Authors: Ioakeim Perros, PhD, Justin Tran, Mohammed Saeed, MD, PhD, A. Mark Fendrick, MD, John Guttag, PhD, and Zeeshan Syed, PhD
Fraud, waste, and abuse (FWA) remains a persistent and formidable challenge within the U.S. health care system, reducing affordability and safety while perpetuating health disparities by disproportionately impacting socially vulnerable individuals. This study reports on the prospective results of the deployment of real-time AI-based FWA screening for a large and well-characterized population served by Personify Health, an independent third-party administrator based in Plano, Texas, and serving hundreds of employers in the United States. Between July 1, 2022, and March 31, 2023, incoming claims for 276,833 members were screened for FWA in real time during prepayment using AI-based FWA screening by Health at Scale, a health care machine intelligence company based in San Jose, California. For claims flagged, medical records were requested from billing providers followed by clinical review of the records received and final physician-led adjudication for appropriateness, thus ensuring alignment of medical services and interventions with established guidelines and standards of medical care. Performance was evaluated across the full population and within subgroups including members stratified by social vulnerability. The study found that 3,013 (0.1%) of 2,657,597 claims were flagged in real time by AI-based FWA screening for clinical review. Of those flagged claims, 1,623 (53.9%) were adjudicated for a reduction in the amount paid. The reduction in paid amounts for these claims totaled US$11.8 million (a US$3,914 average reduction per flagged claim and a US$7,267 average reduction per adjudicated reduced claim paid), corresponding to a 1.2% reduction in the total spend for this period (US$11.8 million of US$981.3 million). Compared with members with the lowest social vulnerability (where AI-based FWA screening reduced inappropriate health care reimbursements by 0.9% of overall spend), members with the greatest social vulnerability saw a greater reduction in inappropriate health care reimbursements (1.3%). The results of this study demonstrate that real-time AI-based FWA claims screening for clinical review during prepayment offers potential to reduce inappropriate reimbursements.