Keywords: Assortativity, Diffusion, Fraud, Medicare, Peer effect, Social network
The Federal government is pursuing Medicare fraud, yet the prosecuted cases are only the tip of the iceberg relative to potentially fraudulent or “gray area” activities. Little is known about the diffusion of such behavior and whether it transmits through physician networks, making the identification of fraudulent activity through networks one approach to improving the quality of care while reducing costs dramatically. We develop methods to assess whether: 1) physicians who are the highest billers of Medicare preferentially share patients; 2) non-outlier physician with strong connections with outlier physicians are more likely to become outliers. We use 100% of national Medicare claims data from 2008-2014 to identify unusual Medicare billing activity for home health care services (HHC). For each treatment, we build a network of the physicians in which edges in the network quantify the number of distinct patients treated by each pair of physicians. Motivated by an initial finding of positive degree assortativity of the networks, we estimate a peer effects regression model to gauge whether gauge whether there is evidence of physician-to-physician diffusion in the network.