OBJECTIVES: The Johns Hopkins Overuse Index (JHOI) was designed as a composite measure of systematic overuse that was operational in claims data. Unlike measures of geographic variation, it uses detailed algorithms to identify instances where the procedure is used in patients that are unlikely to benefit. Using multi-level modeling techniques to measure overuse across 306 Hospital Referral Regions (HRRs), the JHOI has been shown to be highly predictive of higher costs and worse outcomes in an HRR. The purpose of this study was to explore possible drivers of overuse by comparing the overuse index to structural indicators available in the Dartmouth Atlas.  METHODS: A normalized JHOI was derived for 306 HRRs using a 5% sample of Medicare claims from 2008 and reflected overuse within each HRR. Inpatient and outpatient claims were used and claims from nursing homes were excluded. Potential drivers of overuse were derived from Dartmouth Atlas and regressed upon JHOI using ordinary least squares (OLS). Lagged covariates were used to prevent reverse causation. Huber-White standard errors were used to address clustering of HRR within states. RESULTS: Exploratory analyses identified that the JHOI was positively associated with the number of acute-care beds (p<0.001), medical specialists (p<0.005), surgeons (p<0.001), and the number of Medicare beneficiaries (p<0.01) in an HRR. Regions with higher rates of appropriate testing of patients with diabetes (a proxy of process quality) (p<0.05), more physicians (p<0.001), and more nurses (p<0.001) had less systematic overuse. CONCLUSIONS: The use of cross-sectional data, collinearity between possible explanatory variables, and the relatively large unit of observation limit the interpretation of these findings as causal. This said, the intuitive nature of these findings suggests that there are likely structural drivers of overuse across HRRs in the US. Structural variations imply that the causes of widespread overuse can be identified and potentially controlled.