Patient Mix Penalizes Safety Net Hospitals

Relying only on risk-adjustment characteristics approved by Medicare—age, sex, and diagnosis—puts safety net hospitals in line for significant financial penalties for higher readmissions rates, researchers say.

More research is suggesting that a hospital's patient mix has a direct bearing on readmissions rates.

The latest evidence comes from Harvard researchers, who published a study this month in JAMA Internal Medicine that factors race, education level, poverty, disability and other socio-economic factors when measuring hospital readmissions.  

Study lead author Michael L. Barnett, MD, a fellow in general internal medicine and primary care at Harvard Medical School and Brigham and Women's Hospital, says that relying only on risk-adjustment characteristics approved by Medicare—age, sex, and diagnosis—puts safety net hospitals in line for significant financial penalties for higher readmissions rates.

Barnett recently spoke with HealthLeaders Media about his study findings and offered some recommendations that could level the playing field for hospitals with more challenging patient demographics. The following is an edited transcript.

HLM: What prompted you to do this study?

Barnett: The hospital readmissions reduction program is very much top of mind for hospitals across the country. The third round of penalties was just announced and almost 2,600 hospitals face penalties of $420 million for excess readmissions and 90% of those hospitals were penalized last year.

It appears that hospitals that serve disproportionate numbers of safety net patients get higher penalties and are disproportionately penalized more severely than other hospitals. There has been an active debate to what extent should the readmissions reduction program—which does not take into account any factors other than age, sex and diagnoses—account for other social and clinical factors.

HLM: How did you compile your study, and what did you find?

Barnett: We used a national representative survey of Medicare patients admitted to the hospital from 2009 to 2012. From this survey, we abstracted a very comprehensive set of 29 different characteristics encompassing a lot of social and clinical factors that CMS currently doesn't account for, including illness, disability, race, poverty, and social supports. We asked to what extent all of these characteristics that are not currently used by Medicare are associated with the risk of readmission after we account for everything that Medicare accounts for.

We found that a majority of them were significantly associated with readmission risks—obvious stuff like education, income, race, and disability. Then we asked which patients are admitted to hospitals with high versus low publicly reported readmission rates, and to what extent are these characteristics distributed differently across these hospitals.

For instance, are hospitals that have high publicly reported readmission rates more likely to have patients with lower education, lower income, more disability, or more illnesses? That is what is going to determine whether or not adjusting to those factors will change the hospitals' expected readmission rate.

We took all these factors we collected and we added them into the factors that Medicare adjusts for in a model to predict readmission rates. We found that incorporating all these characteristics would decrease the difference in the readmission rates between the highest and lowest performers by 48%.

When we adjusted for everything that Medicare uses, we found that the top and bottom quintile hospitals readmission rates were 4.4 percentage points apart. When we incorporated everything that we found in our study we shrunk that difference down to 2.3 percentage points. Even though the numbers seem small, the magnitude of the penalty is directly tied to how far off your readmission rates are from the national average.

So any adjustment that is going to change your readmission rate relative to the national average is going to affect your penalty. Our analysis suggests that the penalty is being disproportionately leveled on hospitals serving those who are at a higher risk of admission and readmission, [and those who are sicker] and poorer.

HLM: Were you able to put a dollar figure on the cost of these penalties for safety net hospitals?

Barnett: It is something we wanted to do and a lot of people have asked for it. Unfortunately, our data is a national representative survey. We don't have a big sample in any individual hospital, so we can't tie a dollar amount to it.

HLM: Why is CMS reluctant to make concessions for socio-economic factors in patient populations?

Barnett: CMS's position has been that adjusting for socio-economic factors potentially holds hospitals that are serving disadvantaged populations to a lower standard. The concern is that it could dis-incentivize hospitals to address healthcare disparities.  

What we propose is a more sophisticated risk adjustment and program design that can lead to a program where you can address disparities and incentivize quality.

There are a number of pieces of legislation recently passed and currently in Congress addressing this issue. There is the IMPACT Act passed in 2014 which directs HHS to examine the effects of socio-economic status on quality and resource use for Medicare patients. If those studies find a relationship between those factors and quality measures, they require the Secretary to make recommendations for how Medicare can adjust for them.

HLM: What can be done to ensure more balanced risk adjustment?

Barnett: We have a couple of suggestions. We acknowledge that risk adjustment is difficult. We have to accept that the data we access isn't the sort of stuff you can gather for every single patient on Medicare. One or our recommendations is based on an alternative payment models like global payments that preserve incentives for hospitals to reduce readmissions without unfairly penalizing them based on their patient populations.

Of course, those kinds of programs require some sort of risk adjustment for how much you pay the hospital.

Another model we put forward is inspired by the ACO program, where ACOs are rewarded based on their improvement versus their historical average, which in the case of the ACO program is based on prior cost growth. We would have to come up with a different kind of metric for something like readmissions, but we want to incentivize improvement at each hospital, and what the readmissions reduction program does is compare every hospital to a national average.

There is an incentive for every hospital to improve, but to have every hospital held to a similar standard. We think that it could be better to incentivize improvement based on prior performance and come up with a new system where the benchmark you are performing against is either held constant or gradually increases over time so hospitals have an incentive to improve and maintain improvements. But we are also not holding them to an unrealistic standard. It's not like every hospital can improve readmissions every year. That's not possible.