Videos

Speakers: • Sal Aguinaga, Master Data Scientist, Deloitte | AI Center of Excellence • Dr. Sanmitra Bhattacharya, AVP, Data Science, Deloitte | AI Center of Excellence Session type: Full Length Abstract: Each year billions of insurance claims are submitted by healthcare providers. U.S. healthcare spending continues to grow over five percent year-over-year and accounts for approximately 20 percent of the Gross Domestic Product. The National Health Care Anti-Fraud Association conservatively estimates healthcare fraud at three percent of total health care costs, which in 2019 represented over a hundred billion dollars in fraud. The Centers for Medicare & Medicaid Services and other regulators mandate fraud, waste, and abuse (FWA) surveillance by payors of healthcare claims. Screening providers based on their risk profiles across various dimensions of FWA is a key component of such surveillance. Our project identifies providers sharing common risk signatures with other providers – uncovering pairwise similarity using graph-theoretic algorithms and graph neural network (GNN) methods. This two-pronged solution works with Neo4j’s graph engine at its core by applying Graph Data Science and serving quality graph datasets to external state-of-the-science GNN training workflows. The objective of these two approaches is to produce complementary groupings of providers with common risk signatures. Our analyses reveal the likelihood of hidden or unknown relationships between providers across various FWA dimensions.