improvement.

SHAP (SHapley Additive exPlanations)

A method to explain individual predictions by computing the contribution of each feature to the prediction. Provides transparent, interpretable AI explanations.

PAS (Prior Authorization Support)

HL7 FHIR implementation guide for electronic prior authorization workflows, enabling seamless integration between providers and payers.