Can Incentives to Improve Quality Reduce Disparities?

Each year the Agency for Healthcare Research and Quality (AHRQ) submits annual reports to the Congress on the state of health care quality and health care disparities. Results of the 2008 National Healthcare Quality Report and National Healthcare Disparities Report reflect continued trends in both areas. Across all settings and patient groups, quality continues to improve at a modest rate (1.4 percent for core measures). Yet disparities associated with individual race, ethnicity, socioeconomic position, and other factors remain pervasive (Agency for Healthcare Research and Quality 2008a,b;). In recent years, impatience with the slow pace of quality improvement has inspired numerous private-sector initiatives that make a portion of payments to physicians and hospitals dependent on improved quality, often referred to as “pay for performance” or “p4p” (Agency for Healthcare Research and Quality 2006). But can such financial incentives to improve quality reduce care disparities?

At this writing, the U.S. Congress is working hard to pass health care reform legislation. In addition to expanding access to insurance for the 47 million Americans who currently are uninsured, the bills reflect policy makers' recognition that sustaining health care reform will require accelerated efforts to achieve high-quality, affordable care for all Americans. Current bills include multiple sections and initiatives to encourage improved quality. These include a strong focus on public reporting and transparency, as well as pilot and demonstration projects that evaluate the feasibility of linking financial incentives to achieving high-quality care. The demonstration led by Premier, in which participating hospitals agreed to link clinical performance with financial bonuses or penalties, represents an important prototype of future demonstrations and pilot projects under discussion as a core component of health care reform (Lindenauer et al. 2007).

Efforts to improve quality of care can lead to reductions in disparities (Sehgal 2003). For example, AHRQ's 2008 National Healthcare Disparities Report found that improvements in preventive care, chronic care, and access to care significantly reduced disparities in mammography among Asian, American Indian, and Alaska Native women; counseling for smoking cessation low-income adults; and appropriate timing of antibiotics to prevent surgery-related infections among American Indians and Alaska Natives. However, much quality improvement that we track in the National Healthcare Quality Report and National Healthcare Disparities Report is not associated with significant decreases in care disparities across populations. Moreover, some have speculated that activities to improve quality could actually increase disparities if some groups are less able to respond to such activities than others (Casalino et al. 2007; Hasnain-Wynia and Jean-Jacques 2009;). For example, mailed patient reminder cards may be less effective in populations with limited literacy or limited English proficiency. Public reporting can in theory empower patients to switch from lower performing to higher performing providers. But if minorities have less provider choice than whites, larger disparities could ensue. Decision support tools could also exacerbate disparities if providers in low-income neighborhoods are less likely to have the health information technologies needed to use them.

In this issue of the journal, Ryan examined the impact on disparities of a hospital-based pay-for-performance demonstration, the Premier Inc. Hospital Quality Incentive Demonstration (PHQID) (Ryan 2010). The incentive program paid a 2 percent bonus on Medicare reimbursement rates to hospitals performing in the top decile of a composite quality measure for certain conditions. Ryan's study focused on whether hospitals engaged in patient avoidance behavior, which occurs when providers avoid treating patients perceived to be likely to reduce their performance in areas where public reporting or financial incentives are in place. Patient avoidance behavior may result in reduced access to care for racial and ethnic minorities, who may be perceived to have higher risk for poor outcomes than white patients.

Ryan's results demonstrate that this hospital-based pay-for-performance program had only a minimal impact on access to care for racial and ethnic minority Medicare beneficiaries. Adjusted admission rates to PHQID hospitals for acute myocardial infarction (AMI), congestive heart failure (CHF), and pneumonia showed no significant pre–post demonstration differences. For example, 17.5 percent of whites, 18.3 percent of blacks, 15.8 percent of Hispanics, 17 percent of other races, and 17.9 percent of nonwhites were admitted for heart failure to PHQID hospitals before PHQID implementation, compared with 17.7, 18.8, 16.1, 16.7, and 18.3 percent, respectively, after implementation. Similarly, 20.8 percent of whites, 20.9 percent of blacks, 17.6 percent of Hispanics, 21.3 percent of other races, and 20.6 percent of nonwhites were admitted for AMI before PHQID, compared with 21.1, 21.4, 17.9, 20.0, and 20.7 percent, respectively, post-PHQID. The only significant difference after implementation of the pay-for-performance program was for “other race” Medicare beneficiaries with AMI admissions. However, this difference was small (only 1.5 percentage points) and occurred largely during the predemonstration period. Moreover, there was little evidence that minority patients diagnosed with AMI became less likely than whites to receive coronary artery bypass graft surgery at PHQID hospitals after the program was implemented. The authors concluded that there was no evidence of minority patient avoidance and that the program did not reduce access for minority patients overall.

Ryan's findings suggest additional avenues of inquiry. For example, would results differ in hospitals with dedicated hospitalists compared with hospitals with attending physicians, who may not be affected directly by pay for performance? Would results differ outside hospitals, in ambulatory care settings where providers may have less control over patient adherence to medications and instructions? Would results differ if the size of bonuses were larger than the 2 percent provided by the Medicare demonstration? Would results differ for other conditions where minority patients may have greater unmeasured risk than whites? For the conditions included in Ryan's study — AMI, CHF, and pneumonia — minorities often have lower inpatient mortality rates compared with whites. In this case, informed hospitals may engage in selection that favors minority patients over whites under pay-for-performance programs.

It is also important to consider how pay-for-performance approaches could be modified to promote reductions in disparities, rather than just minimizing adverse risk selection. Hospitals located in the poorest neighborhoods have much worse performance than those in more affluent areas (Benincasa and Brooks 2006). Straight pay-for-performance could simply make the rich hospitals richer and the poor hospitals poorer, to the detriment of the residents of poor neighborhoods. One solution is to provide funds for quality improvement infrastructure and technical assistance to poor performers. Another solution is to include incentives related to pay for improvement. This would enable providers at the bottom, who are improving but have not yet attained high performance, to receive bonuses. Maryland includes such an option in its hospital pay-for-performance program.

Another way to use incentives to reduce disparities is to pay for disparities reduction. Massachusetts Medicaid has included elements of this as part of its health care reform. The challenges with this approach include what to do with hospitals that treat few minority or low-income individuals and how to balance overall quality and disparities goals. With respect to sample size, it would be possible to aggregate performance on quality at the institutional level and disparities at a community or state level. However, we have no information about how shared accountability for disparities would be perceived. For organizations that serve larger minority populations, other questions loom large. For example, is mediocre care with no disparities better or worse than overall superior care with larger disparities?

Regardless of the approach, all pay-for-performance policies share one critical need: data on racial and ethnic minorities to track policy effects on disparities. Unfortunately, Medicare data are notoriously poor at identifying people of color (Eggers and Greenberg 1998). Large numbers of Hispanics are misclassified as white, which would tend to mask differences across groups. “Other race” is a mix of groups and of limited utility. In addition, the recently released Institute of Medicine report highlights the limitations of broad race and ethnicity categories and the importance for communities to identify more fine-grained ethnic groups (Institute of Medicine 2009). These specific “granular” ethnicities may allow more efficient targeting of intervention to groups experiencing the largest disparities in their localities.

Health services researchers have articulated multiple concerns about the potential for unintended consequences of pay-for-performance programs, particularly the potential to widen disparities. Notwithstanding limitations, Ryan's results should offer some reassurance. However, the urgency underlying current health care reform efforts should also suggest that monitoring the impact of future initiatives on both quality and disparities is critical but not sufficient. Use of pay for performance and other financial incentives to improve quality may be able to reduce care disparities. Explicitly recognizing disparities reduction as a quality metric under pay for performance and financially rewarding providers that achieve equity in health care may be the most direct path to ensuring high-quality health care for all Americans.