Groundbreaking Study Reveals Impact of AI in Tackling Healthcare Claims Errors, Saving $11.8Million

Health at Scale’s AI-driven system deployed by Personify Health proven to save money and address upstream source of inequities in care delivery

Fraud, waste, and abuse (FWA) has long been a challenging issue within the U.S. healthcare system, contributing to rising costs and compromised patient care. Today, researchers from Personify Health, Health at Scale, Massachusetts Institute of Technology, and University of Michigan announce a new peer-reviewed study on the reduction of FWA in healthcare insurance claims. Published in NEJM Catalyst Innovations in Care Delivery, the study highlights the successful deployment of AI-driven technology by Personify Health, the industry’s first personalized health platform company, resulting in an $11.8 million reduction in paid claim amounts over an eight-month period.

Also Read: Unilever and Accenture Join Forces to Establish a New Industry Standard in Generative AI-Powered Productivity

Key study findings include:

• More than half of flagged claims reduced – 54% of the claims flagged for subsequent clinical review saw a reduction in paid amounts, leading to an average savings of $3,916 per flagged claim – totaling more than $11.8 million, a 1.2% reduction of the total spend of $981.3 million.
  
  
• Meaningful reductions seen at all dollar ranges – Noteworthy reductions were observed across different claim value ranges, including for healthcare services that are low-cost but high-volume:
  • Claims over $50,000: 58% of these high-cost claims saw a decrease in paid amounts, with an average savings of $20,513 per claim.
  • Claims $5,000 – $49,999: Between 56% – 68% of these claims experienced a reduction in paid amounts, resulting in an average savings range from $1,504 for lower dollar claims to $10,420 for higher dollar claims.
  • Claims under $5,000: 42% of claims falling below $5,000 had a reduction in paid amounts, saving an average of $236 per claim.
• Specialties top the list for highest reductions – 10 service types are responsible for 56% of overall spend but accounted for 64% of all flagged claims. The greatest average reductions in paid amounts per flagged claim were observed for cardiovascular procedures ($10,332 average reduction), critical care services ($10,332), and musculoskeletal procedures ($9,604).

Also Read: The Role of AI and Machine Learning in Streaming Technology

“The study spotlights the significant savings being left on the table by traditional methods for FWA identification based on rules and high-dollar thresholds,” said Justin Tran, chief strategy officer at Personify Health. “Using AI to fight FWA has been a key investment for many for the past decade, but most models have only provided incremental improvements and have failed to perform sufficiently to transform processes. With real-time, AI-enabled FWA screening, we are unlocking substantial savings and operational improvements for healthcare stakeholders. We’re now seeing the transformative potential of real-time AI solutions in combatting FWA and promoting cost-effective, patient-centered care delivery.”

With estimated annual costs of medical waste in the U.S. amounting to more than a trillion dollars, the study underscores the critical need for innovative approaches to prevent overspending on unnecessary services and overpaid or mismanaged claims in the healthcare industry. Additionally, FWA analysis has typically focused on manual high-dollar claim review ($100,000 or more), leaving FWA in lower cost claims largely unchecked. Until now, no one has cracked the code for solving this problem before money gets spent.

Also Read: AiThority Interview with Paul Fipps, President, Global Industries and Strategic Growth at ServiceNow

Impact on Socially Vulnerable Populations

FWA disproportionately impacts vulnerable populations, perpetuating health disparities among these groups. The study highlights the role of AI in protecting underserved individuals from substandard and harmful healthcare practices. Researchers found that the AI-based FWA claim screening resulted in the most socially vulnerable members saving an average of $3,973 per claim, an 18% greater avoidance of inappropriate healthcare reimbursements.

“Addressing FWA can be a painful, time-consuming process when it occurs after payment has been made. But more than just the dollars spent, harm can occur through provider mistakes, bad behavior, and unnecessary care resulting in negative side effects or health complications,” said Zeeshan Syed, lead study author and co-founder & CEO of Health at Scale. “Too often, inaccurate claims and inappropriate care go unnoticed – especially among vulnerable populations who have limited care options. By leveraging a more effective approach to FWA screening, we can protect vulnerable populations disproportionately impacted and help create a more equitable healthcare system.”

To access the full study, “Reducing Fraud, Waste, and Abuse Through Real-Time AI-Based Screening: Prospective Results in Deployment,” visit the October 2024 issue of NEJM Catalyst Innovations in Care Delivery.

Study Methodology
A prospective FWA study was conducted to evaluate the potential for using AI-driven technology to reduce FWA, improve affordability, enhance safety, and ensure equitable access to healthcare for socially vulnerable individuals. Researchers analyzed more than 2.6 million professional and institutional claims from over 276,000 members of employer-sponsored health plans received by the largest third-party administrator in the U.S. (now Personify Health, formerly HealthComp) between July 1, 2022 through March 31, 2023. Health at Scale’s commercially available real-time AI-based technology was then applied to look for FWA in claims above a dollar threshold of $4,000 in the pre-payment period.

By incorporating the context of members’ medical histories including acute and chronic health diagnoses, procedures, prescriptions and orders, and other related care encounters including providers seen, facilities, and care settings, the AI-based system was able to flag irregularities in claims with a high degree of accuracy, and with special attention to high-dollar anomalies. Flagged claims were then characterized by dollar amounts, service types, and member sub-groups determined by social determinants of health parameters.

Also Read: Want to Beat FOIA Backlogs? Embrace AI