Essays on Anomaly Detection and Referral Behavior Among Medicare Providers

ABSTRACT

This thesis comprises three studies that contribute to the understanding of anomaly detection (Chapter 1) and referral behavior (Chapters 2 and 3) among Medicare providers.

Chapter 1 examines how often Medicare hospitals’ yearly (2011-2023) cost reports deviate from Benford’s Law and potentially indicate fraud. Using tests of the first, second, and first two digits, we assess Benford conformity within individual reports and across all reports within each year. We then examine how well the tests flag simulated, anomalous cost reports. Using the observed cost reports, we also examine policy-relevant factors that influence whether the tests flag hospitals.

Chapter 2 addresses the ongoing debate about whether expanding nurse practitioners’ scopes of practice harms patients. We evaluate the effects of state-level scope expansions on NPs’ total referrals, referrals to specific entity types, and Medicare Part B utilization and spending. Examining these referrals captures one step towards the coordination of patients’ care between NPs and other providers. The largely null effects we find are consistent with closely related studies that suggest little harm to patients.

Chapter 3 examines homophily—the tendency for individuals to connect with similar others—in referrals from primary care physicians to cardiologists. Unlike previous research, we study homophily along five characteristics: gender, medical school, medical school graduation year, and predicted race and ethnicity. Our framework makes two main contributions. First, we examine whether “excess” homophily, beyond local cardiologist availability, emerges across PCPs’ referrals. Second, we examine which PCP characteristics influence excess homophily within PCPs’ referrals.

BIOGRAPHY

Carlo Duffy is a fifth-year PhD candidate at IDSS advised by Roy Welsch and Stan Finkelstein. Carlo’s research uses statistical methods to measure and evaluate healthcare provider behavior. He has collaborated with researchers at the MIT-IBM Watson AI Lab and Massachusetts General Hospital. He also received the IDSS Teaching Award for his contributions to 15.363 Strategic Decisions in Life Sciences in Spring 2025 and has served as a teaching assistant for five additional master’s-level MIT courses in healthcare and analytics. Carlo graduated Phi Beta Kappa from Carnegie Mellon University in 2021 with a B.S. in Economics and Statistics.

COMMITTEE

Roy Welsch (co-advisor), Stan Finkelstein (co-advisor), Joseph Doyle

EVENT INFORMATION

Hybrid event. To attend virtually, please contact the IDSS Academic Office (idss_academic_office@mit.edu) for connection information.