Hospital Geography Helps Account for Racial Disparities in Health Care, Economist Argues
African-Americans may get poorer medical treatment because they are more likely to go to bad hospitals, said Dartmouth economics professor Amitabh Chandra at a Nov. 19 talk sponsored by the Sadie Lewis Webb Program in Law and Biomedicine.
Chandra's assertion that geographic variations in care explain most racial disparities in treatment offers an alternate explanation to an Institute of Medicine report that placed more blame on physician-based discrimination. "Instead of pushing the physician-discrimination story, we're going to push a differences-in-hospital-quality story," he said. The IOM lambasted southern hospitals in particular, but Chandra charged that racial disparities are a "local phenomenon," with problems in northern hospitals as well. He based his arguments on a series of papers he and others authored on the topic.
"The key fact that comes from [the literature on race and health care] is that African-Americans, once you've controlled for all the clinical characteristics at the time of presentation, appear to get less-so it could be less of kidney dialysis, it could be less of renal transplantation, it could be less of bypass surgery," he explained.
Chandra studied the results of heart attacks in Medicare beneficiaries across the country and found wide variations in treatment depending on the hospital. Many measures that good hospitals take-providing beta blockers when a heart attack patient first arrives at the hospital, for example-are not necessarily expensive. Forty trials published in the Journal of the American Medical Associationtrumpet giving beta blockers to heart attack patients, yet many hospitals have not made the practice standard for all patients.
In charting hospital quality on a map of the United States, Chandra and co-author Katherine Baicker found that many low-quality hospitals were in fact in the South, but they were also in New Jersey and California. Top states for health care included New Hampshire, Maine, Vermont, North Dakota, Utah, Iowa, Wisconsin, and Colorado. Part of what drives quality, they found, was the composition of the physician mix-there were a high number of specialists in low-quality states, suggesting that physicians were not getting a good overall picture of patients' health.
Chandra also looked at the geography of hospital referral regions-where people actually go to get care, even if it's across state lines. While some have argued that there are simply dumb doctors in some regions, Chandra said one scholar who studied the five leading academic medical centers in Boston found significant variations in how they treat heart attacks and strokes, suggesting that regional differences are not as important to care as individual hospital differences.
From 1998-2001, 4,690 hospitals treated heart attacks; 1,000 hospitals account for almost 80 percent of all African-American patients, but only 35 percent of white patients.
"If I go after the first 40 hospitals, I can improve outcomes for 20 percent of blacks," he said.
About 2,000 hospitals didn't see a single African-American patient. Although Chandra was not allowed to tell the ranking of hospitals regarding heart attacks due to Medicare regulations, he is writing a paper that will name the 40 worst hospitals for treating African-Americans.
A model of the 30-day mortality rate shows that "mortality goes up for both blacks and whites" in hospitals that treat more African-American patients. Hospitals that saw no blacks were also low-quality centers. All-white hospitals tend to be located in Appalachia-eastern Kentucky, eastern Tennessee, and West Virginia. Forty percent of the white population is treated by small regional hospitals.
"The one thing we know, in terms of markers of quality, is that hospitals which have big volume are better," he said, perhaps because they attract better physicians or because they learn more from the volume of patients.
Overall, blacks are about 12 percent more likely to die 90 days after a heart attack, when adjusting for gender and age. Some doctors have told Chandra that African-American patients have more severe illnesses in general. "They are much sicker, I don't dispute that, but it's not driving the outcome," Chandra said.
That 12 percent falls to about 10 percent when hospital characteristics are accounted for, and just 5 percent when considering the actual hospital. "I can explain 60 percent of the racial disparity in care by controlling for the kind of hospital that you go to," he said. Thus physician discrimination may account for 40 percent of the white-black mortality gap, but 60 percent is due to geography. Whites treated in hospitals that care for a disproportionate number of African-Americans have a 30 percent lower survival rate after 90 days, while blacks treated at the same hospital have a 17 percent lower survival rate in comparison to those treated at top hospitals.
While Chandra's data on heart attack patients provides a clear picture because the outcome of bad treatment is premature death, racial disparities in regards to other medical problems are not as easy to assign to geography, as hospitals have strengths in different subspecialties. "There really are no clean, simple answers," Chandra said. He plans to write similar papers on hip fractures and strokes, as soon as the data is compiled, which may take three to four years.
He suggested that hospitals could help matters by focusing on the rates of procedures, such as heart catheterization, for whites and blacks. On average, there is a 9 percent gap between whites and blacks in receiving heart catheterizations, although catheterization rates as a whole vary regionally.
"The policy prescription isâ¦.you want to match the rate at which a procedure is being done to inform preferences for that procedure," he said.
Chandra said data suggests that Hispanics are also going to lousy hospitals and are likewise suffering from poor treatment, but Medicare's race coding for Hispanics is less reliable when you compare it to Census reports. "The race field in the Medicare data is populated by Social Security Administration data," entered in the 1940s and 1950s, when classifications for Hispanics were less likely to be used. "I think there's more work to be done on Hispanics."