Map. Proportion of hospitals penalized in FY2019 by state under the Medicare Hospital Readmission Reduction Program. Source: Authors’ analysis of the Kaiser Health News Readmission by State Data File.
Going to the hospital is more than a drag. For patients, it can be a frightening experience, dangerous to one’s health, a burden on family and caregivers, and very expensive. Policies to reduce preventable return visits to the hospital are good for patients – and good for Medicare’s bottom line.
Medicare’s Hospital Readmission Reduction Program (HRRP) has penalized hospitals with excess readmissions for selected conditions to incentivize fewer preventable hospital readmissions. In 2012, the first year of the program, more than 2,597 hospitals were penalized; in 2019, 2,599 hospitals are slated to be penalized (Map shows the proportion of hospitals in each state that will be penalized in 2019). In an evaluation of the HRRP, the Medicare Payment Advisory Commission estimated that potentially preventable 30-day hospital readmissions have declined from 11.2% in 2010 to 9.7% in 2016 [pdf]. With these lower readmission rates, MedPAC estimates that Medicare spent $2.28 billion less in 2016 than it would have otherwise.
Despite these generally positive outcomes, and limited evidence that the HRRP has resulted in higher rates of mortality or observation stays (though a recent study found higher readmission rates in safety net hospitals among black patients with non-HRRP targeted disease), the HRRP has come under considerable criticism. One concern is that hospitals serving low-income and racial and ethnic minority Medicare patients are more likely to be penalized than other hospitals, effectively taking resources away from the hospitals that may most them. Another related concern is that hospitals are being penalized for readmissions that may be driven by patient and neighborhood factors outside of their control. Individuals living with social risk factors such as lower education and income are more likely to return to the hospital within 30-days.
In this month’s Medical Care, Darrell Gaskin, Hossein Zare, Roza Vazin, DeJa Love, and Donald Steinwachs address the concern that the HRRP may be negatively affecting racial and ethnic minority communities. In their paper, the authors examine the relationship between the racial and ethnic composition of hospital service areas on the probability of being penalized under the HRRP.
In their analysis, Gaskin and colleagues push beyond the typical binary categorization of race (white vs non-white or black vs white) to examine patterns by the proportion of residents identifying as black, Asian, Hispanic, and other (which is a catchall category including Native American, Alaskans, Other Pacific Islanders, and people identifying as multiple racial groups).
Using this approach, they find different patterns by the concentration of each racial and ethnic group. As the proportion of black and Asian residents in a community increase, the hospital’s probability of receiving a penalty also increased. However, as the proportion of Hispanic residents in a community increased, the likelihood of a hospital being penalized was flat, and as the proportion of residents identifying as other increased, the likelihood of a penalty for a hospital decreased.
This study finds that hospitals in some minority communities may be disproportionately affected by the HRRP penalty. This finding raises important concerns regarding the cumulative impacts of the HRRP penalties on access to hospital care in these communities and racial and ethnic differences in health.
The authors recommend that adjusting for social risk factors in the HRRP calculation may lessen the impacts on hospitals serving communities with higher social risk factors. However, they caution that risk adjustment will not fully mitigate this issue. Gaskin and colleagues’ recommendations are consistent with the recommendations from the National Quality Forum and other research groups.
Congress, through the 21st Century Cures Act [pdf], directed Medicare to use a different approach: peer group stratification where the peer group is defined by the proportion of dual enrollees served (people enrolled in both Medicaid and Medicare).
Briefly, a hospital’s penalty status is determined by its performance on six selected conditions: acute myocardial infarction, heart failure, pneumonia, chronic obstructive pulmonary disease, hip and knee arthroplasty, and coronary artery bypass graft. For each condition, CMS calculates the hospitals’ observed and expected performance, using patient-level data adjusting for clinical risk factors. The ratio of the observed to expected values (the O/E ratio) is then multiplied by the national readmission rate to obtain a risk-standardized readmission rate for each hospital. Within each peer group, a hospital’s risk-standardized readmission rate is compared to the group median. Hospitals above the median in their peer group are penalized; hospitals below the median are not.
The peer-group stratification approach effectively reduces the number of hospitals serving high proportions of dual enrollees that are penalized. According to our analysis of the Medicare FY2019 HRRP impact file, we estimate that 45 hospitals treating the highest proportion of duals will not be penalized in 2019, even though they would have in previous years.
While one goal has been achieved, it is important to note that peer-group stratification may create other problems. If hospital quality and the proportion of duals treated are correlated then, as Roberts and colleagues noted: “CMS’s approach would set lower standards for hospitals serving those patients.” By setting different standards—effectively setting different benchmarks for each peer group—CMS may slow or even reverse the HRRP’s impact on reducing overall readmission rates and reducing disparities in readmission by race.
An alternative approach is to use a regression model that accounts for patient and hospital-level factors, called a mixed or hierarchical model. In the context of hospital readmissions, many factors may contribute to the variation between hospitals: administration procedures, provider treatment choices, and discharge processes. Variation may also be due to patient-level factors including clinical factors, education level, and income. The hierarchical regression model can account for this using a hospital-specific random effect, which represents the hospital’s likelihood of readmission accounting for patient-level factors. This is the modeling approach currently used by CMS to calculate the risk-standardized readmission rates, and it is sufficiently flexible to account for other area-level risk factors collected through the US Census or patient-reported data.
Other policy approaches may also reduce some of the downsides of peer-group stratification, including using fixed benchmarks, rewarding improvement instead of absolute achievement, and using a hospital-wide readmission measure instead of condition-specific measures.
As CMS and hospitals roll out the new peer-group stratification approach, researchers will have the opportunity to examine whether this new approach affects the downtrend in readmission rates and its impact on health disparities. Following Gaskin and colleagues’ lead, researchers should also be vigilant about the impacts on access to care and quality across communities, not just within the Medicare program.
edugoff