Anomaly Detection Models for Outlier Provider Behavior in Cost and Treatment Patterns
Keywords:
Healthcare Fraud Detection, Anomaly Detection Models, Graph Neural Networks, Temporal Sequence Analysis, Explainable Artificial IntelligenceAbstract
Healthcare fraud detection tools are constantly challenged by the ability to detect advanced billing anomalies that cost billions of dollars each year in insurance initiatives. The conventional rule-based approaches are shown to have inadequate capability to identify changing fraud schemes, which utilize convoluted temporal arrangements and network relationships between claims data. State-of-the-art machine learning architectures that combine variational autoencoders, long short-term memory networks, and graph neural networks offer improved detection by differentiating between normal provider behavior and indicators of possible fraud
References
National Health Care Anti-Fraud Association, "The Challenge of Health Care Fraud". [Online].
Available: https://www.nhcaa.org/tools-insights/about-health-care-fraud/the-challenge-of-health-care
fraud/
Centers for Medicare & Medicaid Services, "2020 Medicare Fee-for-Service Supplemental Improper Payment Data," U.S. Department Of Health & Human Services, 2020. [Online]. Available: https://www.cms.gov/files/document/2020-medicare-fee-service-supplemental-improper-payment
data.pdf
