The study, based on four years of data from the University of Michigan Health System's unique repository of clinical patient information, found that patients transferred to a major center from other hospitals are sicker and more likely to die than others, and that this "transfer effect" is not accounted for by hospital-rating tools.
The effect has long been suspected by large, referral private, public and university-affiliated medical centers, but was never fully quantified until now.
The new results, published by a UMHS team in the June 3 issue of the Annals of Internal Medicine, show the effect is enough to damage a hospital's score on measures that patients, insurers, employers and the government use to make health care choices.
The authors call upon those who prepare hospital rankings and reports to either include a hospital's transfer rate in the statistical analysis of its patient outcomes, or to analyze outcomes with and without transfer patients included. This, they say, will improve the representation of hospital quality that patients and their families can use to guide their health care decisions.
"Hospitals are now asked to prove the quality of their care through hard numbers, and the results are used in everything from advertising to contract negotiations with major employers," says lead author Andrew Rosenberg, M.D., assistant professor of anesthesiology and internal medicine at UMHS. "But we find that such assessments may be biased against the most advanced medical centers, because they're the hospital of last resort for patients with the most complex and severe problems. That means quality models need to change, to make scores and rankings more accurate."
Rosenberg led the study along with Rodney Hayward, M.D., a professor of internal medicine at UMHS and the director of the Health Services Research Center at the Ann Arbor VA Healthcare System, while Rosenberg was a Robert Wood Johnson Clinical Scholar at the U-M Medical School.
The data were from 4,579 ICU admissions to the U-M adult medical ICU, involving 4,208 patients -- 25 percent of whom were transferred to UMHS from other hospitals.
Even with the most modern and valid tools to account for differences in diagnoses, severity of illness and other predictors of outcome, compared with patients who were admitted to the ICU directly, the transferred patients had 38 percent longer ICU stays and 41 percent longer hospital stays, and were twice as likely to die in the hospital.
Though the transfer patients' rates of complications and deaths were still below the expected ranges set by quality models, meaning the care they received at UMHS was of high quality, Rosenberg explains that their poor health when they arrived from other hospitals put them -- and the hospital's scores -- at a disadvantage.
As a result, the authors report, a hospital that gets 25 percent of its ICU patients as transfers from other hospitals would show an extra 14 deaths for every 1,000 admissions, as compared with a hospital that accepts no ICU transfer patients and provides exactly the same quality care. This seemingly small 1.4 percent difference would be enough to drive down the hospital's score.
Rosenberg believes the "transfer effect" holds true for other hospital departments – but that it will be harder to prove because the most thoroughly validated tools to predict outcomes exist mainly for intensive care units. It is also very costly to collect the amount and type of data needed for these analyses and is not done routinely in many medical settings.
He and his colleagues have already begun to analyze data from other UMHS units, starting with other ICUs that serve surgical patients. They also hope that their study can be replicated at other major medical centers, though they recognize that some may not have sufficient data.
"We were able to do this study in large part because our hospital has a dedicated data warehouse for ICU data, and collects more than the basic parameters -- including patient transfer status," says Rosenberg, who directs research in critical care medicine for the U-M Department of Anesthesiology and treats patients in the U-M's Cardiothoracic Intensive Care Unit.
He credits co-author Charles Watts, M.D., former UMHS chief of clinical affairs and now associate dean for academic affairs at Northwestern University's Feinberg School of Medicine, with championing the effort to collect this information at UMHS.
UMHS uses one of the most popular computerized systems for collecting and assessing intensive care patient health, and scoring the quality of care. It scores based on comparisons of actual patient outcomes with expected outcomes generated by a computer model. The system, called APACHE III for Acute Physiology and Chronic Health Evaluation, does consider patient origin to some extent in the way it calculates a hospital's predicted morbidity and mortality rates. But Rosenberg notes that there are few, if any other, methods of accounting for large numbers of transferred patients in clinical quality benchmarks.
The whole idea of assessing hospitals based on hard data on patients' death and complication rates started in the 1980s, when the federal agency in charge of the Medicare system began releasing death rates for Medicare patients treated in hospitals around the country. Realizing that some hospitals took care of sicker patients than others, and provided more complex and sophisticated care, the agency and other entities began taking a hospital's "case mix" into account, and including other data such as numbers of certain operations performed each year.
Today, the number and variety of ranking and scoring programs has ballooned, with everyone from the major hospital accreditation organization -- called JCAHO -- to the magazine U.S. News & World Report issuing report cards on hospital quality.
The hospitals, in turn, use those scores and rankings to negotiate payment rates with public and private insurers, to attract doctors and staff, and to convince potential patients to choose them.
"If you're not in the top 10, top 100, best-of-the-best list, there's a perception that there's something wrong," Rosenberg says, adding that UMHS touts its top rankings just like others.
In this climate, academic medical centers and similar institutions such as major urban hospitals have claimed that their role as the "provider of last resort" puts them at a disadvantage, even though their case mix scores reflect the fact that they take patients who are near death and use everything from organ transplants to advanced life-support systems to try to save them.
The addition of these "extra" severe illnesses and deaths among patients who come to the major centers only after smaller and less well-equipped hospitals fail or give up, has been seen as a drag on quality scores. In the study, the transferred patients had higher Acute Physiology Scores at admission and discharge than directly admitted patients, and were more likely to have complex problems such as severe infections and upper gastrointestinal bleeding.
Rosenberg emphasizes that the transfers seen in the study were not merely the result of "dumping" -- the practice of transferring critically ill patients out of a hospital to another, specifically to keep them out of a hospital's length-of-stay and mortality statistics. "Dumping may happen, but accepting transfer patients is what major medical centers are here for," he says.
Which is why, Rosenberg says, it's so important that this new "transfer effect" study puts numbers behind those centers' claims. He and his colleagues hope their findings will be taken into account by those who score and rank hospitals, if the effect can be found to exist at other medical centers that take large numbers of transfer patients.
"We have to respect that the public, the payors and the insurers have a right to quality and value for their health care dollar," Rosenberg says. "But we also need the assessments of that quality and value to take all factors into account, especially factors unique to major academic medical centers. We provide complex, intensive and costly care, and develop life-saving technologies and treatments, that no other hospitals have -- and we get the sickest patients as a result."
In addition to Rosenberg, Hayward and Watts, the paper's authors include Cathy Strachan, MSRN, who runs the UMHS APACHE database service; and Timothy Hofer, M.D., an associate professor of internal medicine at UMHS.
Both Hayward and Hofer are also members of the Robert Wood Johnson Clinical Scholars Program, and of the Department of Veterans Affairs Health Services Research & Development Service at the VA Ann Arbor Healthcare System.
The study was funded by Rosenberg's former Robert Wood Johnson Foundation grant, and by VA grants to Rosenberg and Hofer.
Reference: Annals of Internal Medicine, Vol. 138, No. 11, 3 June 2003, pp 882-890.
Journal
Annals of Internal Medicine