There is growing recognition that medical education must prepare physicians to deal competently and compassionately with patients from all backgrounds. Recent research highlights health disparities among more-privileged versus less-privileged population groups, along lines of race/ethnicity, sex and gender, socio-economic status, sexual orientation, and other factors.1 Cultural competence among health care providers is an important strategy for eliminating these disparities.2 In addition, new accreditation standards for medical schools from the Liaison Committee on Medical Education state that students and faculty should understand cultural diversity and that students should learn to recognize and address sex/gender and cultural bias in health care.3 Formal opportunities for learning about cultural diversity exist fairly widely. An AAMC survey reported that 86% of U.S. medical schools provide at least one opportunity in multicultural medical education.4
But in addition to the formal curriculum, a broader and subtler “hidden curriculum” (variously referred to as “unwritten,” “latent,” and “tacit”) is always at play in educational settings, consisting of knowledge, attitudes, behaviors, and skills that may be learned but not overtly intended by those who created a specific curriculum.5 In medical education, the hidden curriculum may consist of unintended messages communicated in lectures and other formal teaching; interpersonal interactions between faculty and students (both in clinical settings and in more informal aspects of classroom interactions); and the larger culture and structure of the medical establishment.6,7 This aspect of the curriculum may well be more influential than the structured, formal elements.6
In this study, we sought to characterize one important aspect of the hidden curriculum—the pattern of demographics, risk factors, and disease conditions presented in case examples. Case examples are a component of the curriculum that are planned by individual course directors and course instructors, typically independently, but received as a whole by the students. The cases, as a whole, can convey messages that are not explicitly planned. We examined all of the case examples available in written curriculum materials and course notes in a first- and second-year medical school curriculum, asking the questions: What is the representation of patient groups traditionally marginalized within the medical school curriculum? What are the relationships among demographic variables, behavioral factors, and disease processes that may perpetuate stereotypes in conflict with stated goals of medical education for cultural competence?
In our curriculum, students participate in several required sessions on culture and health in their first year, and have the opportunity to take a number of electives related to culture and health in years three and four. Part of the purpose of this study was to determine whether this aspect of our hidden curriculum (i.e., case examples) supports or detracts from our objectives for these formal curriculum components.
We analyzed 983 of the case examples that were presented in the 1996-97 first-year and 1997-98 second-year curriculum at the University of Minnesota Medical School; these constituted most of the case examples presented in those years. A case example was defined as the use of an individual to illustrate a normal or abnormal process or condition, for which at least two demographic or behavioral descriptors (e.g., age and sex) were included. Case examples were drawn from lectures, small groups, laboratory sessions, exam questions, assignments, class handouts and study guides, and clinical presentations involving guest or videotaped patients. For case examples given in lectures, we used detailed lecture notes from the student cooperative note-taking service (which were distributed to nearly all students) or lecture handouts from faculty.
Demographic and behavioral variables coded were sex and gender, sexual orientation, sexual behavior, number of sexual partners, homeless status, living arrangement, parental status, relationship status, age, race and ethnicity, place of origin, occupation, public assistance use, tobacco use, other chemical use, and HIV status. Two trained coders separately coded each case. Coders were instructed to record information exactly as presented in the case (for example, recording the exact ethnic or racial descriptor used, rather than fitting it into one of a number of arbitrary racial categories). Disagreements in coding were arbitrated by the authors. Inter-coder reliability was 98.1%. We also coded the diagnosis or presenting problem, course, body system, and document source for each case.
Three types of analysis were conducted. Frequency distributions were determined for major demographic variables. Rates of behavioral and diagnostic variables were then calculated in subpopulations of interest. Finally, a qualitative content analysis was conducted of the language used to describe sexual orientation, sexual behavior, and racial and ethnic identifiers.
In each case, we examined the presentation of information about the person or persons mentioned in the case according to three major demographic categories: sex and gender, sexual orientation, and race and ethnicity.
Sex and Gender
Of all 983 cases, 416 (42.3%) featured female patients, 526 (53.5%) featured males, and 12 (1.2%) featured intersexed patients (i.e., patients with conditions such as androgen insensitivity). Sex was unspecified in only 29 (3%) of the cases. No case featured an individual identified as being transgender. For the portion of the curriculum that is organized by organ system, representation of males and females by organ system segment was also examined (Figure 1). The differential representation of females and males varied across courses, from a nearly equal distribution (endocrine system courses, 79 female versus 78 male) to courses in which cases featuring males outnumbered those with females by nearly a factor of 3 (respiratory system courses, 13 female versus 42 male).
To ascertain whether this differential representation might be due to differences in disease prevalences between men and women, three cardiovascular diagnoses common in the cases as well as in the U.S. population were examined in more detail. For each of the diagnoses, representation of women in the curriculum's cases was compared with U.S. prevalence data from the National Hospital Discharge Survey.8 In the curriculum cases,
- 35% of those with the diagnosis of hypertension were female; in the U.S. population 59% were female;
- 33% of those with myocardial infarction were female, versus
- 40% in the U.S. population; and
- 14% of those with heart failure were female, versus 57% in the U.S. population.
Thus, for each of these three diseases, the percentage of curriculum cases featuring females was lower than the percentage of diagnosed individuals in the population identified as female.
Sexual orientation may be stated explicitly, or implied by reference to sexual behavior or relationship status. Since the latter two variables do not always correlate with self-identified orientation, and to avoid inferring the case-writers' intentions, these aspects were examined separately. Details of the representations of these three variables are presented in Table 1.
Examples of the language used to describe sexual orientation and/or behavior included
“Sexual preference: gay … multiple partners in last year”
“Gay male who engaged in unprotected sex with multiple partners”
“An avowed heterosexual … indicated he always used a condom”
“Heterosexual … permissive in the early 80's … but had no homosexual contact”
Denies homosexual exposure
This language was also examined qualitatively for recurring themes. First, in all of the five cases in which sexual orientation was specified, it appears to have been mentioned in the context of a risk assessment. Second, language used in describing heterosexual versus non-heterosexual identity or behavior seems to attach a relatively more positive value to heterosexuality and a relatively more negative value to homosexuality (e.g., “avowed heterosexual”). In addition, associations were found between sexual orientation and certain behavioral and/or disease factors: of the four cases featuring individuals identified as gay or men who have sex with men, three described the individuals as having multiple sexual partners (in the fourth this variable was unspecified). And of these four cases, three of the individuals were HIV-positive (unspecified in the fourth). While these numbers are not large enough for statistical analysis, they suggest important patterns.
Race and Ethnicity
Racial or ethnic identifiers were specified in only 45 (4.5%) of the cases. Among these, the racial identifier “white” was the most commonly used (26 cases), followed by “black” (four) and “Caucasian” (two). In the area of ethnic or national labels, persons of Swedish descent were mentioned in two cases; members classified under 11 other labels were mentioned in one case each.
Rates of tobacco use, other chemical use, homeless status, public assistance use, and HIV status were calculated for three subgroups: unspecified race or ethnicity (938 cases), “white” or “Caucasian” race (28 cases), and racial or ethnic identification other than “white” or “Caucasian” (17 cases). Rates for all variables were similar across subgroups.
A qualitative content analysis of the language in cases that specified race or ethnicity was conducted. In the language related to ethnic descriptions, several themes emerged. A number of health-related population variables were described in association with the person's ethnic identification. These variables included such factors as cultural practices (e.g., diet), population genetic links (e.g., heritable diseases), and geographic health hazards (e.g., origin in an area with less access to or use of biomedical health care)—see List 1. Among the cases in which a racial or ethnic identifier other than “white” or “Caucasian” was specified, more than two thirds (12/17) included such a variable. However, in the cases in which the racial identifier “white” or “Caucasian” was used, there was no such link (0/28).
The demographic information presented in case examples conveys messages, sometimes unintended ones. In this discussion, we examine the patterns of demographic information described above, and discuss the messages that may be conveyed. Last, we discuss the implications this study has for educational practice.
Sex and Gender
Our data show that women were underrepresented in case examples used in this preclinical curriculum. There was significant variation in sex/gender representation across courses. In at least one course, the disparities in such representation did not appear to be accounted for by differences in the prevalence rates of the diseases being discussed.
These results were consistent with content analyses of anatomy textbooks conducted by other investigators. For example, Lawrence and Bendixen9 reported that illustrations representing males have outnumbered those representing females consistently throughout the 20th century, as has the amount of text space devoted to male-specific descriptions versus that devoted to female-specific descriptions. These authors also found that the language used portrayed males as the norm, often defining females as exceptions to or variations on this norm.
There has been widespread agreement that undergraduate medical education has been lacking in representation of women's health issues.10,11 Efforts have been described that create women's health curricula and/or integrate women's health issues into preclinical and clinical courses.12,13 Objectives for content with regard to women's health have been outlined,10 and guidelines have been proposed for thorough review of women's health content in medical curricula.14 Our results suggest that any such review should include an analysis of sex and gender in case examples.
As stated earlier, we found that most case examples did not specify the sexual orientation of the patient being discussed. While sexual orientation may not be explicitly relevant to the teaching purpose of many cases, this lack of inclusion of gay, lesbian, bisexual, and transgender (GLBT) sexuality occurs in a societal context of “heteronormativity.”—15 Hetero-sexuality is often considered the norm—when no mention is made of sexuality, many people assume that the individual is heterosexual. This contributes to the invisibility of GLBT persons and stands in contrast to the widely held recommendation that teaching about GLBT issues be integrated throughout the curriculum.16
There has been a call for inclusion of GLBT issues in medical education. Recent research has shown that real and perceived discrimination and ignorance by health care providers and systems may have negative effects on these patients' access to and quality of care, and that these patients may have particular health care needs.17 A survey of GLB physicians found that 88% had heard medical colleagues disparage GLB patients, and 57% had observed colleagues provide diminished care or deny care to patients because of sexual orientation.18 Surveys of explicit medical school curricula have found limited teaching on this issue. In one study, such teaching took place mainly in lectures on human sexuality; several schools reported this issue's being addressed only or partly in the context of HIV/AIDS-related issues,16 a trend also noted by others.17 Some authors have documented bias among students or surveyed students for general perceptions of bias in the curriculum,19 but no research has examined components of the hidden curriculum with regard to this issue.
In our study, when we looked more closely at the language of the cases in which sexual identity or behavior was included, several themes emerged. One is that of valence, that is, having a relatively more positive value attached to heterosexual identity/behavior and a relatively more negative value attached to GLBT identity and/or same-sex behavior. This pattern may unwittingly feed into societal bias against GLBT persons, which some students may share.19 We also found that all of the cases that included sexual orientation did so in what appeared to be a risk assessment. Additionally, the trends in the cases were to associate gay identity or male same-sex behavior both with large numbers of sexual partners and with HIV-positive status. The latter finding is consistent with reports by other authors.17,18 These patterns are a source of concern related to medical education for a number of reasons. First, students may receive implicit messages that all or most GLBT-identified persons have multiple partners and/or are at risk for HIV/AIDS. And automatically associating GLBT persons with HIV/AIDS risk may lead students to neglect other health care needs of these patients, or to alienate patients already concerned about bias in the medical profession.
Most important, this pattern illustrates confusion over the concepts of identity versus behavior. It is well known that particular behaviors are risk factors for HIV infection; male same-sex sexual activity is one of these. But a man who has sex with men may identify as gay, or may think of himself as heterosexual. A gay-identified man may have one or many male and/or female sexual partners, may or may not engage in high-risk sexual behavior, or may not be sexually active. (Thus identity may be a marker for a risk factor, but is not an actual risk factor, and is an imperfect marker). Many of the cases in our sample discussed identity, but not behavior, as a risk factor for HIV infection, thus appearing to let identity stand in for the actual risk factor of behavior. Others used ambiguous language that leaves unclear whether it is identity or behavior being discussed (see the Results section for examples). The conflation of these distinct concepts may lead to misunderstanding among students about the important epidemiologic concepts of risk and markers for risk, and specifically about behavior and identity, particularly in the context of HIV infection.
Race and Ethnicity
We found that most cases did not indicate the race or ethnicity of the individual featured. This absence of racial and ethnic information occurs within a cultural environment in which whiteness is often assumed, yet goes unnamed,20 similar to the phenomenon of heteronormativity mentioned above.
While a small editorial literature addresses the concept of race as it is used in medicine and medical education,21 few data have been published to examine the actual patterns of use of racial and ethnic identifiers and their impact on student learning. Our literature search yielded only two studies of the use of racial and ethnic labels in case presentations. The one published study is a small but intriguing look at racial specification in residents' patient presentations. The authors found that overall race was included in the presentations more often when the patients were black, and that among the patients “who were presented as having potentially unfavorable characteristics,” race was presented more frequently for black patients than for white patients.22 The other study was an unpublished “in-house” study of several aspects of racial bias at a Canadian medical school. The medical student who led the study looked at the 87 case tutorials presented and found that “in some cases, the stereotypes that went along with the race were just overwhelming.”23
We found a number of health-related themes discussed in the majority of cases using ethnic or racial descriptors other than “white” or “Caucasian.” It seems reasonable to conclude that the use of an ethnic identifier was meant to illustrate something about the disease or health-related factor in these cases.
It is intriguing to note that no such theme was present when the racial identifiers “white” and “Caucasian” were given. One possible explanation for this is that the attribution of “white” or “Caucasian” is meant to signify the neutrality of the case subject, as in the fabled “70-kilogram white male.” In the context of the social sciences, it has been noted that whiteness is “simultaneously ignored and universalized: studies of members of the dominant race or culture, unless focused on racism per se, bracket the issues of race and culture and presume by implication the racial neutrality of the subjects of study.”24 Although we did not assess the effects of these patterns on medical students, it is possible that the patterns of ethnic and racial identification may reinforce the tendency to see white patients as the norm, while marginalizing others.
From our analysis of the treatment of sex/gender, sexual orientation, and race and ethnicity in case examples in a medical school curriculum, it is apparent that the overall pattern of demographics in the case examples conveys certain messages. The relative representations of different demographic groups in the sum total of cases are one lesson; associations of certain groups with particular diseases or risk factors are another. The lack of mention of sexual orientation and race or ethnicity in the overwhelming majority of cases carries meaning as well, with implicit reinforcement of the normalcy of whiteness and heterosexuality. Numerous authors have contributed to the deconstruction of the “70-kilogram white male” that for decades served as the misguided standard in medical research. Our results are suggestive of a similar standard in medical education, in which white, male, and heterosexual are placed in a central, normative position.
This hidden aspect of the curriculum is inconsistent with the formal curriculum, and indeed may undermine formal multicultural teaching. Discussion of the importance of women's health may be counteracted by the underrepresentation of women in case studies. Explicit teaching about gay, lesbian, and bisexual human rights and health issues may be contradicted by invisibility, value-laden terminology, and stereotypes perpetuated elsewhere in the curriculum. Subtle messages about the central position of whiteness may marginalize health issues of ethnic groups of color, weakening planned curricular emphasis on multicultural medicine.
These data suggest that we must go beyond the development of formal curriculum threads in multicultural education to a more transformational approach that actually shifts the lens of the curriculum as a whole. Peggy MacIntosh, in her work on curriculum transformation, identified several phases of curricula reform related to multicultural issues.25 In later phases of her model, MacIntosh calls for an examination of what populations are at the center of our curricula, and urges us to place into the center those individuals and communities that are at the margins.
This shift will require what may be for many of us a new level of curriculum management, in which faculty come together to examine the patterns in the cases presented in the curriculum, as well as any messages that those patterns may convey. Compiling case demographics within and across courses for review by faculty and curriculum leaders, and providing a forum for faculty to compare case demographics, are logical first steps. Reviews should focus first on comparing the epidemiology of cases in the curriculum with the actual epidemiology of the diseases being discussed, and balancing the selection of cases to approximate the actual patterns of disease. A second function of such forums should be to identify unintentional repetitions of similar case profiles (e.g., gay man with multiple sex partners) that may reinforce stereotypes. Finally, it would help students become prepared to address the full range of health concerns encountered by their patients. It would allow them to see a gay man dealing with diabetes, rather than with HIV, or an African American mother dealing with her child's otitis media, rather than with sickle-cell disease.
It may also be important to deliberate, as a faculty, the issue of when to include demographic information in formal case presentations. One possibility is to include standard demographic information in all cases. Naming a demographic characteristic from a dominant group, such as whiteness or heterosexuality, rather than assuming it, tends to remove that characteristic from its central, normative position.26 This practice would have an added benefit of raising faculty awareness of the patterns of case demographics.
If demographic information will not be presented in all cases, we suggest the development of standard decision rules to guide what information will be included. If we use this approach, it is important to develop a set of reasons when it is appropriate to include demographic information in the case. For example, it would seem appropriate to include demographic information if part of the purpose of the case is to teach students about culture and health, for example, meanings of illness in a particular community. Teaching about the effects of racism, homophobia, and other systems of oppression on health and health care access is another case in which the need for including demographic information is clear.
If the purpose is to teach students about a population link to health factors, as we presented in List 1, is it more appropriate to include demographic information, or to describe the actual health-related factor without including demographic information? If we describe the patient's diet as including a high intake of salted and smoked fish (as in the example presented in List 1), is it necessary or appropriate to include the patient's ethnicity (i.e., Japanese)? We must avoid allowing demographic information to stand in for actual risk factors, for example, race for socioeconomic status, or sexual orientation for sexual behavior.
If we take this approach, we must select the demographic information we present carefully. For example, in some cases, ethnicity may be a source of important “clinical clues” about a patient, including “geographic origin and migratory status, housing and employment patterns, dietary preferences, cultural and environmental factors, and genetic ancestry.”21 Race, however, conveys little information, and may be appropriate only when discussing the effects of racism on health and health care. It is worth noting that these clues regarding ethnicity must be derived from discussion with the patient. Contrast this with the imprecise way race is typically determined: by a quick glance at the patient.
If we take the approach of including demographic information only when we deem it relevant, we should realize that in some cases the use of demographic information in cases may reveal more about our own biases than about the concepts we are trying to illustrate.
This study has a number of limitations. One of these relates to the written materials from which the case examples were taken. Eighteen percent (18%) of the cases were drawn from detailed lecture notes from the student cooperative note-taking service (therefore distributed to nearly all students). The coders, who were all medical students, were asked to record any discrepancies between the notes and their recollections of the lectures. While no such discrepancy was recorded, it is possible that there were some inaccuracies in the cases contained in the notes, or some cases presented in lecture that were not recorded in the notes. In future research, faculty should review class notes to ensure their accuracy.
A second limitation concerns the small number of cases in which demographic or behavioral information was included (beyond, typically, age and sex). While this is an interesting point of discussion in itself, it limits the conclusions that can be drawn about associations made in the curriculum between demographic variables and particular diseases or risk factors. And finally, while case examples are a commonly used and important component of the curriculum, the patterns described here are only one element of the larger hidden curriculum; the other elements of this hidden curriculum were not examined.
The present study revealed the ways that race and ethnicity, sex and gender, and sexual orientation are presented in case examples in one undergraduate medical curriculum. The next logical question is how this affects students' learning. For example, if the demographic profile of the cases presenting heart failure in the curriculum differs significantly from the actual profile, how does this difference affect students' learning? When they have completed their course on cardiovascular disease, would students' understanding of the prevalence of heart failure among women be closer to the epidemiologic information presented as a part of the formal curriculum, or closer to the prevalence suggested by the actual demographic profile of cases?
Another question that is raised is how students interpret the absence of demographic information. When ethnicity and sexual orientation are not presented in a case, how does this affect how students view the patient featured in the case? In the absence of other identifying information, do students interpret patients to be a part of the dominant group? If so, how will this affect their understanding of the epidemiology and risk factors for the diseases presented in those cases?
Clearly, the manner in which a medical school curriculum represents traditionally marginalized groups in case examples conveys messages that may stand in contrast to formal teaching. We suggest that curriculum leaders examine this and other aspects of the hidden curriculum, and make deliberate choices about the messages conveyed.