In today’s healthcare fraud, waste, and abuse (FWA) landscape, Special Investigations Units face mounting pressure to do more with less—identifying sophisticated fraud schemes across exploding claims volumes while managing limited resources. At a recent industry roundtable, SIU leaders from across the country gathered to share how they’re leveraging AI-driven analytics to transform their fraud detection capabilities and deliver unprecedented results.
The Breaking Point: Why Traditional Approaches Are No Longer Enough
Healthcare spending has surpassed $4 trillion annually, and with it comes an administrative burden that’s fracturing the system. Traditional payment integrity approaches—heavily reliant on post-payment audits, manual chart reviews, and reactive investigations—are creating downstream challenges that impact everyone:
- Provider abrasion from denials, appeals, and resubmissions
- Increased administrative costs that negatively impact Medical Loss Ratio (MLR)
- Broken trust between payers and providers
- Delayed identification of fraud schemes that continue to drain resources
As one SIU director noted during the discussion, “We’re not the only ones that are going to be using AI to try and track down the fraudsters. You’re going to have fraudsters that I’m sure are already coming up with new schemes and new ways to use AI. So we certainly have to stay ahead of the curve.”
Real Results: AI-Driven Detection in Practice
Case Study 1: The Impossible Days Provider
One health plan identified a mental health counselor through AI-powered outlier detection showing impossible billing patterns—the provider was seeing patients across a geographic spread that would require being in multiple locations simultaneously, averaging 50-60 patients per day.
The AI Advantage: Traditional reviews might have flagged high volume, but AI analytics identified the geographic impossibility and corroborated it across multiple data points, including:
- Member locations at time of service
- Provider location documentation
- Travel time feasibility
- Pattern consistency across quarters
The Outcome: The investigation revealed systematic overbilling, leading to significant recoveries and corrective action.
Case Study 2: The COVID Testing Scheme
Multiple independent pharmacies began flooding the system with at-home COVID test distribution claims. While individually they might not have triggered alerts, AI pattern recognition identified a coordinated scheme across seemingly unrelated entities.
The AI Advantage: The platform detected:
- Unusual spike in specific procedure codes
- Correlation between multiple provider NPIs
- Shared billing addresses and patterns
- Timing anomalies in claim submissions
The Outcome: The scheme was identified and stopped before significant losses occurred, and policy changes were implemented to prevent future abuse.
Case Study 3: The Pulmonologist’s Pattern
A regional plan discovered a pulmonologist billing an average of 16 hours per day using repetitive, unique billing patterns—specifically billing the same specialized lung test (CPT 94729) multiple times per patient visit using modifier 76, often on new patient encounters.
The Investigation Approach:
- AI identified the outlier (receiving 95 times more reimbursement than peers—$207 per member vs. $40 for peers)
- Medical record review revealed vague documentation that didn’t support the frequency of services
- The provider was also billing allergy testing (CPT 95004)—outside their specialty
- Pattern showed suspicious same-day billing of E&M visits with the procedures
The Outcome: Through comprehensive medical review and member interviews, the plan established a 73.4% error rate and secured approximately $51,000 in recoupments for just 24 member records. The provider was placed under ongoing monitoring.
Case Study 4: The Personal Care Services Investigation
A California-based plan used AI-generated reports to identify personal care services providers billing for services while members were in inpatient settings—a clear impossibility.
The Multi-Layered Approach:
- AI Detection: Generated overlap reports identifying HCPCS code T1019 (personal care services) billed concurrently with inpatient stays
- Documentation Review: Requested and reviewed medical records showing vague, cloned documentation
- Member Interviews: Conducted soft-approach interviews to verify services
- Member Surveys: Deployed follow-up surveys for documentation
Key Investigation Best Practice: The SIU team emphasized their approach to member interviews—making members feel comfortable, building trust, and explaining how the verification protects their member ID from billing fraud. This approach led to significantly higher response rates and cooperation.
The Outcome: One provider investigation alone yielded $85,000 in recoupments, with four additional investigations ongoing showing similar patterns.
Case Study 5: The Modifier 25 & KX Manipulation
An SIU team investigating podiatrists identified a sophisticated scheme involving custom foot orthotics (CPT L3000) where providers were:
- Billing new patient E&M visits on the same day as orthotic dispensing
- Consistently using modifiers 25 (separate E&M) and KX (medical necessity met)
- Manipulating diagnosis codes to support billing
- Potentially targeting specific patient populations
The AI Advantage: Rather than manually reviewing thousands of claims, investigators used natural language queries to:
- Identify top billers by procedure code
- Filter same-day E&M and orthotic billing
- Compare provider patterns against peers
- Analyze modifier usage trends
The Investigation Status: This case demonstrates the power of AI for ongoing investigations—the team can continuously refine their analysis, expanding or narrowing focus based on initial findings, all while the investigation remains active.
The Game-Changer: Natural Language Query Capabilities
Perhaps the most significant advancement discussed was the introduction of natural language AI query tools that allow investigators to ask questions in plain English and receive comprehensive analysis in minutes—not days or weeks.
Real-World Query Examples:
Scenario 1: InfoShare Alert Response “Show me all claims for CPT codes 97110, 97112, 97116, 97124, and 97140 for chiropractic and physical therapy providers in Q1 2025.”
Result: Within 1-3 minutes, investigators received:
- Key observations identifying top two providers as outliers by a factor of 5-10x compared to peers
- Payment pattern analysis showing highest-paid procedure codes
- Provider concentration metrics
- Downloadable detailed data for Excel pivot analysis
The Value: Instead of spending hours or days requesting and waiting for data pulls, investigators got immediate actionable intelligence while still on the InfoShare call—allowing them to quickly determine exposure and prioritize resources.
Scenario 2: Proactive Monitoring “Which providers are billing the highest amounts for custom orthotics with same-day E&M visits?”
Result: The AI identified specific billing patterns, code combinations, and outliers that would have required weeks of manual analysis, complete with peer comparisons and statistical significance indicators.
Looking Ahead: The Future of AI in Fraud Detection
As one SIU director summarized: “We’re not the only ones that are going to be using AI to try and track down the fraudsters. You’re going to have fraudsters that I’m sure are already coming up with new schemes and new ways to use AI. So we certainly have to stay ahead of the curve.”
The healthcare fraud landscape is changing rapidly:
- AI-generated medical records are already appearing in investigations
- AI-created business entities may soon be submitting claims
- Sophisticated networks are coordinating across multiple providers and locations
- Manipulation of documentation is becoming more subtle and harder to detect manually
The question isn’t whether to adopt AI-powered fraud detection—it’s how quickly health plans can implement these tools to protect their programs and members.
Key Takeaways for SIU Leaders
- Start with high-value targets: Use AI to quickly identify your biggest exposures and prioritize resources accordingly.
- Corroborate, corroborate, corroborate: The most successful cases come from multiple data points all indicating the same issue.
- Don’t abandon traditional techniques: Member interviews, medical record reviews, and provider engagement remain critical—AI just makes them more targeted and effective.
- Build internal collaboration: Break down silos between SIU, prepayment review, post-payment audit, and provider relations.
- Get leadership buy-in: When providers push back hard (and they will), organizational support is essential.
- Document everything: AI-powered case management keeps audit trails clean and defensible.
- Keep learning: The technology is advancing rapidly—stay connected with peers to share best practices and new approaches.
The healthcare fraud landscape is complex and constantly shifting, but with AI-powered analytics, SIU teams are gaining the tools they need to stay ahead—protecting program integrity, reducing administrative burden, and fostering better relationships with providers. The future of fraud detection isn’t just about catching bad actors; it’s about building a more transparent, efficient, and trustworthy healthcare system for everyone.
Learn more about Codoxo’s AI-driven fraud, waste, and abuse platform, Fraud Scope.