Is research into ethnicity and health racist, unsound, or important science?BMJ 1997; 314 doi: https://doi.org/10.1136/bmj.314.7096.1751 (Published 14 June 1997) Cite this as: BMJ 1997;314:1751
- Raj Bhopal, professor and head ()a
- a Department of Epidemiology and Public Health, University of Newcastle, Newcastle upon Tyne NE2 4HH
- Correspondence to: Professor Bhopal Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC 27599 7400, USA (until 10 June 1997).
- Accepted 18 December 1996
Much historical research on race, intelligence, and health was racist, unethical, and ineffective. The concepts of race and ethnicity are difficult to define but continue to be applied to the study of the health of immigrant and ethnic minority groups in the hope of advancing understanding of causes of disease. While a morass of associations has been generated, race and ethnicity in health research have seldom given fundamental new understanding of disease. Most such research is “black box epidemiology.” Researchers have not overcome the many conceptual and technical problems of research into ethnicity and health. By emphasising the negative aspects of the health of ethnic minority groups, research may have damaged their social standing and deflected attention from their health priorities. Unless researchers recognise the difficulties with research into ethnicity and health and correct its weaknesses, 20th century research in this subject may suffer the same ignominious fate as that of race science in the 19th century.
Epidemiology aids health policy and planning and helps discover the laws governing health and disease. As with other sciences,1 2 3 epidemiology has been beguiled by ethnicity and race4 5 and has become racialised. Racialisation consists of the idea that race is a primary, natural, and neutral means of grouping humans and that racial groups are distinct in other ways, such as their behaviours.6 Racialism is the belief in the superiority of some races. In this paper I draw lessons from the racialised research of the 19th century, discuss the terms race and ethnicity, and analyse the value of and problems with research into ethnicity and health.
Research on race: a historical look
Racialised research has an inglorious history: scientists have been besotted by race and ethnicity, while politicians and social commentators have encouraged them.1 2 3 4 5 6 7 8 9 10 11 In the 19th century scientists ranked races on their biological and social worth, particularly using measurements of the size and shape of the head and the contents of the brain to measure intelligence (northern European groups always ranked top).7 Such research was used to justify slavery, imperialism, anti-immigration policy, and the social status quo.1 6 7 8 9 10 11 12 One underlying value of this research was that biology determined social position–that is, biological determinism. The power behind scientific racism is shown by the prowess of some of the researchers, who included Louis Agassiz, Francis Galton, Paul Broca, and John Down (see Gould7 for details of their contributions).
Medical practitioners contributed to racialised science. “Diseases” such as drapetomania (irrational and pathological desire of slaves to run away) and dysaethesia Aethiopica (rascality) were invented.6 To quote a textbook, “the pelves becomes increasingly lower and broader the more civilised the race from which it is obtained,” and, “coloured children weigh considerably less than white, a fact which, in large cities at least, is indicative of the physical degeneration which characterises the race.”13 The importance of race research and the innate inequality of races was considered self evident, and few scientists questioned whether their work was ethical.1 2 7 11
Current views on race and ethnicity
Humans are one species: races are not biologically distinct, there is little variation in genetic composition between geographically separated groups, and the physical characteristics distinguishing races result from a small number of genes that do not relate closely to either behaviours or disease.2 Massive effort over 150 years to classify races has largely failed, though we use crude classifications which trace their heritage to Linnaeus, based on the division of populations as Homo Afer (synonyms, black, Negro, Negroid), Homo Europaeus (synonyms, white, Caucasian, Caucasoid), Homo Asiaticus (Mongoloid), and Homo Americanus (American Indian). Variants of these classifications also have a grouping for Australian aborigines.2 3 12 14 Most complex classification has been forgotten2.
Haddon and Huxley recommended that the race be replaced by ethnic type,1 an idea enjoying much support4 5 15 16 and some criticism.17 None the less, race remains important in modern thinking, though increasingly it reflects geographical, social, and class divisions rather than biological ones.16 18 19 The term race is often used alongside ethnicity.4 6 While arguing for abandoning race, Huth did not see problems with ethnic identification.15
The taboo surrounding research into race,1 2 11 greater understanding of social and cultural factors in health and disease,4 5 6 15 17 19 20 and the need to describe the health and health care of people from ethnic minorities created the spur for new terminology,16 17 and ethnicity is at the fore. In the context of health it means a group that people belong to because of shared characteristics, including ancestral and geographical origins, cultural traditions, and languages.5 6 15 17 19 Ethnicity is a complex idea that has become a euphemism for race, and writers have not separated the concepts clearly.5 21 22 For example, a paper by Hopkinson constructed around race uses the ethnic groups as classified by the census.22 Inability to use a clear definition of ethnicity echoes the past, when a consensus on the definition of race could not be achieved but was too important an idea to discard.1
Ethnicity is a fluid concept and depends on context. For practical and theoretical reasons, the current preference is for self assessment of ethnicity.17 19 People change their self assessment over time, as is their prerogative. The alternatives include skin colour, birthplace, ancestry, names, geographical origins, or a mixture of these. Ethnicity is not measurable with accuracy or validity.17 The question on ethnicity in the 1991 census worked only in that people were willing to answer it, and the classification was arbitrary.
Research into ethnicity and health
Expectations of researchers
Scientists want to discover the causes and processes of disease, while health policy makers and planners want to meet the needs of ethnic minority groups. Historical analysis reveals motives such as a wish to reverse the health and social disadvantages of ethnic minority groups, curiosity about racial and ethnic variation, and an interest in ranking races and ethnic groups.
Studies of migrant groups help to separate the effects of environmental and genetic factors.4 5 20 23 24 Leaving aside problems of bias and the difficulties of making comparable measurements across long distances, studies of migration could be a powerful means of generating and testing hypotheses. When both migrants and their offspring are compared with other ethnic groups the design is enriched. Changing circumstances within and between generations in different migrant and ethnic groups can be linked to changing health.
The message from most publications on ethnicity and health is that this opportunity must not be missed.4 5 20 23 24 In Biocultural Aspects of Disease Henry Rothschild offered ethnicity as a paradigm for understanding diseases of complex aetiology.20 Marmot and colleagues' report Immigrant Mortality in England and Wales opens with the statement: “Studies of mortality of immigrants are useful for pointing to particular disease problems of immigrants, investigating aetiology and validating international differences in disease.”23
Black box epidemiology
Does such research discover aetiology? Thousands of associations between racial and ethnic groups and disease have been published with the promise that they will help in elucidating aetiology. The data are usually published in the style of aetiological epidemiology to show relative frequency of disease by means of standardised mortality ratios or similar measures (see Senior and Bhopal17 for a fuller discussion). Few variations have been explained in a way that gives new insight into aetiology.4 5 19 23
Most ethnicity and health research is “black box” epidemiology–what Skrabanek described as epidemiology where the causal mechanism behind an association remains unknown and hidden (“black”) but the inference is that the causal mechanism is within the association (“box”).25 Skrabanek argued that science must open and understand the black box. He cited a review of 35 case-control studies of coffee drinking and bladder cancer which failed to provide important information and likened such epidemiology to repeatedly punching a soft pillow. David Savitz defended black box epidemiology, particularly for exploring new subjects, arguing that epidemiology may not be needed when other sciences have elucidated causal paths.26
Many studies have investigated patterns of cancer in immigrant, racial, and ethnic minority populations.4 23 24 27 28 29 30 31 Marmot and colleagues' analysis of cancers in immigrants in England and Wales found many differences, but, overall, immigrants had lower cancer rates.23 The researchers' aetiological focus is illustrated by their emphasis on causal hypotheses, of which many of interest were developed. They noted that international data for cancers of the large intestine and female breast showed high correlations with heart disease and fat consumption. Their observation of low rates of these two cancers in Indian immigrants but high rates of heart disease led them to question the assumption that dietary fat was the common factor in cancer of the large bowel and breast, and they queried whether the high fibre content of the Indian diet modified the effect of fat on large bowel cancer.
Balarajan and colleagues' study of immigrant populations by region of origin also found many differences from which they developed aetiological hypotheses, and they urged that data on ethnicity and health be used to develop more.24 Donaldson and Clayton found numerous ethnic differences in patterns of cancer registration in Leicestershire health district.28 The authors rightly concluded: “The results indicate the need for formal epidemiological study to test specific aetiological hypotheses which may account for these apparent differences.” This type of work has been repeated–for example, by Barker and Baker in Bradford,29 by Matheson et al in Scotland,30 and by Balarajan and Bulusu.3 Similar work has been done on children. The conclusion is almost invariate– differences exist and need detailed study.27
However, there has been little progression beyond this black box epidemiology, since few studies have explored the ideas generated.27 One exception is the study of diabetes and insulin resistance in south Asian communities as the possible basis of their surprisingly high rates of coronary heart disease. Marmot et al observed that “The high rate of diabetes could contribute to the high rate of ischaemic heart disease in Indians. This explanation would then pose the problem of why immigrants from the Caribbean, with their high rate of diabetes, do not also have a high rate of ischaemic heart disease.”23 This question is being pursued tenaciously.32
We need to move from the repetitious demonstration of disease variations that have already been shown in research into ethnicity and health or in work on international variations or in social and sex variations –that is, stop punching the pillow25–and move to new territory.26 Studies of ethnicity and health should be able to provide models and contexts for advancing aetiological knowledge if questions for research are clearly articulated and pursued with sound methods.
Is such research unsound epidemiology?
Much research into ethnicity and health is unsound.1 The key variables of ethnicity and race are vaguely defined, and the underlying concepts are poorly understood and hard to measure.17 There is inconsistent use of terminology: for example, Asian, white, Caucasian, and Hispanic are common terms in research but have inconsistent and non-specific meanings.33 There are difficulties in collecting comparable data across cultural groups: for example, do questions on stress or alcohol consumption have equivalence across cultures? There are problems in recruiting representative and comparable population samples.
Data need to be adjusted for known confounding variables and interpreted with the recognition that adjustment is probably incomplete. These issues have been detailed elsewhere.17 34 Rigour is needed for sound epidemiology in ethnicity and health, but the literature is littered with elementary errors (see box).
Basic errors in epidemiological studies of ethnicity
Inventing ethnic groups–A study labelled a group as Urdus on the basis of the language spoken, thus inventing an ethnic group35
Not comparing like with like–Inner city populations are different from whole population samples,36 but studies of ethnicity and health continue to focus on them for convenience–as in the recent Health Education Authority survey, in which the comparison population was not an inner city sample37
Lumping groups together–A paper on smoking and drinking habits in British residents born in the Indian subcontinent did not describe sex and regional variations, creating the impression that smoking and drinking were unimportant in the “Asian” population.38 As has been shown,37 and long known by people knowledgeable about populations of Indian origin, smoking and drinking are important problems in some subgroups. Heterogeneity in the prevalence of disease and risk factors has even been shown among different Hindu castes in one city in Tanzania.39 Yet journals still publish comparisons as crude as white and non-white.19 The British attitude before 1940 was to blur the racial specificity of colonial populations11
Not adjusting for confounding factors–Inferences can change radically once interacting and confounding factors are accounted for: Lillie-Blanton et al challenged the observation that crack smoking was commoner in African Americans and Hispanic Americans and showed that once social and cultural factors were accounted for there were no differences40
There is little evidence that criticism of the methods and concepts of research into ethnicity and health17 19 34 41 42 43 44 45 has paid dividends. For example, while Marmot and colleagues' analysis of mortality in immigrants attempted to analyse ethnicity because country of birth was too crude,23 an update using mortality data for l980-2 did not even though there were then far more British born people in ethnic minority groups.24
While methodological errors may be apparent, it is more difficult to judge whether the research questions are valuable and whether the conceptual basis of the research (largely comparative) is sound.
Harm from such research
Osborne46 answered yes to the disturbing question of whether race based research in medicine is racist. His review cites projects that focused on differences between blacks and whites in diseases associated with promiscuity, underachievement, and antisocial behaviour and which implied that the underlying explanation lay in race rather than class, lifestyle, or socioeconomic status.
Perceiving ethnic minorities as unhealthy–The perception that the health of ethnic minority groups is poor47 can augment the belief that immigrants and ethnic minorities are a burden. The perception is at least partially false for some migrant groups, especially men, as shown in table 1.23 24 There are variations by cause of disease, but overall standardised mortality ratios hover around the average for England and Wales. Bearing in mind inaccuracy in the denominator, the fact that those born in Britain have not usually been included, and that some deaths and illnesses are among visitors rather than residents, it is not clear whether the true rates are higher in most ethnic minority groups. The perception of poorer health arises from a focus on differences where the excess of disease is in the ethnic minority population.17 47 For many causes, morbidity and mortality are lower.
The focus on a few “ethnic” problems (such as high birth rates, “Asian rickets,” the haemoglobinopathies, and congenital defects said to be linked to consanguinity) has been at the expense of major problems.17 48 Health education material for ethnic minority groups in the 1980s tackled birth control, lice, child care, and spitting, but there was nothing on heart disease and little on smoking and alcohol.48 The idea of a package of specific “ethnic” diseases has echoes in history: Negro susceptibility to particular diseases such as leprosy, tetanus, pneumonia, scurvy, and sore eyes was instrumental in “branding blacks as an exotic breed,” and the differences were explained by nonsensical hypotheses on causation.49
The comparative approach–Most research into ethnicity and health (including mine) is based on the comparative paradigm and presents data using the “white” population as the standard.17 Inevitably, attention is focused on diseases that are commoner in ethnic minority groups than in the white population, thereby displacing problems like cancer and respiratory disease that are very common but less so than in the white population from their rightful place as high priorities for ethnic minority groups. A bibliography by Karmi and McKeigue stated: “Although cancer is one of the key areas specified in the Health of the Nation white paper, it is not especially relevant to ethnic groups in Britain.”50 This shows the danger of the comparative approach. Cancers are a major cause of death and disability in ethnic minority groups, and there is an opportunity to prevent some cancers reaching the high levels seen in the general population.27
Ignoring quality of services–The implications of comparative research, including the risk of ethnocentrism, is discussed in more detail elsewhere,17 and a strategy for setting priorities for ethnic minority groups is forthcoming.51 The misperception that the needs of ethnic minorities are so different from those of the majority that separate strategies are necessary (but which may not materialise) provides a rationale for national strategy to exclude consideration of ethnic minority groups.52 The promise of aetiological understanding has meant a focus on variation in diseases, as opposed to the quality of services. There is a huge gap in the research record on the quality of care received by ethnic minority groups.19
Fuelling racial prejudice–Finally, racial prejudice is fuelled by research portraying ethnic minorities as inferior to the majority. Infectious diseases, population growth, and culture are common foci for publicity. Following the release of statistics on the ethnicity of single mothers, the Sunday Express of 13 August 1996 had the headline “The ethnic time bomb”. Toni Morrison wrote that “A whip of fear broke through the heart chambers as soon as you saw a Negro's face in a paper,” for this signalled exceptionally bad news.53 Researchers cannot be responsible for media reporting but must be aware of the potential impact of their work on race relations.
With hindsight, we can see that much race oriented science in the past was unethical, invalid, racist, and inhumane though it was perceived to be of great importance.1 2 11 The Bell Curve is a reminder that research which purports to demonstrate the innate inferiority of some racial groups continues and that race science is alive.54 Researchers need to understand how research into race and health was misused in the past. Epidemiologists should remember that warnings from disciplines incorporating anthropology and psychology may be based on harsh experience, for these disciplines played a leading part in racialising science.1 7 11 Epidemiologists who remain unpersuaded that racial prejudice could influence science should read about the Tuskegee syphilis study, which examined the natural course of syphilis in 600 poor “negroes” in Alabama, denying them effective treatments and hastening many deaths.55
Knowledge of the interplay of cultural, genetic, and environmental factors is valuable, and research into race and ethnicity is one way to achieve it. Contemporary researchers also justify such research as necessary to help meet the needs of ethnic minority groups and point out that lack of data can hinder health policy.19 56 Inequalities in the health status of ethnic minority groups demand attention.15 21 23 24 31 For these reasons, scientists' interest in the relation between race, ethnicity, and health will increase.
Participation by ethnic minorities in research, policy making, and the development of services might be one safeguard against repeating the mistakes of the past. The American College of Epidemiology has called for a greater contribution to epidemiology by researchers from ethnic minority groups, who are underrepresented.21 57 When Jews and black people spoke on race, however, their views were interpreted as representing special interests.11 A partnership between scientists from ethnic minority and ethnic majority groups is needed.
My view that the potential for testing hypotheses is rarely realised and that the aetiological value of such research has been exaggerated17 34 52 remains controversial. Wider and constructive debate on mounting criticisms is essential as a step towards agreement on the way forward. This debate is more advanced in the United States than in Europe, but on both sides of the Atlantic writings intended to stimulate change6 19 41 42 43 44 45 46 have had surprisingly little impact.
With regard to the question in this paper's title, I believe that few people researching ethnicity and health are racist and that most hold humanitarian views. Many, however, are working to a racialised research agenda (my own work is no exception). If our work is racist as well as racialised most of it is unwittingly so, but that probably applies to much historical work. In 100 years' time will our successors judge our research to be racist–that is, bolstering the case for the innate superiority of some racial groups? Applying this to ethnicity, might our work be seen as “ethnicised” research and the fuel of “ethnicism”? Millions of people had their skulls measured by craniologists to no benefit.1 Let us ensure that late 20th century research does not suffer a similar fate.
Most research on race, ethnicity, and migration has been black box epidemiology, which has potential in planning health care but needs to be presented and interpreted with a different perspective from that of aetiological research.17 A great deal of research on race, ethnicity, and migration is unsound because the questions posed were not relevant or answerable or because the methods were not adequate.
My message is not of despair but a call for corrective action, of the kind so ably documented by many scholars,6 19 41 42 43 44 45 46 but which has been deftly evaded by researchers and editors alike. The research paradigm should be adjusted away from ethnicity and race as the key to unlocking the secrets of the causes of disease20 to being a tool for assessing needs and inequality and guiding practical action.
Senior and I made nine recommendations to help make ethnicity a sound epidemiological variable (see box).17 To these I would add (or re-emphasise) the following:
In the absence of consensus on the nature of ethnicity and race, researchers must state their understanding, describe the characteristics of both the study and comparison populations, and provide and justify the ethnic coding17 58 59
There should be wide recognition that, like data on social class, information on race and ethnicity has a key role in raising awareness of inequalities and stimulating policy and action.19 56 57 61 62
Summary of recommendations to improve the value of ethnicity as an epidemiological variable (from Senior and Bhopal17)
Ethnicity should be perceived as different from race and not as a synonym for it
Ethnicity's complex and fluid nature should be appreciated
The limitations of methods of classifying ethnic groups should be recognised, and reports should state explicitly how such classifications were made
Investigators should recognise the potential influence of their personal values, including ethnocentricity
Socioeconomic differences should be considered as an explanation of differences in health between ethnic groups
Research on methods for ethnic classification should be given higher priority
Ethnicity's fluid and dynamic nature means that results should not be generalised except with great caution
Results should be applied to the planning of health services
Observations of variations in disease should be followed by detailed examination of the relative importance of environmental, lifestyle, cultural, and genetic influences
I thank the many authors cited here, but particularly Stephen J Gould, for inspiration and insight; Lorna Hutchinson and Alison Etherington for help in preparing the manuscript; Drs B Charlton, R Thomson, B Elliott, W Chin Leung, and Professor L Donaldson for critical comments. An anonymous referee and editorial staff of the BMJ provided extensive and important critical comment. The views expressed are entirely my responsibility, though they have evidently been shaped by other scholars.