Purpose: Confounding of racial, socioeconomic, and behavioral characteristics may lead to the misidentification of race as a risk factor for certain diseases. The authors evaluated the validity of attributing race as a risk factor in a widely used pathology textbook.
Method: In 2009–2010, the authors searched Robbins and Cotran Pathologic Basis of Disease, Professional Edition (8th ed) for assertions that African Americans have a different disease profile than do individuals of other races. They evaluated the references cited in the textbook, assessing the quality of the sources and whether those authors controlled for confounding variables, such as socioeconomic status and behavioral characteristics. For statements that remained unconfirmed, the authors searched the literature for evidence supporting or contradicting the claims made in the textbook. The authors classified each statement from the textbook as supported, unconfirmed, or contradicted.
Results: Of the 31 relevant statements, 11 were supported by the literature review, 17 remained unconfirmed, and three were contradicted.
Conclusions: About two-thirds of the assertions that different risk factors exist for African Americans found in a widely used pathology textbook could not be supported by the published literature. Failure to separate race from other socioeconomic and behavioral characteristics as risk factors for certain diseases may contribute to the misdiagnosis of patients and interfere with efforts to identify and remove causes of health disparities.
Mr. Sheets is a fourth-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Ms. Johnson is a fourth-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Mr. Todd is a fourth-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Mr. Perkins is a third-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Mr. Gu is a third-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Ms. Rau is a third-year medical student, University of Missouri–Columbia School of Medicine, Columbia, Missouri.
Correspondence should be addressed to Mr. Sheets, 707 E. Delmar Street, Springfield, MO 65807; telephone: (417) 860-1197; e-mail: LincolnSheets@gmail.com.
First published online August 24, 2011.
Epidemiology is an important part of medical education, and risk stratification by race can inform a clinician's differential diagnosis. However, when medical textbooks identify race as a risk factor for certain diseases, that information may be misleading if the authors confounded race with income, education, or behavior. Few studies have examined how well the authors of medical textbooks control for these potential confounding variables.
In the United States, race is often linked with socioeconomic status (SES); for example, 24.2% of African Americans lived below the poverty line in 2006 compared with only 8.2% of non-Hispanic Caucasians.1 For middle-class African Americans, it is likely that the risk factors for certain diseases resemble more closely the risk factors for middle-class Americans of other races than those for impoverished African Americans. Are the putative associations between race and certain diseases, then, actually nonexistent? To test the hypothesis that authoritative sources in medical education designate race as a risk factor for certain diseases without adequately considering the potential confounding by SES and certain behaviors, we examined a widely used pathology textbook, subjecting the authors' assertions regarding race as a risk factor to reviews of the pertinent literature.
In 2009–2010, we searched a digital copy of Robbins and Cotran Pathologic Basis of Disease, Professional Edition (8th ed)2 for any use of the words “African,” “black,” “dark skin,” “dark-skinned,” “ethnic,” “race,” or “racial” that might indicate that African Americans have a different disease profile than do individuals of other races. Counting multiple uses of these words within one topic as a single statement, our search identified 94 statements in which at least one of these words or phrases was used. We then excluded 36 statements that used the word “black” in reference to color (e.g., “Melanin, derived from the Greek [melas, black], is an endogenous, nonhemoglobin-derived, brown-black pigment”2(p36)), 15 statements that used the word “African” as a geographic term (e.g., “Only rare homozygotes for the mutation [in HIV] have been found in African or East Asian populations”2(p240)), and one statement that asserted an equality in disease profiles by race rather than a difference (“Blacks and whites are equally affected [by myocardial infarcts]”2(p547)). By excluding these 52 statements from the 94 total search results, we were left with 42 statements that asserted that African Americans have a different disease profile, either in prevalence or severity, than do individuals of other races (almost always non-Hispanic Caucasians). We next excluded 11 statements about diseases with a demonstrated genetic etiology (e.g., “Studies of whites and blacks suggest linkage of the primary form [of Sjogren syndrome]”2(p221)). We were left with 31 relevant statements, for which we noted the sources referenced by the authors of the textbook.
We examined each source, recording whether or not the potential confounding by SES and certain behaviors was controlled for in the analysis. We designated income, education, and health insurance status as variables indicating SES and considered consumption of alcohol and tobacco as well as diet as behavioral factors. We interpreted studies demonstrating different genetic profiles between ethnic groups as having controlled for confounding variables. We scored all the sources for overall quality using the 22 STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) criteria3 and excluded any sources that failed to meet five or more of these criteria. For every statement about the effects of race as a risk factor that was neither supported nor contradicted by one or more published sources that controlled for confounding variables, we searched DynaMed4 for any relevant evidence to support the statement's claims about the effects of race on a disease profile and evaluated those studies in the same way that we evaluated the sources cited in the textbook.
Medical students, who had received three hours of training, served as raters. Two raters, who changed for each statement, independently analyzed each relevant Robbins and Cotran statement that asserted that African Americans have a different disease profile than do individuals of other races; in cases where the two coders did not agree about the risk factor being supported, the risk factor being confounded, or the quality of the studies being used to validate the statement, a third independent rater functioned as a tie-breaker. We counted a statement from the textbook as supported if we identified at least one agreeing published study that controlled for confounding variables. Similarly, we counted a statement as contradicted if we identified at least one contradictory published study. We identified no statements that fell into both categories. See Figure 1 for our search and assessment process.
The latest edition of the widely used Robbins and Cotran pathology textbook asserts that 24 separate diseases are more prevalent or severe in African Americans than in non-Hispanic Caucasians, and 7 are less so. Of these 31 assertions that race is a risk factor, we confirmed 11 to be true because they were supported by published research that controlled for SES and behavioral risk factors. For 3 diseases (squamous cell carcinoma of the esophagus, cholangiocarcinoma, and malignant hypertension), our literature review found evidence to the contrary of the claims in the textbook—The racial disparity stated could be explained by SES or behavioral risk factors. The 17 remaining statements were supported only by studies that failed to control for possible confounding by SES or behavioral characteristics and thus remain unconfirmed according to our research methodology. See Table 1 for these results. However, we have excluded 1 of the 17 unconfirmed statements, regarding the pathologic aspects of repair (keloid formation), from Table 1. Neither the source referenced by the Robbins and Cotran textbook nor the sources that we discovered in DynaMed controlled for SES or behavioral risk factors, and so we were unable to confirm that race is a risk factor, yet we are aware of a significant body of literature that does cite race as a risk factor for keloid formation. To prevent confusion about this one disease, we have chosen to omit it from Table 1. For 9 of the 31 statements, the first two raters did not arrive at the same results, requiring a third independent rater to break the tie.
Our finding that approximately two-thirds of the statements affirming that race plays a role in the differences in disease prevalence or severity found in a widely used pathology textbook are not supported by published evidence in the literature has implications for medical education. Accurate information about the epidemiology of specific diseases is important for the understanding of pathogenesis and etiology, the delivery of patient care, and the improvement of public health. Our findings suggest that other resources commonly used in medical education misidentify race as a risk factor for certain diseases because of a failure to adequately account for confounding by SES or behavioral characteristics.
Although we found no similar previously published research, our results are consistent with the findings from other studies that have attributed racial discrepancies in disease prevalence to differences in SES or other confounding variables.5,6 The widespread inclusion of race as a patient characteristic on rounds7 and in clinical case presentations8 suggests that knowledge of a patient's race affects physicians' assessments. Race is not unique in this regard, however, and similar studies of other purported risk factors could produce comparable results.
Assessing whether a study supported a statement from the textbook, determining whether a study adequately controlled for confounding variables, and evaluating the quality of a study were all susceptible to subjectivity. We addressed potential inconsistencies by relying on two raters and enlisting a third when a disagreement occurred. The need for a third rater in 9 of 31 assessments reflects the subjective nature of this type of methodology. In contrast, by using the STROBE criteria as a tool for objectively judging observational epidemiological research,9 there were no instances of disagreement between the two raters in evaluating the quality of the studies.
In addition, our analysis was limited to one textbook. Similar examinations of other widely used educational resources are needed to determine if the potential mislabeling that we found is an isolated phenomenon or if it is representative of a spectrum of educational materials. We excluded eight diseases with demonstrated genetic etiology on the assumption that they had little potential for confounding by SES or behavioral characteristics. This exclusion has made the ratio of unsupported to supported statements from the textbook greater than it otherwise would have been.
Finally, it is possible that our literature review missed evidence that supported some of the statements that we concluded to be unsupported. We used DynaMed for our literature review because of its concise and highly regarded summations of the best available evidence about specific diseases, including their epidemiology.10 Failure to control for confounding variables precludes the definitive identification of race as a risk factor for certain diseases; however, it does not discount it entirely.
Recently, considerable attention has focused on racial disparities in health care and health status.11–13 In clinical practice, the erroneous association of race with certain diseases could contribute to the misdiagnosis of patients with those diseases. Failure to disentangle race from related socioeconomic and behavioral characteristics may interfere with efforts to identify and remove causes of health disparities. We hope that our study findings will serve as an alert to medical textbook authors and will motivate them to evaluate more closely what their textbooks say about race as a risk factor for certain diseases. We also hope that other investigators will build on our findings to more fully explore the issues that we have raised in this report.
The authors wish to thank Daniel C. Vinson, MD, and Robert L. Blake, Jr, MD, of the Department of Family and Community Medicine at the University of Missouri–Columbia School of Medicine for their generous guidance, encouragement, and critical review; medical students Lane Wilson, Emily Hoffman, Brian Hilliard, Alexandra Weir, Lindsey Martin, Marianne Lopez, Anjali Patel, and Emily Griffard of the University of Missouri–Columbia School of Medicine for their careful assistance with data collection; Dr. Jane McElroy and Dr. Robin Kruse of the University of Missouri–Columbia School of Medicine for their careful and creative assistance in presenting results; and Rebecca Graves, Educational Services Librarian at the University of Missouri–Columbia Health Sciences Library, for her patient and professional training in expert search techniques.
The authors previously presented some of this material in poster form at the University of Missouri–Columbia Health Sciences Research Day on November 12, 2009.
1 U.S. Census Bureau. Table B-1. Poverty status of people by family relationship, race, and Hispanic origin: 1959 to 2006. In: Income, Poverty, and Health Insurance Coverage in the United States: 2006. http://www.census.gov/prod/2007pubs/p60-233.pdf
. Accessed May 5, 2011.
2 Kumar V, Abbas AK, Fausto N, Aster JC. Robbins and Cotran: Pathologic Basis of Disease, Professional Edition. 8th ed. Philadelphia, Pa: Saunders Elsevier; 2009.
3 von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. Ann Intern Med. 2007;147:573–577.
5 Subramanian SV, Kawachi I. The association between state income inequality and worse health is not confounded by race. Int J Epidemiol. 2003;32:1022–1028.
6 Kaufman JS, Cooper RS, McGee DL. Socioeconomic status and health in blacks and whites: The problem of residual confounding and the resiliency of race. Epidemiology. 1997;8:621–628.
7 Nawaz H, Brett AS. Mentioning race at the beginning of clinical case presentations: A survey of US medical schools. Med Educ. 2009;43:145–154.
9 Manchikanti L, Singh V, Smith HS, Hirsch JA. Evidence-based medicine, systematic reviews, and guidelines in interventional pain management: Part 4: Observational studies. Pain Physician. 2009;12:73–108.
10 Alper BS, White DS, Ge B. Physicians answer more clinical questions and change clinical decisions more often with synthesized evidence: A randomized trial in primary care. Ann Fam Med. 2005;3:507–513.
11 Flores G, Tomany-Korman SC. Racial and ethnic disparities in medical and dental health, access to care, and use of services in US children. Pediatrics. 2008;121:e286–e298.
12 LeCook B, McGuire TG, Zuvekas SH. Measuring trends in racial/ethnic health care disparities. Med Care Res Rev. 2009;66:23–48.
13 Bagchi AD, Schone E, Higgins P, Granger E, Casscells SW, Croghan T. Racial and ethnic health disparities in TRICARE. J Natl Med Assoc. 2009;101:663–670.