Research Summary: From Prediction to Explainable Provider Behavior Profiles for Fraud, Waste, and Abuse Review

The research from arXiv highlights limitations in current predictive modeling for identifying fraud, waste, and abuse (FWA) in provider behavior, particularly within claims data analysis. While claims data can indicate changes in behavior, it often fails to explain the underlying reasons. The paper suggests that predictive models, especially those relying on forecasting expected utilization, provide limited value beyond simple persistence and do not adequately explain the significance of observed deviations. It notes that residuals in such models often conflate various factors, making it challenging to isolate potentially concerning behaviors effectively for FWA review.

Why it matters

This analysis is crucial for organizations relying on data analytics for compliance and risk management, particularly in domains susceptible to fraud, waste, and abuse. It underscores the need to move beyond purely predictive models to explainable frameworks that can clarify 'why' behavior changes occur, thereby enabling more targeted and effective interventions.

Key insights

• Claims data indicates changes in provider behavior but does not explain the underlying causes.
• Effective FWA review requires identifying material behavior, locating specific codes and financial drivers, and testing plausible explanations.
• Predictive modeling, which flags deviations from expected utilization forecasts, offers limited value unless it surpasses simple persistence and provides clear explanations for deviations.
• In the analyzed quarterly provider-procedure data, the latest observation accounts for most forecastable variation, with additional model structure yielding little accuracy improvement.
• Residuals in predictive models often combine growth, service-line shifts, code maintenance, and incomplete observation, obscuring potentially concerning behaviors.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.28477