Physician Networks And Low-Value Cancer Screening: A Cross-Sectional Economics Study Of 2008 Medicare Claims Data

Abstract

Recent developments in healthcare economics literature suggest that various low-value healthcare services can be reliably detected through querying claims data; further, advances in the study of physician networks relationship with the cost and quality of care provoke the hypothesis that physician networks may be associated with the usage of low-value services. In this study, we investigate the distribution of low-value cancer screening services amongst physician networks in 5 states, testing for differences between networks using econometric methods and exploring potential drivers of low-value service usage amongst networks. Using data mining techniques on 2008 Medicare claims data, we identify three low-value cancer screening services amongst 417 physician referral networks developed in prior research. A fixed effects negative binomial model and a multivariate regression model were used to test for significance of low-value service usage variation between networks and for investigating potential associative significance of network-level characteristics, respectively. We found statistically significant variation (p<0.01) in the usage of two of the three cancer screening services between the physician networks after controlling for patient and physician characteristics. Additionally, we identified three network variables which significantly predict the dependent variable (low-value services per 100 “at-risk” beneficiaries in the network): percentage of network physicians trained in the U.S.; percentage of physicians with an MD (compared to DO); and the mean-adjusted valued degree (a measure for the level of “patient-sharing” within the network).