乌鸦传媒

SAN DIEGO -- The inability to account for wide variation in the use of early do-not-resuscitate (DNR) orders -- and the underlying reasons for the orders -- complicates comparisons of stroke mortality among hospitals, researchers found.

Across hospitals in California, the percentage of stroke cases associated with a written DNR order in the first 24 hours of admission varied from 2.2% in the fifth of hospitals with the lowest use to 23.2% in the fifth with the highest use, according to , of the University of Rochester Medical Center in New York.

The hospitals with the highest use also had the highest in-hospital mortality rate, which was about twice as high as that seen in the hospitals with the lowest percentage of DNR orders (8.7% versus 4.2%), he reported at the International Stroke Conference here.

When he and his colleagues ranked hospitals according to their in-hospital mortality rates, they found that a model that incorporated DNR use provided a much different listing compared with one that did not.

The findings -- which were published simultaneously online in -- indicate that failing to account for DNR orders provides inaccurate assessments of mortality, which is tracked as a marker of quality of care. The Centers for Medicare & Medicaid Services (CMS), for example, recently implemented a system to track 30-day stroke mortality and readmission rates.

But correcting the problem might not be as simple as statistically adjusting for variation in rates of DNR use, Kelly told 乌鸦传媒, because early DNR orders can reflect either excellent or substandard quality of care depending on the patient or family preferences in a specific situation.

A DNR order might reflect a pre-existing wish of the patient that was written before the stroke -- in which case, an outcome of death would be consistent with good, patient-centered care. Or it might reflect pessimism on the part of the treating physician that a poor outcome is inevitable and further treatments would not have any benefit -- in which case, death could be an indicator of lower quality of care.

"We have this wide variation and we don't understand why," Kelly said. "So to me that [uncertainty] just seems like a reason not to use mortality as a marker of quality of care because I think this study would show that there's a lot that goes into this that we still don't understand."

Kelly and his colleagues used information from the California State Inpatient Database to identify admissions for ischemic stroke in patients 50 and older that occurred from 2005 through 2011. The analysis included 252,368 stroke cases in 355 hospitals.

Overall, 13.3% of the cases had a DNR order placed within 24 hours of admission; the researchers could not distinguish between DNR orders that existed before admission and those that were newly written after the stroke. The unadjusted in-hospital mortality rate was higher for patients with an early DNR order (19.2% versus 3.9%, P<0.001).

Not accounting for differences in DNR use had a large impact on ranking hospitals according to their inpatient mortality rates. A model that included DNR plus comorbidities, vascular risk factors, and demographics -- versus one without DNR use -- was associated with relative decreases in in-hospital mortality of 10% or more in 190 hospitals, 20% or more in 37 hospitals, and 30% or more in seven hospitals, and with relative increases of at least 10% in 84 hospitals, at least 20% in 32 hospitals, and at least 30% in three hospitals.

"I think the study is very important because it adds to a growing list of factors that need to be accounted for in any analysis of mortality that are not available in most records, particularly administrative claims data," according to , director of stroke services at Massachusetts General Hospital in Boston and a spokesperson for the American Stroke Association.

The concern about the appropriate risk adjustment of the CMS models for stroke mortality and readmissions is an issue Schwamm is very familiar with because he is an author of an American Heart Association/American Stroke Association presidential advisory that came out this week dealing with that issue.

The advisory in the models -- which, the organizations stated, would unfairly skew any rankings against primary stroke centers and other hospitals caring for the sickest patients. But, Schwamm said, stroke severity and DNR use are strongly intertwined because it is the patients with the most disabling strokes on admission that are most likely to have a DNR order written.

Unless mortality "can be adequately adjusted for these kinds of risks," he said, mortality should not be used to judge quality of care at hospitals treating patients with stroke.

The AHA/ASA are in continuing discussions with CMS to better refine the models, he said.

From the American Heart Association:

Comment