Market Power, Transactions Costs, and the Entry of Accountable Care Organizations in Health Care

Abstract

ACOs were promoted in the 2010 Patient Protection and Affordable Care Act (ACA) to incentivize integrated care and cost control. Because they involve vertical and horizontal collaboration, ACOs also have the potential to harm competition. In this paper, we analyze ACO entry and formation patterns with the use of a unique, proprietary database that includes public (Medicare) and private ACOs. We estimate an empirical model that explains county-level ACO entry as a function of: physician, hospital, and insurance market structure; demographics; and other economic and regulatory factors. We find that physician concentration by organization has little effect. In contrast, physician concentration by geographic site—which is a new measure of locational concentration of physicians—discourages ACO entry. Hospital concentration generally has a negative effect. HMO penetration is a strong predictor of ACO entry, while physician-hospital organizations have little effect. Small markets discourage entry, which suggests economies of scale for ACOs. Predictors of public and private ACO entry are different. State regulations of nursing and the corporate practice of medicine have little effect.

Appendix: Data Sources

1.1 Proprietary ACO Data

The Optum Institute proprietary data on ACOs was generated by its global searches of published peer reviewed, grey literature (such as working papers); government publications; news media and systematic Internet-web searches. Searches were conducted in May 2011, October 2011, and May 2012 to identify and document all ACOs that were operating or were in development across the US.³⁰ We found 230 entities that met inclusion criteria and are represented in our May 2012 database as having been in existence in either May 2011 or May 2012.

We included public and private ACOs, ACOs at all stages of development, and participants in an ACO implementation collaborative.³¹ All organizations that clearly used population health management and accountability for population quality measures and that imposed financial risk on providers were included.

We excluded potential ACOs where we found no evidence of investments or documented steps towards ACO formation. Approximately 12 % of entities that were identified as “ACOs” in the general press were excluded from the data due to conflicting information or the absence of this evidence. We excluded older organizations that were formed before 2005. Geographically specific searches were conducted for each of the 50 states and the District of Columbia. We believe that our criteria have resulted in conservative estimates of the number of ACOs. We compared our California data to data that were collected for the same time period by Cattano & Stroud (CS). Nearly all counties with an ACO in our scan data also have one in the CS data. There are 12 counties where our scan data indicate an ACO but the CS data do not. This makes sense because our data use much broader sources. Our scan data are likely to pick up small-scale ACO entry that is not captured in the CS data. Further, our data include ACO in-development and in-pilot entry while the CS data do not.

1.2 Independent Variables

We include many county-level control variables that account for variation in healthcare market structure, health care practice, and regulatory and demographic environments. Median household income and Medicare spending are both normalized for county-level differences in price level, using Medicare’s Geographic Adjustment Factor (Edmunds et al. 2012).

We include hospital, insurer, and physician concentration, as measured by the Herfindahl–Hirschman Index (HHI). The hospital data are from 2010 AHA Annual Survey of Hospitals, and the insurer HHI and HMO penetration are from the 2009 Health-Leaders-InterStudy. The physician site and group HHIs are calculated using 2010 physician census data from SK&A.³²

From the AHA Survey, we obtain several measures of physician-hospital organization structure. Following Cuellar and Gertler (2006), we include the percentage of hospitals that operate as a “medical group without walls”; hospitals that operate as closed hospital-physician organizations; hospitals that operate as open hospital-physician organizations; fully integrated hospitals; and independent practice association (IPA) hospitals. We employ two measures of technology use: the proportion of eligible providers that have received payments from the Medicare and Medicaid Electronic Health Records Incentive Program, as of March 2012³³; and the percentage of physicians who actively used an electronic health record to e-prescribe via the Surescripts Network, as of December 2011.³⁴ These are the only two independent variables that we could not lag, due to data limitations.

Using 2009 data from the Dartmouth Atlas, we include the number of Medicare discharges for ambulatory care sensitive conditions per 1000 Medicare enrollees (rescaled for convenience)³⁵ and also age, sex, and price-adjusted total Medicare Parts A and B per enrollee spending.³⁶

We include corporate practice of medicine laws using 2006 data from Michal (2006), and we use 2007 data on the ability of nurse practitioners to authorize tests and operate without physician oversight from Catherine Dower et al. (2007).

From the 2009 to 2010 Area Resources file, we obtained each county’s 2009 population. Finally, the Area Resources file provided the share of the population eligible for Medicare and Medicaid). See Table 2 for a list of variables, sources, and dates and Table 3 for maximums and minimums of the continuous variables.