Physicians’ Participation In ACOs Is Lower In Places With Vulnerable Populations Than In More Affluent Communities

Research Brief

Physicians’ Participation In ACOs Is Lower In Places With Vulnerable Populations Than In More Affluent Communities

Risks exacerbating disparities in quality of care

Laura C. Yasaitis, William Pajerowski, Daniel Polsky, and Rachel M. Werner, Health Affairs, August 2016


Key Findings: In 2013, physician participation in accountable care organizations (ACOs) was inversely related to the percentage of the local population that was black, living in poverty, uninsured, or disabled or that had less than a high school education. This risks exacerbating disparities in the quality of care received by these vulnerable populations.

The Question

Early evidence suggests that accountable care organizations (ACOs) - networks of doctors and hospitals whose members share responsibility for providing coordinated care to patients - improve health care quality and constrain costs. ACOs are increasingly common in the U.S., both for Medicare and commercially insured patients. However, there are concerns that ACOs may worsen existing disparities in health care quality if disadvantaged patients have less access to physicians who participate in them. Does physicians’ ACO participation relate to the sociodemographic characteristics of their patient population, and if so, why?

The Findings

Nationally, nearly 26 percent of physicians reported that they participated in an ACO. Participation was more common among female physicians than their male counterparts; physicians in large and multispecialty practices; and primary care physicians, compared to those in other specialties.

The authors examined physicians’ ACO participation at the levels of hospital referral region (HRR) as well as Zip Code Tabulation Area, the general geographic area covered by a zip code.

After adjusting for physician characteristics, they find significantly lower rates of ACO participation in areas where a higher percentage of the population was black, living in poverty, uninsured, or disabled or had less than a high school education. 

In the Zip Code Tabulation Area with the lowest quartile of black population, 30.8 percent of all physicians (35.7 percent of primary care physicians) were participating in an ACO, compared to the areas with the highest proportion of black population where 22.9 percent of all physicians (26.4 percent of primary care physicians) were participating in an ACO. There was no consistent relationship of ACO participation to Hispanic population levels. Areas with higher physician ACO participation tended to have a greater supply of all types of physicians but a lower supply of acute care hospital beds.

The Implications

This study suggests that vulnerable patients have less access to physicians who participate in ACOs, and therefore less access to any of the potential benefits of ACOs. This could exacerbate existing disparities in health care quality.

The authors offer two potential explanations for why physicians serving populations with relatively high rates of disadvantaged people could be excluded from emerging ACOs: ACOs may be more likely to form in regions where the overall patient population is more affluent; or ACOs may be more likely to contract with physicians who serve more affluent patients.

They expand on this to suggest possible mechanisms contributing to the variation in physician participation in ACOs:

Given that it may be hard to meet some benchmarks for quality of care among hard-to-treat, vulnerable populations, ACOs may be less likely to locate in regions with these populations than elsewhere. At the same time, even within areas with ACOs, the organizations may exclude physicians likely to care for vulnerable populations in an effort to ensure that the organizations care for populations that will make it possible to achieve high scores on specific quality measures. Rates of ACO participation may also be lower among physicians serving vulnerable populations, compared to other physicians, because of physician choice. Individual physicians or physician groups may choose to hold off on joining an ACO if they deem it would be too difficult to achieve specific quality and spending goals with their patient populations. Finally, physician leaders of early ACOs reported that limited capacity was a major hurdle to entering an ACO contract.

As ACOs proliferate, policy interventions are needed to prevent them from contributing to the worsening of health disparities. The authors suggest providing physician groups that serve vulnerable populations with additional incentives to form an ACO or assistance with start-up costs. Participation might be encouraged by risk-adjusting ACO-linked quality indicators for patients’ sociodemographic characteristics, or rewarding improvements over time.

The authors looked at a snapshot of participation in 2013, so they could not test whether rates of ACO participation were increasing over time among physicians who serve vulnerable populations. Current ACO expansion in Medicaid and federally qualified health centers may also serve to alleviate potential disparities.

The Study

The authors used a telephone-verified national database of U.S. office-based physicians that included self-reported information about their participation in an ACO, both public and commercial. The data used was collected during 2013 and included more than 500,000 physicians, the vast majority of those currently practicing. The authors estimated physicians’ ACO participation at the level of the hospital referral region (HRR) as well as the more targeted Zip Code Tabulation Area level, and found that variations in ACO penetration largely followed HRR boundaries.

The authors used summary statistics of the Zip Code Tabulation Area [of the physician’s practice location] from the American Community Survey for 2009-2013 as a proxy for the sociodemographic characteristics of the population that the physician was likely to care for. The population characteristics included: the percentages of the population that had less than a high school education and that were black, Hispanic, living in poverty, uninsured, or disabled. They also examined and controlled for physician-level variables: sex, number of physicians in the practice, whether or not the practice was multispecialty, rurality of the practice location, and medical specialties.

Lead Author: Dr. Laura Yasaitis

Laura Yasaitis, PhD, is a Postdoctoral Researcher in Health Policy at the Perelman School of Medicine. Her research interests include the study of large datasets, including surveys and administrative data, to better understand the quality and outcomes of health care in the United States. She is also interested in the application of spatial analysis methods to further our understanding of local trends in health and health care. 

LDI Research Briefs are produced by LDI's policy team. For more information, please contact Janet Weiner at weinerja@mail.med.upenn.edu.