- Email: [email protected]
The Impact of Different Measures of Socioeconomic Position on the Relationship Between Ethnicity and Health
The Impact of Different Measures of Socioeconomic Position on the Relationship Between Ethnicity and Health MARGARET KELAHER, SHEILA PAUL, HELEN LAMBERT, WAQAR AHMAD, AND GEORGE DAVEY SMITH
PURPOSE: In this study we explicitly examine the effect of using asset-based and conventional measures of socioeconomic position (SEP) on ethnic differences in health. METHODS: The survey of white (n Z 227), Indian and Pakistani (n Z 233), and African Caribbean (n Z 213) adults aged 18 to 59 years living in Leeds, United Kingdom aimed to examine the relationship between ethnicity, health, and SEP. SEP variables included perceived ability to get £10,000 if needed, car ownership, level of education, and home ownership. Health variables included self-reported health status, presence of a long-term illness or disability, presence of limitations arising from a long-term illness or disability, one or more limitations in mobility, obesity, and being anxious, worried, or depressed. Logistic regression analysis was used to assess the relationship between ethnicity and SEP and health. Five models were run for each health variable so that the effects of changing the SEP measure could be ascertained. The first model included only ethnicity and the remaining 4 models tested the effects of the perceived ability to get £10,000, car ownership, level of education, and home ownership separately. RESULTS: The results suggest that the statistical inclusion of asset-based SEP measures, such as car ownership and ability to obtain £10,000, which reflect an individual’s ability to mobilize resources, tend to increase ethnic differences in health, whereas more conventional steady-state indicators, such as education level and home ownership, tended to have little effect or to reduce ethnic differences in health. CONCLUSIONS: Overall, this study suggests that the choice of SEP measure may affect the conclusion of research on ethnicity and health and that choice of SEP measures should in turn be informed by the research problem being examined. Ann Epidemiol 2008;18:351–356. Ó 2008 Elsevier Inc. All rights reserved. KEY WORDS:
Ethnicity, Socioeconomic Measures, Health.
INTRODUCTION There is a vast body of research linking socioeconomic position (SEP) to health (1–10). However, there is increasing evidence that not all indicators of SEP are interchangeable (11). A number of studies report variation in the apparent relationship between SEP and health dependent on the type of measures used (11–22). This has resulted in a call
From the Centre for Health Policy, Programs and Economics, School of Population Health, University of Melbourne, Carlton, Victoria, Australia (M.K.); UCL Centre for International Health and Development, Institute of Child Health (S.P) and Social Policy Research Centre, Middlesex University (W.A.), London, UK; Department of Social Medicine, University of Bristol (H.L, G.D.S.), UK. Address correspondence to: Dr Margaret Kelaher, School of Population Health, University of Melbourne, 207 Bouverie St, Carlton Vic 3010, Australia. Tel.: 61 3 8344 0648; fax: 62 3 9348 1174. E-mail: mkelaher@ unimelb.edu.au. The authors have no competing interests This project was funded through the Economic and Social Research Council Health Variations Programme and the Leeds Health Authority. Margaret Kelaher was supported in part by an Australian National Health and Medical Research Council Career Development award and VicHealth. Received August 7, 2007; accepted December 28, 2007. Ó 2008 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
for greater attention to the SEP indicator used and the causal processes through which it might affect health (7, 8). Concerns about the measurement of SEP have also become an important issue in research examining ethnicity in health. There is still considerable debate about the causal pathway between ethnicity and health and about the way in which ethnicity should be included in health research (23– 33). Attribution of the causes for ethnic differences in health has fluctuated over time between SEP and cultural and genetic differences. In recent research, there has been a tendency for residual variance in health due to ethnicity to be attributed to cultural and genetic differences, once SEP has been taken into account. It has been suggested that this assumes that measures of SEP are sufficiently accurate to account for all socioeconomic variation and that this is unlikely to be true partly because measures of SEP are unlikely to be equally applicable to all ethnic populations (30–33). A number of studies suggest that particular measures of SEP might not be equally applicable to all ethnic groups (33). For example, a study of variation within class groups suggested that minority ethnic groups had lower incomes than other British people in the same class (30). Similarly, 1047-2797/08/$–see front matter doi:10.1016/j.annepidem.2007.12.006
352
Kelaher et al. ETHNICITY AND HEALTH
research in the United States has found that calculating socioeconomic position on the basis of income understates the true magnitude of racial differences in economic resources because there are large racial gaps in wealth at every income level (34). Our previous research suggests that there are differences between white, African Caribbean, and Indian and Pakistani groups in economic priorities and opportunities, particularly in relation to car ownership, home ownership, investment, and debt (35). Differences in living conditions, household assets, and debt between ethnic groups were dependent on differences in education; however, differences in car and home ownership and perceived ability to obtain £10,000 remained after adjustment for education (35). This finding suggests that these alternative indicators of SEP might be able to make a unique contribution to understanding the relationship between SEP ethnicity and health. In this study we explicitly examine the effect of using different measures of SEP (car ownership, perceived ability to obtain £10 000, home ownership, and education level) on ethnic differences in health measures using both asset-based measures and more conventional measures. Car ownership and ability to obtain £10,000 reflect current ability to mobilize health-relevant resources. Home ownership and education are more likely to reflect the cumulative effects of SEP on health.
AEP Vol. 18, No. 5 May 2008: 351–356
and a detailed review of existing published and unpublished survey instruments used to assess socioeconomic position, reference to a national advisory group and piloting of questions using 12 interviews and 6 focus groups. The questionnaire included sections on SEP, social resources, discrimination, and health (35). Demographic and health questions were informed by the Census and Health Survey for England (37). Household and individual income questions were taken from Breadline Britain (38). An Urdu version of the questionnaire was available where required. Variables Ethnicity was coded into 3 categories: white, Indian and Pakistani as a group, and African Caribbean. All SEP variables were coded dichotomously. SEP variables included perceived ability to get £10,000 if needed, car ownership, level of education (secondary or below [reference] versus above secondary), and home ownership. Health variables included self-perceived health status; presence of a long-term illness or disability; presence of limitations arising from a long-term illness or disability; one or more limitations in mobility; obesity; currently anxious, worried, or depressed; visits to general practitioners (GPs) in the last 12 months and visits to GPs taking into account the presence of long-term conditions or disabilities. Analysis
INSTRUMENTS AND PROCEDURE We report on a community survey of 247 white, 233 Indian and Pakistani, and 212 African Caribbean adults aged 18–59 years, living in Leeds, United Kingdom. Potential respondents were identified in 7 electoral wards with more than 3% concentration of black and minority ethnic groups by using general practice lists as well as random walks and snowball sampling (35). Electoral wards were grouped according to the Townsend score into areas of low, medium, and high deprivation and sampling was conducted in each electoral ward in proportion to the number of people in each ethnic group within each deprivation grouping. This is important in order to avoid disproportionate sampling of ethnic minority populations from high-deprivation wards where they are overrepresented. Approximately two thirds of the sample were from ethnic minority groups compared with an average of 18% in the electoral wards where sampling took place (36). Because of the high representation of ethnic minority groups, the sample had slightly lower levels of education and car ownership than would be expected on the basis of overall ward statistics (36). There were similar proportions of men and women in the sample and in the electoral ward. The instrument used in the community survey was developed following preliminary analysis of the qualitative data that had been collected in the previous phase of the research
Logistic regression analysis was used to assess the relationship between ethnicity and SEP and health. Five models were run for each health variable so that the effects of changing the SEP measure on ethnic differences in health could be ascertained. The first model included only ethnicity as an independent variable and the remaining 4 models tested the effects of perceived ability to get £10,000, car ownership, level of education, and homeownership separately. All models controlled for age. The analyses were conducted in Intercooled Stata 8. RESULTS Table 1 shows the breakdown of SEP measures by ethnicity. SEP among white respondents tended to be higher than among the minority ethnic groups on all measures, with the exception of home ownership where rates were higher among Indians and Pakistanis as a group. Respondents in minority ethnic groups were more likely than respondents in the white group to report being in fair or poor health. Indians and Pakistanis were more likely than white respondents to report a long-standing illness, disability, limitation in mobility, or being worried or depressed. African Caribbean respondents were more likely than white respondents to be obese.
AEP Vol. 18, No. 5 May 2008: 351–356
353
Kelaher et al. ETHNICITY AND HEALTH
TABLE 1. Socioeconomic position by ethnicity Age-adjusted OR (95%CI) White SEP/Health measures SEP measures Able to get £10,000 Car ownership Level of education Home ownership Health measures Fair/poor health Long-standing illness Disability One or more limitations in mobility Obesity Being anxious, worried, or depressed
Indian and Pakistani No.
African Caribbean
No.
%
%
No.
%
151/247 136/247 97/232 141/247
61.1 55.1 41.8 57.1
122/233 143/233 55/197 167/233
52.4 61.4 27.9 71.7
80/212 69/212 58/200 93/212
37.7 32.5 29.0 43.9
44/247 43/247 24/247 29/247 32/247 17/247
17.8 17.4 9.7 11.7 13.0 6.9
69/233 45/233 30/233 44/233 30/233 26/233
29.6 19.3 12.9 18.9 12.9 11.2
46/212 28/212 17/212 24/212 42/212 13/212
21.7 13.2 8.0 11.3 19.8 6.1
OR Z odds ratio; 95%CI Z 95% confidence interval; SEP Z socioeconomic position. p ! 0.05.
Table 2 shows that when SEP was not controlled for, the Indian and Pakistani group was more likely than the white group to report fair or poor health. These differences remained, indeed apparently increased, when the SEP measure used was perceived ability to get £10,000, car ownership or home ownership (see Table 2). However, if the SEP measure used was level of education, then greater attenuation was observed. There was no robust difference between the African Caribbean group and the white group (see Table 2). There were no ethnic differences in the presence of longterm illness or disability or limitation due to long-term illness or disability when SEP was not taken into account (Table 3). However, the Indian and Pakistani group was more likely than the white group to report the presence of one or more limitations to mobility (see Table 3). Differences in disability levels between the Indian and Pakistani group and the white group apparently increased, when the SEP measure used was perceived ability to get £10,000 or car ownership (see Table 3). Adjustment for education did not influence the differences in disability levels between the Indian and Pakistani group and the white group in a consistent way, whereas adjustment for home ownership attenuated the effects to around the null. Table 4 shows that there were no differences in obesity levels between the Indian and Pakistani and the white groups in any of the models. The African Caribbean group was more likely than the white group to be classified as obese when no SEP was used (see Table 4). These differences were little influenced by adjustment for car ownership or ability to obtain £10,000, but were attenuated markedly by adjustment for education or home ownership (see Table 4). Rates of anxiety, worry, or depression were higher in the Indian and Pakistani group than in the white group when no SEP measure was included. These differences remained after
adjustment for each of the SEP measures. There were no differences in rates of anxiety, worry, or depression among the African Caribbean group compared with the white group (Table 5).
DISCUSSION Overall the results of the study suggest that adults in both broad minority ethnic groups studied have a somewhat less favorable profile of physical and mental health than their white counterparts. The relationship between ethnicity and health is modified by SEP, but the effects of including SEP vary depending on the measure used. Variation in the apparent relationship between ethnicity and health because of differences in the choice of SEP measure supports the contention that measurements of SEP are not sufficiently precise TABLE 2. Relationship between ethnicity and self-perceived health status adjusted by different SEP measures Health status: Age-adjusted OR (95%CI) SEP measure in each model* No SEP measure Fair/poor health Able to get £10,000 Fair/poor health Car ownership Fair/poor health Level of education Fair/poor health Home ownership Fair/poor health
Indian and Pakistani
African Caribbean
1.9 (1.3–3.0)y
1.3 (0.8–2.0)
2.4 (1.3–4.5)y
1.2 (0.7–2.3) 0.5 (0.3–1.0)
3.4 (1.8–6.5)y
1.7 (0.9–3.2) 1.0 (0.5–1.8)
1.7 (1.0–3.1)
1.5 (0.8–2.6) 0.8 (0.4–1.6)
2.4 (1.2–4.8)y
1.3 (0.7–2.4) 0.7 (0.3–1.3)
SEP measure
For abbreviations, see Table 1. *Each model tests a different SEP measure; ‘‘white’’ Z reference group for all analyses. y p ! 0.05.
354
Kelaher et al. ETHNICITY AND HEALTH
AEP Vol. 18, No. 5 May 2008: 351–356
TABLE 3. Relationship between ethnicity and disability and limitations in activities with different SEP measures* Disability/Limitation: Age-adjusted OR (95%CI)
No SEP measure Long-term illness/disability Limited by illness/disability One or more limitations Able to get £10,000 Long-term illness/disability Limited by illness/disability One or more limitations Car ownership Long-term illness/disability Limited by illness/disability One or more limitations Level of education Long-term illness/disability Limited by illness/disability One or more limitations Home ownership Long-term illness/disability Limited by illness/disability One or more limitations
Indian and Pakistani
African Caribbean
SEP measure
1.1 (0.7–1.8) 1.4 (0.8–2.4) 1.8 (1.1–2.9)y
0.7 (0.4–1.2) 0.8 (0.5–1.5) 1.0 (0.6–1.7)
1.8 (0.9–3.7) 2.0 (0.9–4.5) 2.2 (1.1–4.4)y
1.1 (0.5–2.3) 1.4 (0.6–3.3) 1.1 (0.5–2.4)
1.4 (0.7–2.8) 1.2 (0.5–2.7) 0.7 (0.3–1.5)
1.9 (0.9–3.8) 2.2 (1.0–4.9) 2.6 (1.3–5.1)y
0.9 (0.5–1.8) 1.1 (0.5–2.4) 1.0 (0.5–2.0)
1.2 (0.6–2.4) 1.0 (0.4–2.2) 0.6 (0.3–1.2)
1.6 (0.8–3.1) 1.6 (0.8–3.2) 1.6 (0.9–3.1)
1.3 (0.6–2.5) 1.0 (0.5–2.1) 0.9 (0.5–1.8)
2.3 (1.2–4.7) 0.9 (0.4–2.1) 0.9 (0.4–2.0)
0.8 (0.4–1.8) 1.1 (0.4–2.7) 1.1 (0.5–2.5)
0.5 (0.2–1.1) 0.7 (0.3–1.6) 0.6 (0.3–1.4)
0.7 (0.4–1.4) 0.6 (0.3–1.3) 0.5 (0.2–1.0)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
to enable the attribution of residual differences in ethnicity and health to culture or genetics once SEP is taken into account (30, 33). Testing multiple models is central to the purpose of this study, which is to assess whether the relationship between ethnicity and health is affected by the SEP measure used. However, this is likely to inflate the type 1 error rate; accordingly, the significance levels associated with ethnic differences should be interpreted with caution. Generally these data confirm previous research in terms of health differences between Indian and Pakistani adults and white adults (29, 36). There were no differences between the health status of African Caribbeans and whites in this sample survey, in contrast with the findings of previous research (29), which indicated that African Caribbeans
had worse health than whites. This may have occurred because the current study focused on adults in the 18- to 59-year-old age group, whereas other studies (29, 37, 39) include greater representation from older persons. In this study ethnic differences in health tended to be accentuated in models which included asset-based SEP measures, such as car ownership and perceived ability to obtain £10,000, which reflect an individual’s ability to mobilize resources in comparison to models that did not take SEP into account. In contrast, the inclusion of level of education or home ownership in models tended to have little effect on or reduce apparent ethnic differences in health. Overall, this suggests that choice of SEP measure is an important consideration in research attempting to disentangle
TABLE 4. Relationship between ethnicity and obesity with different SEP measures
TABLE 5. Relationship between ethnicity and the health variables of being anxious, worried, or depressed with different SEP measures Anxious, worried, or depressed: Age-adjusted OR (95%CI)
Obesity: Age-adjusted OR (95%CI) and regression coefficient* SEP measure in each model
Indian and Pakistani
African Caribbean
SEP measure
SEP measure in each model*
Indian and Pakistani
No SEP measure Able to get £10,000 Car ownership Level of education Home ownership
1.0 (0.6–1.7) 1.1 (0.5–2.4) 1.2 (0.5–2.8) 0.9 (0.5–1.6) 0.4 (0.1–1.2)
1.7 (1.1–2.8)y 1.9 (0.9–3.7) 2.5 (1.2–5.1)y 1.2 (0.7–2.2) 1.3 (0.6–2.6)
0.7 (0.3–1.4) 1.4 (0.7–3.1) 0.5 (0.2–1.1) 0.7 (0.3–1.5)
No SEP measure Able to get £10,000 Car ownership Level of education Home ownership
1.7 (0.9–3.2)y 2.7 (1.01–7.2)y 3.5 (1.4–9.0)y 1.7 (0.8–3.8) 2.4 (0.9–6.4)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
African Caribbean
SEP measure
0.9 (0.4–1.8) 1.4 (0.5–4.0) 1.4 (0.5–3.8) 1.3(0.5–3.5) 1.4(0.5–3.8) 1.1 (0.5–2.6) 0.9 (0.3–2.6) 1.0 (0.4–2.8) 0.8 (0.3–2.0)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
AEP Vol. 18, No. 5 May 2008: 351–356
the influence of different types of social determinants on health (11). Ethnic differences in health were generally attenuated following adjustment for education. This could be interpreted as suggesting that differences in health between ethnic groups were attributable to SEP and not ethnicity. However, variations in the relationship between ethnicity and health depending on the SEP measure used suggest that this interpretation is not adequate on its own. Indeed, there are a number of reasons why asset-based measures might be more sensitive to differences in health in general and ethnic differences in health. First, education may reflect structural inequities between the groups to a greater extent than asset-based measures. Car and home ownership and ability to obtain £10,000 were selected for inclusion in this study because ethnic variations in these indicators remained once education was taken into account. Unfortunately, the sample size in this study is too small to assess the effects of different SEP measures simultaneously; therefore their independent contribution to health cannot be assessed. Second, asset-based measures of SEP may be more sensitive to fluctuations in health than education. One possible explanation for this would be ‘‘reverse causation’’; there is a tendency for people in ill health to drift down in SEP. This downward drift is more likely to be reflected in measures of SEP, such as income, which can easily fluctuate with ill health rather than measures, such as education, which do not change in the face of ill health (11). However, car and home ownership would be expected to be relatively robust against all but the most catastrophic changes in health. Ability to obtain £10,000 might be expected to be affected by changes in personal earning capacity but would be expected to be less susceptible to change than income because it is a measure of wealth among social networks as well as individual wealth. Third, asset-based SEP measures might be temporally more closely related than education to an individual’s ability to protect their health from adverse exposures and mobilize health-related resources. This is particularly likely to be true in contexts where the least educated people are still likely to have adequate literacy. If this is the case, then it would be expected that the relationship between SEP and health would vary with the type of health outcome being assessed. For example, the elevation in the rate of anxiety, worry, and depression among Indians and Pakistanis as a group compared to the white group was increased when owning a car was taken into account. Not owning a car was associated with much higher levels of anxiety, worry, and depression among Indians and Pakistanis than in white people. This may be because driving was much more likely to be a primary characteristic of main earners’ jobs in Indian and Pakistani households (15.0% compared to 7.4% in white and 10.2%
Kelaher et al. ETHNICITY AND HEALTH
355
in African Caribbean households) and because private means of transport may be particularly important for facilitating contact among dispersed kin groups and for enabling mobility among South Asian Muslim women. Adjustment for education may have had a particularly marked effect on reducing apparent differences in obesity among African Caribbeans compared to the white group because it functions in part as a measure of social conditions in early life, and deprivation in childhood has consistently been related to higher prevalence of obesity (33). Overall, the results suggest that the relationship between ethnicity and health is modified by SEP but that the type of modification varies depending on the SEP measure used. The results reinforce concerns about measures of SEP being used interchangeably in health studies (11). The inclusion of asset-based SEP measures, such as car ownership and ability to obtain £10,000, in statistical models tended to increase ethnic differences in health, whereas the inclusion of education level and home ownership tended to have little effect or reduce ethnic differences in health. This suggests that there is less overlap between assetbased measures and ethnicity than education and ethnicity in accounting for health differences. Overall, the study suggests that choice of SEP measure will affect the conclusions of research on ethnicity and health. We thank all those who took part in the survey and the Leeds General Practices who participated for providing access to their GP lists. We would also like to thank Ghazala Mir for her help with name sampling and CPRC, University of Leeds for their support.
REFERENCES 1. Assai M, Siddiqi S, Watts S. Tackling social determinants of health through community based initiatives. BMJ. 2006;333:854–856. 2. Baum F. Wealth and health: the need for more strategic public health research. J Epidemiol Community Health. 2005;59:542–545. 3. Davey Smith G. Health inequalities: lifecourse approaches. Bristol (UK): Policy Press; 2003. 4. Davey Smith G, Hart C. Life-course socioeconomic and behavioral influences on cardiovascular disease mortality: the collaborative study. Am J Public Health. 2002;92:1295–1298. 5. Graham H. Social determinants and their unequal distribution: clarifying policy understandings. Milbank Q. 2004;82:101–124. 6. Lynch J, Davey Smith GD, Harper S, Hillemeier M. Is income inequality a determinant of population health? Part 2. U.S. National and regional trends in income inequality and age- and cause-specific mortality. Milbank Q. 2004;82:355–400. 7. Lynch J, Davey Smith G, Harper S, Hillemeier M, Ross N, Kaplan GA, et al. Is income inequality a determinant of population health? Part 1. A systematic review. Milbank Q. 2004;82:5–99. 8. Lynch SM. Explaining life course and cohort variation in the relationship between education and health: the role of income. J Health Soc Behav. 2006;47:324–338. 9. Power C, Hypponen E, Davey Smith G. Socioeconomic position in childhood and early adult life and risk of mortality: a prospective study of the
356
Kelaher et al. ETHNICITY AND HEALTH
mothers of the 1958 British birth cohort. Am J Public Health. 2005;95:1396–1402. 10. Welshman J. Searching for social capital: historical perspectives on health, poverty and culture. J R Soc Health. 2006;126:268–274. 11. Geyer S, Hemstrom O, Peter R, Vagero D. Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. J Epidemiol Community Health. 2006;60:804–810. 12. Alter DA, Iron K, Austin PC, Naylor CD. Influence of education and income on atherogenic risk factor profiles among patients hospitalized with acute myocardial infarction. Can J Cardiol. 2004;20:1219–1228. 13. Araya R, Lewis G, Rojas G, Fritsch R. Education and income: which is more important for mental health? J Epidemiol Community Health. 2003;57:501–505. 14. Kanjilal S, Gregg EW, Cheng YJ, Zhang P, Nelson DE, Mensah G, et al. Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971-2002. Arch Intern Med. 2006;166:2348–2355. 15. Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R. Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). J Epidemiol Community Health. 2003;57:186–199. 16. Lahelma E, Martikainen P, Laaksonen M, Aittomaki A. Pathways between socioeconomic determinants of health. J Epidemiol Community Health. 2004;58:327–332. 17. Maty SC, Everson-Rose SA, Haan MN, Raghunathan TE, Kaplan GA. Education, income, occupation, and the 34-year incidence (1965-99) of Type 2 diabetes in the Alameda County Study. Int J Epidemiol. 2005;34:1274– 1281. 18. Naess O, Claussen B, Thelle DS, Davey Smith G. Four indicators of socioeconomic position: relative ranking across causes of death. Scand J Public Health. 2005;33:215–221. 19. Rognerud MA, Zahl P-H. Social inequalities in mortality: changes in the relative importance of income, education and household size over a 27year period. Eur J Public Health. 2006;16:62–68. 20. Sellstrom E, Bremberg S. The significance of neighbourhood context to child and adolescent health and well-being: a systematic review of multilevel studies. Scand J Public Health. 2006;34:544–554. 21. Shohaimi S, Bingham S, Welch A, Luben R, Day N, Wareham N, et al. Occupational social class, educational level and area deprivation independently predict plasma ascorbic acid concentration: a cross-sectional population-based study in the Norfolk cohort of the European Prospective Investigation into Cancer (EPIC-Norfolk). Eur J Clin Nutr. 2004;58:1432–1435.
AEP Vol. 18, No. 5 May 2008: 351–356
22. Ahmad W. Ethnic statistics: better than nothing or worse than Nothing? In: Dorling D, Simpson S, eds. Statistics in society. London: Arnold; 1999:124–131. 23. Bhopal R. Is research into ethnicity and health racist, unsound, or important science? BMJ. 1997;314:1751. 24. Braun L. Reifying human difference: the debate on genetics, race, and health. Int J Health Serv. 2006;36:557–573. 25. Burchard EG, Ziv E, Coyle N, Gomez SL, Tang H, Karter AJ, et al. The importance of race and ethnic background in biomedical research and clinical practice. N Engl J Med. 2003;348:1170–1175. 26. Frank R. What to make of it? The (Re)emergence of a biological conceptualization of race in health disparities research. Soc Sci Med. 2007;64:1977–1983. 27. Hilton C. Collecting ethnic group data for inpatients: is it useful? BMJ. 1996;313:923–925. 28. Kahn J. Genes, race, and population: avoiding a collision of categories. Am J Public Health. 2006;96:1965–1970. 29. Nazroo JY. The health of Britain’s ethnic minorities: findings from a national survey. London (UK): Policy Studies Institute; 1997 p. 8–11. 30. Nazroo JY. The structuring of ethnic inequalities in health: economic position, racial discrimination, and racism. Am J Public Health. 2003;93:277–284. 31. Risch N, Burchard E, Ziv E, Tang H. Categorization of humans in biomedical research: genes, race and disease. Genome Biol. 2002;3 comment2007. 32. Senior PA, Bhopal R. Ethnicity as a variable in epidemiological research. BMJ. 1994;309:327–330. 33. Davey Smith G. Learning to live with complexity: ethnicity, socioeconomic position, and health in Britain and the United States. Am J Public Health. 2000;90:1694–1698. 34. Williams DR. Race, socioeconomic status and health: the added effects of socioeconomic status and racism. Ann NY Acad Sci. 1999;896:173–188. 35. Kelaher M, Paul S, Lambert H, Ahmad W, Fenton S, Davey Smith G. Ethnicity, health and health services utilisation in a British study. Crit Public Health Issues. 2003;13:231–249. 36. National Statistics (UK). Neighbourhood statistics 2001 Census. Available at: http://neighbourhood.statistics.gov.uk. Accessed December 11, 2007. 37. Department of Health (UK). The health of ethnic minority groups. Health Survey for England 1999. Available at: http://www.archive.officialdocuments.co.uk/document/doh/survey99/hses-02.html, 2002. 38. Gordon D, Pantazis C. Breadline Britain in the 1990s. Aldershot: Ashgate; 1997. 39. Curtis S, Lawson K. Gender, ethnicity and self-reported health: the case of African-Caribbean populations in London. Soc Sci Med. 2000;50:365–385.
PURPOSE: In this study we explicitly examine the effect of using asset-based and conventional measures of socioeconomic position (SEP) on ethnic differences in health. METHODS: The survey of white (n Z 227), Indian and Pakistani (n Z 233), and African Caribbean (n Z 213) adults aged 18 to 59 years living in Leeds, United Kingdom aimed to examine the relationship between ethnicity, health, and SEP. SEP variables included perceived ability to get £10,000 if needed, car ownership, level of education, and home ownership. Health variables included self-reported health status, presence of a long-term illness or disability, presence of limitations arising from a long-term illness or disability, one or more limitations in mobility, obesity, and being anxious, worried, or depressed. Logistic regression analysis was used to assess the relationship between ethnicity and SEP and health. Five models were run for each health variable so that the effects of changing the SEP measure could be ascertained. The first model included only ethnicity and the remaining 4 models tested the effects of the perceived ability to get £10,000, car ownership, level of education, and home ownership separately. RESULTS: The results suggest that the statistical inclusion of asset-based SEP measures, such as car ownership and ability to obtain £10,000, which reflect an individual’s ability to mobilize resources, tend to increase ethnic differences in health, whereas more conventional steady-state indicators, such as education level and home ownership, tended to have little effect or to reduce ethnic differences in health. CONCLUSIONS: Overall, this study suggests that the choice of SEP measure may affect the conclusion of research on ethnicity and health and that choice of SEP measures should in turn be informed by the research problem being examined. Ann Epidemiol 2008;18:351–356. Ó 2008 Elsevier Inc. All rights reserved. KEY WORDS:
Ethnicity, Socioeconomic Measures, Health.
INTRODUCTION There is a vast body of research linking socioeconomic position (SEP) to health (1–10). However, there is increasing evidence that not all indicators of SEP are interchangeable (11). A number of studies report variation in the apparent relationship between SEP and health dependent on the type of measures used (11–22). This has resulted in a call
From the Centre for Health Policy, Programs and Economics, School of Population Health, University of Melbourne, Carlton, Victoria, Australia (M.K.); UCL Centre for International Health and Development, Institute of Child Health (S.P) and Social Policy Research Centre, Middlesex University (W.A.), London, UK; Department of Social Medicine, University of Bristol (H.L, G.D.S.), UK. Address correspondence to: Dr Margaret Kelaher, School of Population Health, University of Melbourne, 207 Bouverie St, Carlton Vic 3010, Australia. Tel.: 61 3 8344 0648; fax: 62 3 9348 1174. E-mail: mkelaher@ unimelb.edu.au. The authors have no competing interests This project was funded through the Economic and Social Research Council Health Variations Programme and the Leeds Health Authority. Margaret Kelaher was supported in part by an Australian National Health and Medical Research Council Career Development award and VicHealth. Received August 7, 2007; accepted December 28, 2007. Ó 2008 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
for greater attention to the SEP indicator used and the causal processes through which it might affect health (7, 8). Concerns about the measurement of SEP have also become an important issue in research examining ethnicity in health. There is still considerable debate about the causal pathway between ethnicity and health and about the way in which ethnicity should be included in health research (23– 33). Attribution of the causes for ethnic differences in health has fluctuated over time between SEP and cultural and genetic differences. In recent research, there has been a tendency for residual variance in health due to ethnicity to be attributed to cultural and genetic differences, once SEP has been taken into account. It has been suggested that this assumes that measures of SEP are sufficiently accurate to account for all socioeconomic variation and that this is unlikely to be true partly because measures of SEP are unlikely to be equally applicable to all ethnic populations (30–33). A number of studies suggest that particular measures of SEP might not be equally applicable to all ethnic groups (33). For example, a study of variation within class groups suggested that minority ethnic groups had lower incomes than other British people in the same class (30). Similarly, 1047-2797/08/$–see front matter doi:10.1016/j.annepidem.2007.12.006
352
Kelaher et al. ETHNICITY AND HEALTH
research in the United States has found that calculating socioeconomic position on the basis of income understates the true magnitude of racial differences in economic resources because there are large racial gaps in wealth at every income level (34). Our previous research suggests that there are differences between white, African Caribbean, and Indian and Pakistani groups in economic priorities and opportunities, particularly in relation to car ownership, home ownership, investment, and debt (35). Differences in living conditions, household assets, and debt between ethnic groups were dependent on differences in education; however, differences in car and home ownership and perceived ability to obtain £10,000 remained after adjustment for education (35). This finding suggests that these alternative indicators of SEP might be able to make a unique contribution to understanding the relationship between SEP ethnicity and health. In this study we explicitly examine the effect of using different measures of SEP (car ownership, perceived ability to obtain £10 000, home ownership, and education level) on ethnic differences in health measures using both asset-based measures and more conventional measures. Car ownership and ability to obtain £10,000 reflect current ability to mobilize health-relevant resources. Home ownership and education are more likely to reflect the cumulative effects of SEP on health.
AEP Vol. 18, No. 5 May 2008: 351–356
and a detailed review of existing published and unpublished survey instruments used to assess socioeconomic position, reference to a national advisory group and piloting of questions using 12 interviews and 6 focus groups. The questionnaire included sections on SEP, social resources, discrimination, and health (35). Demographic and health questions were informed by the Census and Health Survey for England (37). Household and individual income questions were taken from Breadline Britain (38). An Urdu version of the questionnaire was available where required. Variables Ethnicity was coded into 3 categories: white, Indian and Pakistani as a group, and African Caribbean. All SEP variables were coded dichotomously. SEP variables included perceived ability to get £10,000 if needed, car ownership, level of education (secondary or below [reference] versus above secondary), and home ownership. Health variables included self-perceived health status; presence of a long-term illness or disability; presence of limitations arising from a long-term illness or disability; one or more limitations in mobility; obesity; currently anxious, worried, or depressed; visits to general practitioners (GPs) in the last 12 months and visits to GPs taking into account the presence of long-term conditions or disabilities. Analysis
INSTRUMENTS AND PROCEDURE We report on a community survey of 247 white, 233 Indian and Pakistani, and 212 African Caribbean adults aged 18–59 years, living in Leeds, United Kingdom. Potential respondents were identified in 7 electoral wards with more than 3% concentration of black and minority ethnic groups by using general practice lists as well as random walks and snowball sampling (35). Electoral wards were grouped according to the Townsend score into areas of low, medium, and high deprivation and sampling was conducted in each electoral ward in proportion to the number of people in each ethnic group within each deprivation grouping. This is important in order to avoid disproportionate sampling of ethnic minority populations from high-deprivation wards where they are overrepresented. Approximately two thirds of the sample were from ethnic minority groups compared with an average of 18% in the electoral wards where sampling took place (36). Because of the high representation of ethnic minority groups, the sample had slightly lower levels of education and car ownership than would be expected on the basis of overall ward statistics (36). There were similar proportions of men and women in the sample and in the electoral ward. The instrument used in the community survey was developed following preliminary analysis of the qualitative data that had been collected in the previous phase of the research
Logistic regression analysis was used to assess the relationship between ethnicity and SEP and health. Five models were run for each health variable so that the effects of changing the SEP measure on ethnic differences in health could be ascertained. The first model included only ethnicity as an independent variable and the remaining 4 models tested the effects of perceived ability to get £10,000, car ownership, level of education, and homeownership separately. All models controlled for age. The analyses were conducted in Intercooled Stata 8. RESULTS Table 1 shows the breakdown of SEP measures by ethnicity. SEP among white respondents tended to be higher than among the minority ethnic groups on all measures, with the exception of home ownership where rates were higher among Indians and Pakistanis as a group. Respondents in minority ethnic groups were more likely than respondents in the white group to report being in fair or poor health. Indians and Pakistanis were more likely than white respondents to report a long-standing illness, disability, limitation in mobility, or being worried or depressed. African Caribbean respondents were more likely than white respondents to be obese.
AEP Vol. 18, No. 5 May 2008: 351–356
353
Kelaher et al. ETHNICITY AND HEALTH
TABLE 1. Socioeconomic position by ethnicity Age-adjusted OR (95%CI) White SEP/Health measures SEP measures Able to get £10,000 Car ownership Level of education Home ownership Health measures Fair/poor health Long-standing illness Disability One or more limitations in mobility Obesity Being anxious, worried, or depressed
Indian and Pakistani No.
African Caribbean
No.
%
%
No.
%
151/247 136/247 97/232 141/247
61.1 55.1 41.8 57.1
122/233 143/233 55/197 167/233
52.4 61.4 27.9 71.7
80/212 69/212 58/200 93/212
37.7 32.5 29.0 43.9
44/247 43/247 24/247 29/247 32/247 17/247
17.8 17.4 9.7 11.7 13.0 6.9
69/233 45/233 30/233 44/233 30/233 26/233
29.6 19.3 12.9 18.9 12.9 11.2
46/212 28/212 17/212 24/212 42/212 13/212
21.7 13.2 8.0 11.3 19.8 6.1
OR Z odds ratio; 95%CI Z 95% confidence interval; SEP Z socioeconomic position. p ! 0.05.
Table 2 shows that when SEP was not controlled for, the Indian and Pakistani group was more likely than the white group to report fair or poor health. These differences remained, indeed apparently increased, when the SEP measure used was perceived ability to get £10,000, car ownership or home ownership (see Table 2). However, if the SEP measure used was level of education, then greater attenuation was observed. There was no robust difference between the African Caribbean group and the white group (see Table 2). There were no ethnic differences in the presence of longterm illness or disability or limitation due to long-term illness or disability when SEP was not taken into account (Table 3). However, the Indian and Pakistani group was more likely than the white group to report the presence of one or more limitations to mobility (see Table 3). Differences in disability levels between the Indian and Pakistani group and the white group apparently increased, when the SEP measure used was perceived ability to get £10,000 or car ownership (see Table 3). Adjustment for education did not influence the differences in disability levels between the Indian and Pakistani group and the white group in a consistent way, whereas adjustment for home ownership attenuated the effects to around the null. Table 4 shows that there were no differences in obesity levels between the Indian and Pakistani and the white groups in any of the models. The African Caribbean group was more likely than the white group to be classified as obese when no SEP was used (see Table 4). These differences were little influenced by adjustment for car ownership or ability to obtain £10,000, but were attenuated markedly by adjustment for education or home ownership (see Table 4). Rates of anxiety, worry, or depression were higher in the Indian and Pakistani group than in the white group when no SEP measure was included. These differences remained after
adjustment for each of the SEP measures. There were no differences in rates of anxiety, worry, or depression among the African Caribbean group compared with the white group (Table 5).
DISCUSSION Overall the results of the study suggest that adults in both broad minority ethnic groups studied have a somewhat less favorable profile of physical and mental health than their white counterparts. The relationship between ethnicity and health is modified by SEP, but the effects of including SEP vary depending on the measure used. Variation in the apparent relationship between ethnicity and health because of differences in the choice of SEP measure supports the contention that measurements of SEP are not sufficiently precise TABLE 2. Relationship between ethnicity and self-perceived health status adjusted by different SEP measures Health status: Age-adjusted OR (95%CI) SEP measure in each model* No SEP measure Fair/poor health Able to get £10,000 Fair/poor health Car ownership Fair/poor health Level of education Fair/poor health Home ownership Fair/poor health
Indian and Pakistani
African Caribbean
1.9 (1.3–3.0)y
1.3 (0.8–2.0)
2.4 (1.3–4.5)y
1.2 (0.7–2.3) 0.5 (0.3–1.0)
3.4 (1.8–6.5)y
1.7 (0.9–3.2) 1.0 (0.5–1.8)
1.7 (1.0–3.1)
1.5 (0.8–2.6) 0.8 (0.4–1.6)
2.4 (1.2–4.8)y
1.3 (0.7–2.4) 0.7 (0.3–1.3)
SEP measure
For abbreviations, see Table 1. *Each model tests a different SEP measure; ‘‘white’’ Z reference group for all analyses. y p ! 0.05.
354
Kelaher et al. ETHNICITY AND HEALTH
AEP Vol. 18, No. 5 May 2008: 351–356
TABLE 3. Relationship between ethnicity and disability and limitations in activities with different SEP measures* Disability/Limitation: Age-adjusted OR (95%CI)
No SEP measure Long-term illness/disability Limited by illness/disability One or more limitations Able to get £10,000 Long-term illness/disability Limited by illness/disability One or more limitations Car ownership Long-term illness/disability Limited by illness/disability One or more limitations Level of education Long-term illness/disability Limited by illness/disability One or more limitations Home ownership Long-term illness/disability Limited by illness/disability One or more limitations
Indian and Pakistani
African Caribbean
SEP measure
1.1 (0.7–1.8) 1.4 (0.8–2.4) 1.8 (1.1–2.9)y
0.7 (0.4–1.2) 0.8 (0.5–1.5) 1.0 (0.6–1.7)
1.8 (0.9–3.7) 2.0 (0.9–4.5) 2.2 (1.1–4.4)y
1.1 (0.5–2.3) 1.4 (0.6–3.3) 1.1 (0.5–2.4)
1.4 (0.7–2.8) 1.2 (0.5–2.7) 0.7 (0.3–1.5)
1.9 (0.9–3.8) 2.2 (1.0–4.9) 2.6 (1.3–5.1)y
0.9 (0.5–1.8) 1.1 (0.5–2.4) 1.0 (0.5–2.0)
1.2 (0.6–2.4) 1.0 (0.4–2.2) 0.6 (0.3–1.2)
1.6 (0.8–3.1) 1.6 (0.8–3.2) 1.6 (0.9–3.1)
1.3 (0.6–2.5) 1.0 (0.5–2.1) 0.9 (0.5–1.8)
2.3 (1.2–4.7) 0.9 (0.4–2.1) 0.9 (0.4–2.0)
0.8 (0.4–1.8) 1.1 (0.4–2.7) 1.1 (0.5–2.5)
0.5 (0.2–1.1) 0.7 (0.3–1.6) 0.6 (0.3–1.4)
0.7 (0.4–1.4) 0.6 (0.3–1.3) 0.5 (0.2–1.0)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
to enable the attribution of residual differences in ethnicity and health to culture or genetics once SEP is taken into account (30, 33). Testing multiple models is central to the purpose of this study, which is to assess whether the relationship between ethnicity and health is affected by the SEP measure used. However, this is likely to inflate the type 1 error rate; accordingly, the significance levels associated with ethnic differences should be interpreted with caution. Generally these data confirm previous research in terms of health differences between Indian and Pakistani adults and white adults (29, 36). There were no differences between the health status of African Caribbeans and whites in this sample survey, in contrast with the findings of previous research (29), which indicated that African Caribbeans
had worse health than whites. This may have occurred because the current study focused on adults in the 18- to 59-year-old age group, whereas other studies (29, 37, 39) include greater representation from older persons. In this study ethnic differences in health tended to be accentuated in models which included asset-based SEP measures, such as car ownership and perceived ability to obtain £10,000, which reflect an individual’s ability to mobilize resources in comparison to models that did not take SEP into account. In contrast, the inclusion of level of education or home ownership in models tended to have little effect on or reduce apparent ethnic differences in health. Overall, this suggests that choice of SEP measure is an important consideration in research attempting to disentangle
TABLE 4. Relationship between ethnicity and obesity with different SEP measures
TABLE 5. Relationship between ethnicity and the health variables of being anxious, worried, or depressed with different SEP measures Anxious, worried, or depressed: Age-adjusted OR (95%CI)
Obesity: Age-adjusted OR (95%CI) and regression coefficient* SEP measure in each model
Indian and Pakistani
African Caribbean
SEP measure
SEP measure in each model*
Indian and Pakistani
No SEP measure Able to get £10,000 Car ownership Level of education Home ownership
1.0 (0.6–1.7) 1.1 (0.5–2.4) 1.2 (0.5–2.8) 0.9 (0.5–1.6) 0.4 (0.1–1.2)
1.7 (1.1–2.8)y 1.9 (0.9–3.7) 2.5 (1.2–5.1)y 1.2 (0.7–2.2) 1.3 (0.6–2.6)
0.7 (0.3–1.4) 1.4 (0.7–3.1) 0.5 (0.2–1.1) 0.7 (0.3–1.5)
No SEP measure Able to get £10,000 Car ownership Level of education Home ownership
1.7 (0.9–3.2)y 2.7 (1.01–7.2)y 3.5 (1.4–9.0)y 1.7 (0.8–3.8) 2.4 (0.9–6.4)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
African Caribbean
SEP measure
0.9 (0.4–1.8) 1.4 (0.5–4.0) 1.4 (0.5–3.8) 1.3(0.5–3.5) 1.4(0.5–3.8) 1.1 (0.5–2.6) 0.9 (0.3–2.6) 1.0 (0.4–2.8) 0.8 (0.3–2.0)
For abbreviations, see Table 1. *Each model tests a different SEP measure; white Z reference group for all analyses. y p ! 0.05.
AEP Vol. 18, No. 5 May 2008: 351–356
the influence of different types of social determinants on health (11). Ethnic differences in health were generally attenuated following adjustment for education. This could be interpreted as suggesting that differences in health between ethnic groups were attributable to SEP and not ethnicity. However, variations in the relationship between ethnicity and health depending on the SEP measure used suggest that this interpretation is not adequate on its own. Indeed, there are a number of reasons why asset-based measures might be more sensitive to differences in health in general and ethnic differences in health. First, education may reflect structural inequities between the groups to a greater extent than asset-based measures. Car and home ownership and ability to obtain £10,000 were selected for inclusion in this study because ethnic variations in these indicators remained once education was taken into account. Unfortunately, the sample size in this study is too small to assess the effects of different SEP measures simultaneously; therefore their independent contribution to health cannot be assessed. Second, asset-based measures of SEP may be more sensitive to fluctuations in health than education. One possible explanation for this would be ‘‘reverse causation’’; there is a tendency for people in ill health to drift down in SEP. This downward drift is more likely to be reflected in measures of SEP, such as income, which can easily fluctuate with ill health rather than measures, such as education, which do not change in the face of ill health (11). However, car and home ownership would be expected to be relatively robust against all but the most catastrophic changes in health. Ability to obtain £10,000 might be expected to be affected by changes in personal earning capacity but would be expected to be less susceptible to change than income because it is a measure of wealth among social networks as well as individual wealth. Third, asset-based SEP measures might be temporally more closely related than education to an individual’s ability to protect their health from adverse exposures and mobilize health-related resources. This is particularly likely to be true in contexts where the least educated people are still likely to have adequate literacy. If this is the case, then it would be expected that the relationship between SEP and health would vary with the type of health outcome being assessed. For example, the elevation in the rate of anxiety, worry, and depression among Indians and Pakistanis as a group compared to the white group was increased when owning a car was taken into account. Not owning a car was associated with much higher levels of anxiety, worry, and depression among Indians and Pakistanis than in white people. This may be because driving was much more likely to be a primary characteristic of main earners’ jobs in Indian and Pakistani households (15.0% compared to 7.4% in white and 10.2%
Kelaher et al. ETHNICITY AND HEALTH
355
in African Caribbean households) and because private means of transport may be particularly important for facilitating contact among dispersed kin groups and for enabling mobility among South Asian Muslim women. Adjustment for education may have had a particularly marked effect on reducing apparent differences in obesity among African Caribbeans compared to the white group because it functions in part as a measure of social conditions in early life, and deprivation in childhood has consistently been related to higher prevalence of obesity (33). Overall, the results suggest that the relationship between ethnicity and health is modified by SEP but that the type of modification varies depending on the SEP measure used. The results reinforce concerns about measures of SEP being used interchangeably in health studies (11). The inclusion of asset-based SEP measures, such as car ownership and ability to obtain £10,000, in statistical models tended to increase ethnic differences in health, whereas the inclusion of education level and home ownership tended to have little effect or reduce ethnic differences in health. This suggests that there is less overlap between assetbased measures and ethnicity than education and ethnicity in accounting for health differences. Overall, the study suggests that choice of SEP measure will affect the conclusions of research on ethnicity and health. We thank all those who took part in the survey and the Leeds General Practices who participated for providing access to their GP lists. We would also like to thank Ghazala Mir for her help with name sampling and CPRC, University of Leeds for their support.
REFERENCES 1. Assai M, Siddiqi S, Watts S. Tackling social determinants of health through community based initiatives. BMJ. 2006;333:854–856. 2. Baum F. Wealth and health: the need for more strategic public health research. J Epidemiol Community Health. 2005;59:542–545. 3. Davey Smith G. Health inequalities: lifecourse approaches. Bristol (UK): Policy Press; 2003. 4. Davey Smith G, Hart C. Life-course socioeconomic and behavioral influences on cardiovascular disease mortality: the collaborative study. Am J Public Health. 2002;92:1295–1298. 5. Graham H. Social determinants and their unequal distribution: clarifying policy understandings. Milbank Q. 2004;82:101–124. 6. Lynch J, Davey Smith GD, Harper S, Hillemeier M. Is income inequality a determinant of population health? Part 2. U.S. National and regional trends in income inequality and age- and cause-specific mortality. Milbank Q. 2004;82:355–400. 7. Lynch J, Davey Smith G, Harper S, Hillemeier M, Ross N, Kaplan GA, et al. Is income inequality a determinant of population health? Part 1. A systematic review. Milbank Q. 2004;82:5–99. 8. Lynch SM. Explaining life course and cohort variation in the relationship between education and health: the role of income. J Health Soc Behav. 2006;47:324–338. 9. Power C, Hypponen E, Davey Smith G. Socioeconomic position in childhood and early adult life and risk of mortality: a prospective study of the
356
Kelaher et al. ETHNICITY AND HEALTH
mothers of the 1958 British birth cohort. Am J Public Health. 2005;95:1396–1402. 10. Welshman J. Searching for social capital: historical perspectives on health, poverty and culture. J R Soc Health. 2006;126:268–274. 11. Geyer S, Hemstrom O, Peter R, Vagero D. Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. J Epidemiol Community Health. 2006;60:804–810. 12. Alter DA, Iron K, Austin PC, Naylor CD. Influence of education and income on atherogenic risk factor profiles among patients hospitalized with acute myocardial infarction. Can J Cardiol. 2004;20:1219–1228. 13. Araya R, Lewis G, Rojas G, Fritsch R. Education and income: which is more important for mental health? J Epidemiol Community Health. 2003;57:501–505. 14. Kanjilal S, Gregg EW, Cheng YJ, Zhang P, Nelson DE, Mensah G, et al. Socioeconomic status and trends in disparities in 4 major risk factors for cardiovascular disease among US adults, 1971-2002. Arch Intern Med. 2006;166:2348–2355. 15. Krieger N, Chen JT, Waterman PD, Soobader MJ, Subramanian SV, Carson R. Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). J Epidemiol Community Health. 2003;57:186–199. 16. Lahelma E, Martikainen P, Laaksonen M, Aittomaki A. Pathways between socioeconomic determinants of health. J Epidemiol Community Health. 2004;58:327–332. 17. Maty SC, Everson-Rose SA, Haan MN, Raghunathan TE, Kaplan GA. Education, income, occupation, and the 34-year incidence (1965-99) of Type 2 diabetes in the Alameda County Study. Int J Epidemiol. 2005;34:1274– 1281. 18. Naess O, Claussen B, Thelle DS, Davey Smith G. Four indicators of socioeconomic position: relative ranking across causes of death. Scand J Public Health. 2005;33:215–221. 19. Rognerud MA, Zahl P-H. Social inequalities in mortality: changes in the relative importance of income, education and household size over a 27year period. Eur J Public Health. 2006;16:62–68. 20. Sellstrom E, Bremberg S. The significance of neighbourhood context to child and adolescent health and well-being: a systematic review of multilevel studies. Scand J Public Health. 2006;34:544–554. 21. Shohaimi S, Bingham S, Welch A, Luben R, Day N, Wareham N, et al. Occupational social class, educational level and area deprivation independently predict plasma ascorbic acid concentration: a cross-sectional population-based study in the Norfolk cohort of the European Prospective Investigation into Cancer (EPIC-Norfolk). Eur J Clin Nutr. 2004;58:1432–1435.
AEP Vol. 18, No. 5 May 2008: 351–356
22. Ahmad W. Ethnic statistics: better than nothing or worse than Nothing? In: Dorling D, Simpson S, eds. Statistics in society. London: Arnold; 1999:124–131. 23. Bhopal R. Is research into ethnicity and health racist, unsound, or important science? BMJ. 1997;314:1751. 24. Braun L. Reifying human difference: the debate on genetics, race, and health. Int J Health Serv. 2006;36:557–573. 25. Burchard EG, Ziv E, Coyle N, Gomez SL, Tang H, Karter AJ, et al. The importance of race and ethnic background in biomedical research and clinical practice. N Engl J Med. 2003;348:1170–1175. 26. Frank R. What to make of it? The (Re)emergence of a biological conceptualization of race in health disparities research. Soc Sci Med. 2007;64:1977–1983. 27. Hilton C. Collecting ethnic group data for inpatients: is it useful? BMJ. 1996;313:923–925. 28. Kahn J. Genes, race, and population: avoiding a collision of categories. Am J Public Health. 2006;96:1965–1970. 29. Nazroo JY. The health of Britain’s ethnic minorities: findings from a national survey. London (UK): Policy Studies Institute; 1997 p. 8–11. 30. Nazroo JY. The structuring of ethnic inequalities in health: economic position, racial discrimination, and racism. Am J Public Health. 2003;93:277–284. 31. Risch N, Burchard E, Ziv E, Tang H. Categorization of humans in biomedical research: genes, race and disease. Genome Biol. 2002;3 comment2007. 32. Senior PA, Bhopal R. Ethnicity as a variable in epidemiological research. BMJ. 1994;309:327–330. 33. Davey Smith G. Learning to live with complexity: ethnicity, socioeconomic position, and health in Britain and the United States. Am J Public Health. 2000;90:1694–1698. 34. Williams DR. Race, socioeconomic status and health: the added effects of socioeconomic status and racism. Ann NY Acad Sci. 1999;896:173–188. 35. Kelaher M, Paul S, Lambert H, Ahmad W, Fenton S, Davey Smith G. Ethnicity, health and health services utilisation in a British study. Crit Public Health Issues. 2003;13:231–249. 36. National Statistics (UK). Neighbourhood statistics 2001 Census. Available at: http://neighbourhood.statistics.gov.uk. Accessed December 11, 2007. 37. Department of Health (UK). The health of ethnic minority groups. Health Survey for England 1999. Available at: http://www.archive.officialdocuments.co.uk/document/doh/survey99/hses-02.html, 2002. 38. Gordon D, Pantazis C. Breadline Britain in the 1990s. Aldershot: Ashgate; 1997. 39. Curtis S, Lawson K. Gender, ethnicity and self-reported health: the case of African-Caribbean populations in London. Soc Sci Med. 2000;50:365–385.