The Affordable Care Act has propelled health care reform forward, paving a path for more than 20 million Americans to gain access to medical care while catalyzing a decrease in Medicare spending growth, which is projected to save $1 trillion between 2010 and 2020. Part of these savings are expected to come from preventing costly hospitalizations and rehospitalizations through better coordination of care and shifting care to less costly settings.
In the ACA, Congress carefully detailed how “excess” rehospitalization rates should be calculated for individual hospitals, and clearly outlined how these rates should be translated into hospital-specific payment penalties. Following federal statute, the Centers for Medicare & Medicaid Services, or CMS, created the Hospital Readmissions Reduction Program and decreed that acute care hospitals should be penalized based on their unplanned 30-day rehospitalization rate among patients with several common conditions, starting with heart attack, pneumonia and heart failure.
Over the past 3 years, these penalties have been extended to chronic obstructive lung disease, total hip and knee replacement, and coronary bypass graft surgery. Claiming that “hospitals can often prevent these events,” CMS added readmission measures to its public reporting website, Hospital Compare, and created a 5-star quality rating scheme for hospitals in which nearly a quarter of the overall rating comes from readmission measures.
However, recent research has raised serious doubts about some of the assumptions that underlie the Hospital Readmissions Reduction Program.
First, we now know that readmissions are very hard to predict based on current approaches. The methods used to predict if a patient will be readmitted are only modestly more accurate than a coin toss. To illustrate this problem, consider a hospital that has ten sick patients who leave with a 50% risk of returning to the hospital by the 30th day. While a coin flip would predict five of the ten patients’ outcomes correctly, the current approach gets six outcomes right. Factors related to a patient’s home situation, social and economic circumstances, and community are now known to be important predictors of readmissions (but are ignored by CMS).
Second, the 30-day time window is too long to capture quality and safety problems attributable to hospitals. In our study, published in the October issue of Health Affairs, we found that the greatest opportunity to discern differences in the quality of hospitals occurred during the first seven days following the hospital stay. During these seven days, the hospital quality signal dissipates quickly and reaches a minimum by the 10th day. Focusing attention on shorter post-hospital periods will better isolate the impact of hospital care from all of the other drivers of readmissions. Indeed, previous research has shown that when hospitals send patients home too early, while they are still unstable, early readmissions may increase.
Even when considering only this shorter interval, the hospital quality signal is small. In fact, less than about 3 percent of a patient’s readmission risk comes from differences in quality between hospitals. Another recent study by Dr. Daniel Brotman and his colleagues at Johns Hopkins University reported that higher readmission rates were associated with lower mortality rates at the hospital level for three of the six reported conditions, despite the fact that higher quality care would be expected to prevent both readmissions and deaths. Together, these findings suggest that readmission rates are a poor measure of hospital quality, although they are an important component of resource use.
In a recent blog, the chief medical officer of CMS reported that the Hospital Readmissions Reduction Program and the related Partnership for Patients have led to an 8 percent drop in readmissions between 2010 and 2015. However, these reductions appear to have reached a plateau since 2014. Given this observation and emerging evidence from our study and others, it is time for a new approach to the problem of avoidable hospital admissions. National policy cannot continue to be narrowly focused on hospitals as the principal driver of readmissions.
In order to pursue the goals established by Congress, we propose three potential solutions.
First, the 30-day interval for capturing readmissions, which is currently used for hospital profiling, should be shortened and better tailored to the actual time course of preventable complications after each conditions or procedure.
Second, readmission rates should be regarded more as measures of resource use, or as measures of the performance of health care systems (such as integrated health care organizations), and less as measures of hospital quality. Giving readmission rates so much weight in the 5-star hospital rating scheme seems particularly misleading.
Third, federal policy should recognize that readmissions reflect many things that happen to patients beyond the hospital walls, including quality and coordination of care involving physicians and other primary care providers, home health agencies, and nursing homes. A more successful and sustainable approach to improving quality while reducing spending will require that hospital readmissions be considered within a broader context, including ALL potentially avoidable hospital use and ALL of the health care providers who contribute to that overuse.
As high-quality hospital care is an important goal of national health policy, we should take care to verify that our measurement tools truly reveal the quality of care that patients receive. When they don’t, we need to reconsider our measurement tools and reassess how we are using them.
David L. Chin is a postdoctoral scholar at the Center for Healthcare Policy and Research at the University of California, Davis.
Patrick S. Romano is a professor of medicine and pediatrics at the University of California, Davis.