This study by Hannan et al. examined racial differences in the utilization of cardiac angiography, coronary artery bypass graft, and coronary angioplasty for a sample population of patients with coronary artery disease hospitalized between January 1987 and June 1987 in New York State. Several methodological improvements over previous studies were described. First, in addition to using traditional proxies for disease severity, the investigators used the software package Disease Staging, which is characterized as an “objective measure of disease severity developed by a panel of physicians in terms of the probability of death or residual impairment.” Secondly, a longitudinal patient database was utilized to focus on patients, rather than admissions, as the unit of analysis. Since patients may initially visit a non-certified hospital for cardiac procedures and then be referred to a certified hospital, studies that use discharge or admission data double count these patients as two admissions and one procedure. As a result, potential bias is introduced to the study if referrals to certified hospitals differ by race.

Hannan et al. found statistically significant differences in the procedural treatment of black and white patients with coronary artery disease admitted to New York State hospitals. For the first admission during the study period, whites were 1.41 times as likely as blacks to receive angiographies, 3.74 times as likely to undergo coronary bypass, and 2.55 times as likely to receive angioplasties. When extended to include procedures done within four months of the first admission, the black-white odds ratios showed the same pattern. Whites were 1.42 times as likely to receive angiographies, 3.00 times as likely to receive coronary bypasses, and 2.39 times as likely to receive angioplasties. When, in addition to disease severity, the investigators controlled for several independent variables – age, sex, payer, number of secondary diagnosis, admission status, and estimated mean income of the zip code in which the patient resided – the statistically significant results persisted. Disease severity, however, proved to be one of the most significant variables. After controlling for disease severity the odds ratio were as follows: 1.25 for angiographies, 2.06 for bypasses, and 1.69 for angioplasties.

The authors consider several explanations for these results. The first is that white patients may be sicker than black patients and, therefore, have a greater need for the procedures. However, “blacks have a higher prevalence of cardiac risk factors, out of hospital cardiac fatality rates, mortalities for ischemic heart disease, and rates of coronary disease.” Another possible explanation offered is that whites undergo “more unnecessary surgery than blacks,” and the authors suggest, “if unnecessary procedures were eliminated from the set of cases considered, interracial differences would disappear or at least diminish.” The investigators propose using the Rand/UCLA appropriateness criteria to review medical records as an alternative measuring tool. The third potential explanation is that black patients “refuse recommended surgery more often than whites do.” The authors recommend two possible methods to further investigate this possibility. The first is to “review medical records retrospectively for evidence of recommended procedures that were refused” and, secondly, “to conduct a prospective study in which physicians are requested to document refusals.” The fourth explanation considered is that certain hospitals and physicians treat primarily white patients, and these hospitals and physicians may be more aggressive in recommending cardiac procedures. However, the authors found that there “is no clear tendency of hospitals with the highest percentages of whites to have higher procedure rates.” The correlation between percentage of white patients and procedure rate was not statistically significant. The final explanation considered is “that there is some bias toward providing procedures to whites more frequently than blacks.” The authors submit that this “is a difficult hypothesis to explore directly and possibly can be accepted only by disproving the alternative explanations.” In conclusion, they suggest that a study that “involves determination of appropriateness from medical record reviews as well as patient (and possibly surgeon) interviews” can possibly “detect preferences and biases” and therefore “begin exploring the reasons for the differences” observed. (This is precisely what the authors did in a later study. See Hannan et al, 1999 and Van Ryn and Burke, 2000 in this package.)