Socioeconomic Status in Health Research One Size Does Not Fit All

SPECIAL COMMUNICATION

Paula A. Braveman, MD, MPH Problems with measuring socioeconomic status (SES)—frequently included Catherine Cubbin, PhD in clinical and public health studies as a control variable and less frequently Susan Egerter, PhD as the variable(s) of main interest—could affect research findings and con- Sekai Chideya, MD, MPH clusions, with implications for practice and policy. We critically examine stan- Kristen S. Marchi, MPH dard SES measurement approaches, illustrating problems with examples from Marilyn Metzler, RN new analyses and the literature. For example, marked racial/ethnic differ- Samuel Posner, PhD ences in income at a given educational level and in wealth at a given in- HE TERMS “SOCIOECONOMIC STA- come level raise questions about the socioeconomic comparability of indi- tus,” “socioeconomic posi- viduals who are similar on education or income alone. Evidence also shows tion,” and “social class” (col- that conclusions about nonsocioeconomic causes of racial/ethnic differ- lectively, “SES”) are widely used ences in health may depend on the measure—eg, income, wealth, educa- inT health research, reflecting wide- tion, occupation, neighborhood socioeconomic characteristics, or past so- spread albeit often implicit recognition cioeconomic experiences—used to “control for SES,” suggesting that findings of the importance of socioeconomic fac- from studies that have measured limited aspects of SES should be reas- tors for diverse health outcomes. Most health studies that consider SES treat so- sessed. We recommend an outcome- and social group–specific approach to cioeconomic characteristics as poten- SES measurement that involves (1) considering plausible explanatory path- tial confounders of relationships be- ways and mechanisms, (2) measuring as much relevant socioeconomic in- tween other variables and health. Others formation as possible, (3) specifying the particular socioeconomic factors explicitly examine relationships be- measured (rather than SES overall), and (4) systematically considering how tween SES and health, seeking to better potentially important unmeasured socioeconomic factors may affect con- understand the associations that have clusions. Better SES measures are needed in data sources, but improve- been repeatedly observed.1-22 The so- ments could be made by using existing information more thoughtfully and cial science and social epidemiology literature consistently treats SES as a mul- acknowledging its limitations. tidimensional construct comprising JAMA. 2005;294:2879-2888 www.jama.com diverse socioeconomic factors (typically, economic resources, power, and/or complex and multifactorial, most health variables without justifying why a given prestige).6,7,23-30 Different socioeco- studies that consider SES use a single so- measure was selected over others, with- nomic factors could affect health at dif- cioeconomic variable measured at a out explaining its meaning for a given ferent times in the life course,31-33 oper- single period and level. Often, very few analysis, and without discussing how ating at different levels (eg, individual, categories are used (eg, poor/nonpoor, household, neighborhood)30,34-37 and less than high school/high school gradu- Author Affiliations: Center on Social Disparities in Health and Department of Family and Community Medicine, through different causal path- ation or more schooling), which could University of California, San Francisco (Drs Braveman, 3,6,7,9,38-45 ways (eg, by determining expo- obscure important social gradients in Cubbin, and Egerter and Ms Marchi); and Epidemic In- sures, vulnerability, or even direct physi- health that apply across the entire so- telligence Service (Dr Chideya) and National Center for 40,42,44,45 2,8,9 Chronic Disease Prevention and Health Promotion, Di- ological effects). Socioeconomic cioeconomic spectrum. Occupation visions of Adult and Community Health (Ms Metzler) factors can interact with other social is frequently used as a measure of SES and Reproductive Health (Dr Posner), Centers for Dis- 11-14,16,17,52 ease Control and Prevention, Atlanta, Ga. characteristics, such as racial/ethnic in Europe, while income or Corresponding Author: Paula A. Braveman, MD, MPH, group and sex, to produce different education is more commonly used in the Center on Social Disparities in Health, University of 9,10,15,27,46-52 California, San Francisco, 500 Parnassus Ave, MU-3E, health effects across groups. United States. Regardless of the mea- Box 0900, San Francisco, CA 94143-0900 (braveman Despite expert consensus that SES is sure(s) used, most studies include SES @fcm.ucsf.edu).

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

unmeasured socioeconomic differ- proaches to SES measurement indi- phone survey); (2) indicators of differences might have affected findings. cates that these documents have not yet ent aspects of health (status, behaviors, This article critically examines sev- influenced most researchers’ prac- care) in (3) different life stages; and (4) eral widespread standard practices in tices. In many ways, this article paral- different commonly used and reason- SES measurement, discussing key con- lels a recent JAMA Editorial critiquing able measures of income and educa- cepts and providing examples to illus- the ways in which race/ethnicity is com- tion. Data sources, samples, and key SES trate problems with those practices. The monly measured in health research55 and health-related variables for the examples were chosen to illustrate SES and suggesting improved approaches new analyses are summarized in measurement issues, not to gain new based on more sound conceptualiza- TABLE 1.Weused5nationallyorstate- knowledge about the role of socioeco- tion. wide-representative data sources with nomic or other social factors in the se- well-documented strengths and limi- lected health indicators. Our goals were Methods tations: the multistate Behavioral Risk to increase awareness—not only among To illustrate key points, we selected ex- Factor Surveillance System (BRFSS), health researchers but also among amples from existing literature and from 200456; the National Health Interview policy makers and practitioners who new analyses of several national and Survey (NHIS) linked with the use health research findings—about in- statewide surveys, using the most re- National Death Index, 1989-1994, herent problems with standard meth- cent data available that were suitable for with mortality follow-up through ods of measuring and interpreting so- the desired analyses. The new analy- 1997 (mortality follow-up is not cur- cioeconomic characteristics in health ses were planned a priori to examine a rently available for later years of the research, as well as to encourage im- range of (1) widely used, high- NHIS)57;theThirdNationalHealthand proved approaches. While excellent cri- quality, population-based data sources Nutrition Examination Survey tiques have been available,10,25-28,53,54 the that represent a range of data collec- (NHANES III), 1988-1994 (both in- persistence of problematic ap- tion methods (eg, household vs tele- come and education were measured

Table 1. Summary of Data Sources, Samples, Measures of Socioeconomic Status (SES), and Dependent Variables Used in New Analyses of Health Surveys Dependent Variables Data Source Age Group, y Racial/Ethnic Groups Measures of SES* (Health-Related Indicators) Behavioral Risk Factor 18-74 (n = 255 896) Black (non-Hispanic) Poverty level Cigarette smoking Surveillance System, Hispanic Income (categories) Obesity 200456 White (non-Hispanic) Educational level Sedentary National Health Interview 18-64 (n = 380 552) Black (non-Hispanic) Poverty level Fair or poor health status Survey, 1989-1994 Mexican American Income (continuous) All-cause mortality (mortality follow-up White (non-Hispanic) Income/needs (continuous) Heart disease mortality through 1997)57 Income (quartiles) Homicide Income/needs (quartiles) Motor vehicle fatalities Educational level Education, y Third National Health and 1-5 (n = 5493) Black (non-Hispanic) Poverty level Elevated lead levels (ages 1-5 y) Nutrition Examination Ͻ17 (n = 12 388) Mexican American Income (continuous) Asthma (ages Ͻ17 y) Survey, 1988-199458 25-64 (n = 10 029) White (non-Hispanic) Educational level Cigarette smoking (ages 25-64 y) Obesity (ages 25-64 y) Hypertension (ages 25-64 y) Diabetes (ages 25-64 y) National Longitudinal Study 11-21 (n = 14 282) Black (non-Hispanic) Poverty level Cigarette smoking of Adolescent Health, Cuban Income (continuous) Initiation of sexual intercourse 1994-199559 Mexican American Income/needs (continuous) Violent behavior Puerto Rican Income (quartiles) Binge drinking White (non-Hispanic) Income/needs (quartiles) Marijuana use Educational level California Maternal and Infant Ͼ24 (n = 13 952) Asian/Pacific Islander Poverty level Unintended pregnancy Health Assessment, (immigrant or US-born) Income (continuous) Delayed or no prenatal care 1999-200460 Black Income/needs (continuous) Low birth weight Latina (immigrant or US-born) Income (quartiles) Never breastfed White Income/needs (quartiles) Educational level Education, y *Poverty level indicates annual income estimated in categories defined by 100% increments of the federal poverty level according to family or household size (0%-100%, 101%- 200%, 201%-300%, 301%-400%, Ͼ400%, unknown). Income (categories) indicates annual income in categories corresponding to ranges in dollars; income (continuous), annual income in continuous dollars estimated as the midpoints of a given income range; income/needs (continuous), annual income in continuous dollars divided by family size; income (quartiles), annual income in continuous dollars grouped into empirical quartiles; and income/needs (quartiles), annual income in continuous dollars divided by family size, grouped into empirical quartiles. Educational level indicates completed education in levels according to earned credentials (Ͻ9 years, some high school, high school graduate/General Equivalency Diploma, some college, college graduate or more); for California Maternal and Infant Health Assessment, Ͻ9 years and some high school are combined; the National Longitudinal Study of Adolescent Health also includes a category for unknown education. Education, y indicates completed education in years.

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

Table 2. Mean Family Income by Educational Level and Racial/Ethnic Group Among Adults Aged 18-64 Years—National Health Interview Survey, 1989-1994 (n = 380 552) Mean Income (95% Confidence Interval), $ Educational Level, y Black Mexican American White Overall Ͻ9 15 503 (14 851-16 154) 19 104 (17 549-20 659) 22 707 (21 935-23 479) 20 551 (19 868-21 234) 9-11 17 743 (17 051-18 434) 22 377 (20 807-23 948) 28 573 (28 062-29 083) 25 945 (25 486-26 403) 12 25 337 (24 529-26 144) 30 945 (29 791-32 098) 37 853 (37 400-38 305) 36 022 (35 613-36 432) 13-15 33 026 (31 898-34 154) 37 642 (36 089-39 195) 43 197 (42 510-43 884) 41 857 (41 228-42 487) Ն16 46 815 (45 535-48 095) 48 055 (46 137-49 973) 55 277 (54 725-55 829) 54 606 (54 064-55 149) Overall 27 937 (27 123-28 752) 27 927 (26 498-29 356) 42 113 (41 626-42 600) 39 716 (39 274-40 159)

more adequately in NHANES III than tails about methods are included in a proxy for income (or for SES overall). in the most recent NHANES data)58;the technical appendix available from the When both education and income are National Longitudinal Study of Ado- authors on request or at http://www available, researchers may hesitate to in- lescent Health (Add Health), 1994- .ucsf.edu/csdh. clude both in analytic models because 1995 (the baseline survey, when par- Race/ethnicity, regarded as a social of concerns about colinearity. Evi- ents or guardians of the adolescents construct, was assessed in the study pri- dence from the literature and our new responded to questions on income and marily to examine whether observed as- analyses indicates, however, that while education)59;andtheMaternalandIn- sociations between racial/ethnic groups standard measures of education and infant Health Assessment (MIHA), 1999- and health-related indicators varied by come are correlated, these correla- 2004,60 a statewide survey of postpar- the income measures, education mea- tions are generally not strong enough tum women in California supported by sures, or both used as covariables to re- to justify using education as a proxy for the California Department of Health flect SES; we also examined how a range income (or vice versa). Earnings can Services Maternal, Child and Adoles- of socioeconomic factors varied by ra- vary at similar educational levels, par- cent Health Branch.61-64 Using SAS soft- cial/ethnic group. While racial/ethnic ticularly across different social (eg, ra- ware,65 we examined correlations groups were categorized differently de- cial/ethnic, sex, age) groups. among multiple individual- or house- pending on the data source (see Table 1 For example, TABLE 2 shows, based hold-level measures of current educa- or text below), all categories were based on NHIS data, that black and Mexican tion and income. To provide addi- on self-reported information about re- American adults at every educational tional illustrative examples, we used spondents’ primary racial/ethnic iden- level had significantly lower mean in- SUDAAN software66 and the specified tification. The MIHA included Latino/ comes compared with white adults of data sources and samples to construct Hispanic as a separate racial/ethnic similar educational attainment (eg, 33% aseriesofweightedunadjustedand category; the BRFSS, NHIS, NHANES and 18% lower for those with 12 years multivariate logistic regression mod- III, and Add Health surveys included of schooling). This difference may re- els for each of 23 health-related indi- separate questions about race and His- flect not only unequal employment op- cators (listed as dependent variables in panic ethnicity, which we used to cre- portunities and rewards but varia- Table 1, counting obesity and smok- ate mutually exclusive racial/ethnic tions in educational quality.10 While ing as separate indicators in different groups. measuring credentials is generally pre- data sources or samples). These mod- ferred to years of schooling for the pur- els examined how observed associa- Critique of SES Measurement poses of reflecting SES,27,28,68,69 neither tions—expressed as odds ratios with Approaches in Health Research captures potentially dramatic differ- 95% confidence intervals —between ra- Education and Income: Not Inter- ences across schools in prestige or re- cial/ethnic groups and the health- changeable. Socioeconomic status is of- sources, which may contribute to dif- related indicators varied by the in- ten implicitly or explicitly equated with ferences in future earnings. TABLE 3 come and/or education measures (see income, especially in the United States. displays correlations between educa- Table 1 footnotes) used as covariables Despite wide recognition of its impor- tion and income (see Table 3 foot- to reflect SES; models also included age tance for health, income information is notes) overall and within racial/ethnic and sex (or parity, in MIHA), as well considered to be sensitive and is not groups for each data source or sample. as family structure (in Add Health). Ref- measured in many studies. Informa- Consistent with other studies,25,51,70,71 erence groups were those with the high- tion about education (typically mea- income-education correlations— est income or educational attainment. sured as years completed or creden- most less than 0.50—were not strong Log likelihood ratio tests were used to tials of formal schooling) is more easily enough to justify using income and edu- assess goodness of fit.67 Further de- obtained and is frequently treated as a cation as proxies for each other.

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

Table 3. Spearman Correlation Coefficients Between Poverty Level and Educational Level Overall and by Racial/Ethnic Group, by Data Source and Age Group* Data Source/Age Group

BRFSS, NHIS, NHANES III, 1988-1994 Add Health, MIHA, 2004, 1989-1994, 1994-1995, 1999-2004, Race/Ethnicity 18-74 y 18-64 y† 1-5 y Ͻ17 y 25-64 y 11-21 y Ͼ24y Black 0.45 0.42 0.44 0.42 0.42 0.45 0.59 Hispanic 0.50 Mexican American 0.44 0.46 0.52 0.50 0.40 Cuban 0.42 Puerto Rican 0.37 Asian/Pacific Islander Immigrant 0.48 US-born 0.47 Latina Immigrant 0.35 US-born 0.56 White 0.41 0.37 0.54 0.50 0.40 0.44 0.49 Overall 0.48 0.41 0.58 0.55 0.44 0.48 0.67 Abbreviations: Add Health, National Longitudinal Study of Adolescent Health; BRFSS, Behavioral Risk Factor Surveillance System; MIHA, California Maternal and Infant Health Assessment; NHANES III, Third National Health and Nutrition Examination Survey; NHIS, National Health Interview Survey. *Poverty level indicates annual income estimated in categories defined by 100% increments of the federal poverty level according to family or household size (0%-100%, 101%- 200%, 201%-300%, 301%-400%, Ͼ400%, unknown); educational level indicates completed education in levels according to earned credentials (Ͻ9 years, some high school, high school graduate/General Equivalency Diploma, some college, college graduate or more); for MIHA Ͻ9 years and some high school are combined; Add Health also includes a category for unknown education. †With mortality follow-up through 1997.

cluded. Comparing blacks and whites, Table 4. Odds Ratios for Racial/Ethnic Disparities in Fair or Poor Health Among Adults Aged 18-64 Years—National Health Interview Survey, 1989-1994 (n = 380 552) higher risks were seen for blacks re- OR (95% CI)† gardless of the income or education measure used; however, the relevant SES Measure in Model* Black Non-Hispanic Mexican American odds ratios were significantly smaller None (baseline) 2.47 (2.30-2.64) 2.16 (2.01-2.32) after adjustment for income or pov- Poverty level‡ 1.65 (1.56-1.76) 1.30 (1.22-1.39) erty level. Similarly, TABLE 5 reveals dis- Income, per $1000§ 1.67 (1.57-1.78) 1.44 (1.32-1.56) parities in receipt of delayed (after the Educational level ࿣ 1.96 (1.85-2.08) 1.06 (0.99-1.14) first trimester) or no prenatal care Education, per 1 y 2.02 (1.91-2.14) 0.89 (0.82-0.97) among women who gave birth in Cali- Poverty level and educational level 1.53 (1.45-1.61) 0.86 (0.81-0.93) fornia during 1999-2004; all models Abbreviations: CI, confidence interval; OR, odds ratio; SES, socioeconomic status. *All models included race/ethnicity, age, and sex. used MIHA data and included race/ †White non-Hispanic is the reference group. ‡Annual income estimated in categories defined by 100% increments of the federal poverty level according to family or ethnicity, age, and parity. In separate household size (0%-100%, 101%-200%, 201%-300%, 301%-400%, Ͼ400%, unknown). comparisons of black and immigrant §Annual income in continuous dollars estimated as the midpoints of a given income range. ࿣Completed education in levels according to earned credentials (Ͻ9 years, some high school, high school graduate/ Latina women with white women, sig- General Equivalency Diploma, some college, college graduate or more). nificant disparities in delayed or no prenatal care were seen when adjusting for To further illustrate practical impli- covariables (see Table 4 footnotes); all either education measure alone but not cations of treating education as an in- models included race/ethnicity, age, and when adjusting for either income mea- come proxy, we examined how con- sex. When income (measured continu- sure alone or together with education. clusions regarding racial/ethnic ously or grouped by poverty level) was Although differences often were not sta- disparities in health might vary depend- used to “adjust for SES,” Mexican tistically significant, one would have ing on whether income or education Americans appeared to be at signifi- reached different conclusions about the was used to measure SES. TABLE 4 dis- cantly increased risk of fair or poor significance, magnitude, or direction of plays results on racial/ethnic differ- health compared with non-Hispanic racial/ethnic disparities in 10 of the 23 ences in self-rated health among adults whites; however, risks appeared simi- health indicators we examined, depend- aged 18 to 64 years from 6 multivari- lar when educational level alone was in- ing on whether (or less frequently, how) ate models using NHIS data. The mod- cluded and significantly lower when income and education were mea- els differed by whether and how in- either years of education or both pov- sured. For 20 of the 23 health indica- come and education were included as erty and educational levels were in- tors, model fit was significantly better

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

when both income and education were home7,29,46,80 or car46,84 ownership or a been observed in the British civil ser- included, indicating that both factors single question on “liquid assets,”83 even vice hierarchy, with individuals in each should be considered (data available after controlling for income, another so- occupational grade experiencing worse from authors on request or at cioeconomic measure, or both.7,29,46,83,84 outcomes than those in the grade im- http://www.ucsf.edu/csdh). In summary, there are strong concep- mediately above,52 even after adjust- Both income and education can in- tual and empirical grounds for measuring for health-related behaviors2,17; no fluence the etiology of many health out- ing wealth in health studies and for con- group was poor, and financial access to comes, in part through pathways including that income is not an adequate medical care was unlikely to account volving material resources. Education measure of wealth. for the differences. One explanation is also can reflect a range of noneco- Inadequacy of Standard US Occu- variation in the psychosocial charac- nomic social characteristics (eg, gen- pational Categories. Occupational cat- teristics of one’s occupation,17 includ- eral and health-related knowledge, lit- egories based on prestige, skills, social ing control over one’s work.17,91-93 eracy, and problem-solving skills; influence, and/or power have been the Health studies in the United States prestige; influence over others and one’s primary basis for socioeconomic clas- rarely measure occupation, however. own life)68,72-76 with important health sification in western European coun- Occupational information is included in effects; thus, for many health out- tries. Studies have repeatedly found vital statistics and some national health comes, education should be consid- strong relationships between occupa- surveys, but these data were not in- ered in addition to—not instead of— tional status using such classifications tended—and do not appear to be mean- income, accumulated wealth, and other (eg, manual vs nonmanual labor, or ingful—as SES measures, because the more directly economic factors. graded hierarchies according to pres- categories include workers with di- Income: Not a Proxy for Wealth. Im- tige or skills) and diverse health indi- verse prestige, skills, power, and/or earn- portant links between economic re- cators,2,8,12,15-17,46,84-90 even after control- ings. For example, the 2000 Standard sources and health, operating through ling for car ownership84 or income and Occupational Classification System94 diverse causal pathways, are widely rec- education.48 Stepwise gradients in mor- classifies chief executive officers, town ognized. While income is the most com- tality and cardiovascular outcomes have clerks, and tenant farmers under “Man- monly used measure of economic resources in affluent countries, total accumulated economic resources or Table 5. Odds Ratios for Racial/Ethnic Disparities in Delayed or No Prenatal Care Among Childbearing Women Aged 25 Years and Older—California Maternal and Infant Health “wealth” could be at least as impor- Assessment, 1999-2004 (n = 13 952) tant for health; for example, wealth can OR (95% CI)† buffer the effects of temporarily low income due to unemployment or illness SES Measure in Model* Black Latina Immigrant and can reflect power or influence over None (baseline) 1.86 (1.53-2.26) 2.21 (1.90-2.57) others.77 Furthermore, wealth can vary Poverty level‡ 1.16 (0.95-1.43) 1.08 (0.91-1.28) dramatically across different social Income, per $1000‡ 1.12 (0.91-1.40) 1.05 (0.88-1.26) groups with similar incomes. For ex- Educational level§ 1.57 (1.29-1.92) 1.47 (1.22-1.76) Education, per 1 y 1.72 (1.41-2.10) 1.37 (1.12-1.67) ample, TABLE 6,basedon2000data from the Census Bureau, shows that, in Poverty level and educational level 1.13 (0.91-1.39) 0.98 (0.82-1.18) Abbreviations: CI, confidence interval; OR, odds ratio; SES, socioeconomic status. the lowest income quintile, house- *All models included race/ethnicity, age, and parity. †White is the reference group. holds headed by whites had on aver- ‡See Table 4 footnotes for definition. age more than 400 times as much §Completed education in levels according to earned credentials (Ͻ9 years/some high school, high school graduate/ wealth as those headed by blacks, and General Equivalency Diploma, some college, college graduate or more). that, in higher income quintiles, whites had approximately 3 to 9 times the Table 6. Median Net Worth in Dollars (Excluding Home Equity) by Quintile of Monthly wealth of blacks; these differences were Household Income and Householder’s Racial/Ethnic Group, 2000* statistically significant (see Table 6 Quintile Black† Hispanic‡ White Non-Hispanic 78 footnotes). Lowest 57 500 24 000 7,18,25,29,46,79-84 With notable exceptions, Second 5275 5670 48 500 few health studies in affluent coun- Third 11 500 11 200 59 500 tries have measured wealth. Like in- Fourth 32 600 36 225 92 842 come, wealth can be a sensitive topic, Highest 65 141 73 032 208 023 and standard methods of calculating net *From US Census Bureau Survey of Income and Program Participation.78 Standard errors are not reported by the source; however, both the comparisons of blacks with whites and of Hispanics with whites are stated to be significantly dif- worth can be laborious. Some studies ferent. have observed wealth effects on health †Report does not indicate whether blacks are Non-Hispanic. ‡Persons of Hispanic origin may be of any race. using simpler measures such as

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

lege-graduate women had been raised Figure 1. Percentage of College-Graduate Childbearing Women Aged 25 Years and Older With Ն1 College-Graduate Parent—California Maternal and Infant Health Assessment, by a college-graduate parent. Lack of 2003-2004 (n=1702) correspondence between educational attainment of women and their par- 70 ents was seen for women at all educa- 60 tional levels (data available on request 50 or at http://www.ucsf.edu/csdh). Given 1 Parent ≥ 40 the potentially important role that past socioeconomic experiences may play in 30 health outcomes, practical measures of 20 socioeconomic experiences earlier in

Percentage WithPercentage 10 life—eg, the highest educational attain-

Who Had Graduated From College Who Had Graduated From 0 ment of either parent/guardian when an Asian/Pacific Black Latina White All adult respondent was a child79,108— Islander (n = 151) (n = 226) (n = 939) (n = 1702) (n = 386) should be developed and tested in diverse populations. Women’s Racial/Ethnic Group Importance of Neighborhood So- Error bars indicate 95% confidence intervals. cioeconomic Conditions. Accumulat- ing evidence suggests that an individu- agement” and head chefs, waitresses, and act independently or modify the ef- al’s health can be influenced by the dishwashers under “Food Preparation fects of current factors on health.13,39,100 socioeconomic characteristics of the and Serving Related.” Several au- Changes in SES over time, including neighborhood in which she or he lives, thors25-28,53 have provided detailed cri- dramatic loss of income, can affect later above and beyond her or his own indi- tiques of occupational classifications. health,90,101,102 and poverty can have cu- vidual-level SES.34-37 Socioeconomic char- Some researchers have grouped the stan- mulative health effects.99,103,104 Differ- acteristics of neighborhoods could affect dard occupational classifications into ent socioeconomic factors could be health through features of the physical manual and nonmanual work catego- more or less important at different (“built”), social, or service environ- ries and found strong associations with stages of life.83 ments109-113 via multiple pathways.114 health outcomes.48 While this may be Standard practice for measuring SES People with similar individual- or house- worthwhile when better alternatives are in most health studies, however, is to hold-level socioeconomic characteris- unavailable, more meaningful occupa- include only measures of current or re- tics can live in very different local envi- tional classifications (eg, such as ques- cent socioeconomic characteristics; ronments. For example, as seen in tions from Karasek92,93 measuring con- apart from some occupational health FIGURE 2, Add Health data show that trol and demands at work) are needed studies, socioeconomic experiences poor black and Puerto Rican adoles- to capture information about impor- during earlier life stages are rarely mea- cents (in families with incomes at or be- tant differences in occupation-related sured. Socioeconomic status in child- low the poverty line) lived in neigh- prestige or power that could affect hood generally is related to SES in adult- borhoods with higher poverty health. Occupational measures, how- hood,105-107 but important SES concentrations than their poor white ever, present challenges for classifying characteristics earlier in life (eg, par- counterparts; mean differences were persons outside the paid labor force ents’ educational attainment) may not statistically significant and similar pat- (thus disproportionately affecting wom- be reflected in measures of current SES, terns of racial/ethnic differences in en) or the chronically unemployed. particularly for some population neighborhood conditions were appar- Importance of Past Socioeconomic groups. For example, when their own ent among higher-income adolescents Experiences. Childhood SES may in- parents’ education was examined (data available on request or at http: fluence adult health independently of among women older than 24 years who //www.ucsf.edu/csdh) and among na- adult SES.6,13,31-33,79,95-99 Acknowledg- gave birth in California during 2003- tionally representative samples of ing that the relative importance of SES 2004, only 50% of college-graduate adults115 and women of reproductive early in life varies with health out- women overall had been raised by a col- age.116 comes, Smith and Ben-Shlomo13 con- lege-graduate parent. In addition, this Despite increasing recognition that cluded that “studies with data on so- percentage varied significantly across both individual- and neighborhood- cioeconomic circumstances at only one racial/ethnic groups (for all compari- level SES can influence health, few stage of life are inadequate for fully elu- sons except between blacks and Lati- health studies measure neighborhood cidating the contribution of socioeco- nas); as shown in FIGURE 1,50%of features along with—rather than as nomic factors to health and mortality Asian/Pacific Islander, 34% of black, proxies for—individual-level SES mea- risk.” Past socioeconomic factors could 20% of Latina, and 58% of white col- sures. In practice, however, many re-

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

searchers could consider characteris- an even greater extent than simpler mea- Figure 2. Distribution of Neighborhood tics of both individuals and their sures, may not apply over time and across Poverty Among Poor Adolescents by neighborhoods, despite theoretical and populations.26,28 Furthermore, such com- Racial/Ethnic Group—National Longitudinal methodological challenges.117-119 By posite measures, while potentially use- Study of Adolescent Health, 1994-1995 linking residential addresses with cen- ful for classification in some studies, do –1 SD +1 SD sus geographic codes such as census not permit study of how particular SES From Mean Median From Mean tracts, census variables (eg, percent- factors influence health. Health research- ages of poor households or of unem- ers should justify the particular socio- ployed adults) can be used to describe economic measures they have studied, Racial/Ethnic Group neighborhoods; similarly, geographic avoiding claims to have measured SES Black coordinates (longitude/latitude) can be overall. Cuban linked with various data sources to de- Standard measures may not reflect im- Mexican American scribe physical environments in ways portant and relevant aspects of SES. Even Puerto Rican White that are relevant for health research. studies that include multiple standard 0 10 20 30 40 SES measures cannot examine all poten- Neighborhood Poverty, % Conclusions and tially important socioeconomic influ- Recommendations ences on health. Years of schooling re- Neighborhood poverty defined as percentage of families in the census tract who had 1989 income that was Based on the empirical evidence from ceived or earned credentials may not below the federal poverty level; poor adolescents de- new analyses and from the literature, capture significant differences in educa- fined as those aged 11-21 years and living in families with 1994 income that was at or below the federal we have reached 5 general conclu- tional quality that may also be relevant. poverty level. sions, with corresponding recommen- Conventional US occupational group- dations for improving the measure- ings were not designed and are unlikely ment and interpretation of SES in health to adequately measure job-related socio- serve racial/ethnic disparities in health studies. economic characteristics that may affect outcomes should explicitly acknowl- Different socioeconomic measures health. Subjective SES120 and experi- edge the plausible role of unmeasured 121 aspects of SES and other potentially rel- cannot be assumed to be interchange- ences of economic hardship could 55 able. The generally modest correla- plausibly influence health through psy- evant explanations, including insti- tions between income and education chophysiological pathways discussed ear- tutional or personal experiences of dis- indicate that measures of these 2 socio- lier but are not explicitly reflected in stan- crimination. We believe that economic factors are not interchange- dard SES measures. Rather than claiming conclusions from prior research about able; this is further supported by the to have “controlled for SES,” research- the nature of racial/ethnic (and other examples showing that associations ers should acknowledge the potentially social) differences in health should be between racial/ethnic groups and health relevant aspects of SES that could not be thoughtfully reassessed in light of find- indicators can depend on which socio- measured and explicitly consider the im- ings and recommendations in this ar- * economic measures are used as covari- plications of unmeasured socioeco- ticle as well as prior literature. ables. Evidence of racial/ethnic differ- nomic influences when interpreting find- A given SES measure may have dif- ences in income at a given level of ings.70 ferent meanings in different social education, in wealth at a given level of Racial/ethnic differences are likely to groups. Although our examples fo- income, in past SES at a given level of reflect unmeasured socioeconomic dif- cused on racial/ethnic differences, simi- current SES, and in neighborhood SES ferences. The concerns expressed above lar issues may apply to other social at a given level of individual SES also sup- underscore the fact that—without mea- groups categorized, for example, by age,83 sex,50,52,127 or urban/rural loca- ports the noninterchangeability of these suring all relevant SES dimensions, life 128 variables across diverse populations. stages, and aggregation levels (eg, in- tion. Whenever possible, research- Multiple socioeconomic measures often dividual, household, neighborhood, ers should examine how findings us- can be included simultaneously in mul- city)—an observed racial/ethnic dis- ing a given SES measure vary across tivariable models without colinearity parity in health cannot be considered social groups. problems20-22,61-64,70,71;stratifiedanaly- “independent of SES.”122 However, ra- Measures of SES should be selected ses also should be considered. Compos- cial/ethnic differences also cannot be as- and interpreted thoughtfully in the con- ite SES measures, or “indices,” also have sumed to be reducible to socioeco- text of plausible explanatory path- been used to reflect multiple socioeconomic issues; for example, systematic ways through which socioeconomic fac- nomic factors. However, few of the indi- socioeconomic differences between ra- tors may influence health. Researchers vidual- or household-level (distin- cial/ethnic groups can reflect racial dis- should select socioeconomic factors sys- guished from community-level) indices crimination at the institutional/ tematically, considering whether eco-

have been validated. Most involve mul- structural level, personal experience, or *References 10, 27, 54, 55, 70, 114, 122, 123, 125, tiple questionable assumptions and, to both.10,54,123,124 Researchers who ob- 126.

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

nomic resources, education, occupa- ethnic group,55 calls for a careful reas- sity of California, San Francisco, which supported her as a Kellogg Health Disparities Scholar. tion, socioeconomic factors earlier in sessment of conclusions about the etio- Role of the Sponsors: None of the funding sources life, and neighborhood socioeco- logic basis of racial/ethnic differences for this study had any role in the design or conduct of the study; the collection, management, analysis, or in- nomic conditions at any life stage could in health based on studies with lim- terpretation of the data; or the preparation of the plausibly be relevant to the particular ited socioeconomic information. manuscript. Anonymous reviewers at the Centers for health outcome and population of in- These findings highlight the chal- Disease Control and Prevention reviewed and ap- proved the technical quality of the manuscript. The terest. They should assess the feasibil- lenges in capturing the multidimen- role of the California Department of Health Services ity of adequately measuring each po- sional nature of SES and the limita- in the collection of the data from the MIHA is noted above. tentially relevant factor using available tions of standard measures. Should we Acknowledgment: We acknowledge MCAH research- data; when a relevant factor cannot be despair of measuring SES adequately? ers Shabbir Ahmad, DVM, MPH, Eugene Takahashi, PhD, and Moreen Libet, PhD, for their contributions measured, the implications of its ex- We think not. Better measures are to the design and conduct of MIHA. We thank Nicole clusion should be considered thought- needed, along with work to develop and Wojtal and Pinal Shah, Department of Family and Com- munity Medicine, University of California, San Fran- fully and acknowledged. If a theoreti- validate measures and measurement ap- cisco (UCSF), for their assistance with research and pre- cally plausible argument or evidence proaches that are as comparable across paring the manuscript, and Roberto Vargas, MD, MPH, from previous studies indicates that a studies and populations as possible. Department of Medicine, University of California, Los Angeles, Craig Pollack, MD, Department of Medi- particular socioeconomic factor is un- However, health research could be im- cine, UCSF, and Thomas Bodenheimer, MD, Depart- likely to be relevant for the given out- proved significantly with a more con- ment of Family and Community Medicine, UCSF, for come or population of interest, it may ceptually and empirically sound ap- helpful comments on drafts. not be necessary to include that fac- proach to measurement of SES. Building tor. However, the approach we recom- on our own and others’ work, we call REFERENCES mend would shift the burden to the re- for a fundamentally different concep- 1. Isaacs SL, Schroeder SA. Class—the ignored de- terminant of the nation’s health. NEnglJMed.2004; searcher to justify why a study does not tual approach to measuring SES in 351:1137-1142. include measures of economic re- health studies. This approach is out- 2. Marmot MG, Rose G, Shipley M, Hamilton PJ. Em- sources, education, occupation, past come- and social group–specific and ployment grade and coronary heart disease in British civil servants. JEpidemiolCommunityHealth.1978; SES, and neighborhood socioeco- rests on considering explanatory path- 32:244-249. nomic conditions and to discuss fully ways and mechanisms, measuring as 3. Adler NE, Boyce WT, Chesney MA, Folkman S, Syme SL. Socioeconomic inequalities in health: no easy the implications of unmeasured fac- much relevant socioeconomic infor- solution. JAMA.1993;269:3140-3145. tors when stating conclusions. mation as possible, claiming to mea- 4. Kaplan GA. People and places: contrasting per- spectives on the association between social class and sure only what was measured, and sys- health. Int J Health Serv.1996;26:507-519. Final Comments tematically considering how important 5. Backlund E, Sorlie PD, Johnson NJ. The shape of the relationship between income and mortality in the We have presented conceptual argu- unmeasured socioeconomic factors may United States: evidence from the National Longitu- ments and empirical examples from di- affect conclusions. dinal Mortality Study. Ann Epidemiol.1996;6:12-20. verse data sources, populations, and 6. Lynch JW, Kaplan GA, Salonen JT. Why do poor Author Contributions: Drs Braveman, Cubbin, and people behave poorly? variations in adult health be- health-related indicators to illustrate Egerter and Ms Marchi had full access to all of the data haviour and psychosocial characteristics, by stage of several major and pervasive problems in the study and take responsibility for the integrity the socioeconomic life course. Soc Sci Med.1997;44: of the data and the accuracy of the data analysis. 809-820. with the standard ways in which SES Study concept and design: Braveman, Cubbin, Egerter, 7. Macintyre S, Ellaway A, Der G, Ford G, Hunt K. is measured in health research. The ex- Metzler. Do housing tenure and car access predict health be- Acquisition of data: Braveman, Cubbin, Chideya, cause they are simply markers of income or self es- amples presented here highlight the po- Marchi. teem? a Scottish study. JEpidemiolCommunityHealth. tential consequences of inadequately Analysis and interpretation of data: Braveman, Cubbin, 1998;52:657-664. measuring socioeconomic factors, par- Egerter, Chideya, Marchi, Metzler, Posner. 8. Marmot MG, Smith GD, Stansfeld S, et al. Health Drafting of the manuscript: Braveman, Cubbin, Egerter. inequalities among British civil servants: the White- ticularly in studies of racial/ethnic dis- Critical revision of the manuscript for important in- hall II study. Lancet.1991;337:1387-1393. parities; similar caveats apply to stud- tellectual content: Braveman, Cubbin, Egerter, Chideya, 9. Marmot M, Ryff CD, Bumpass LL, Shipley M, Marks Marchi, Metzler, Posner. NF. Social inequalities in health: next questions and ies of other social disparities, eg, by age, Statistical analysis: Cubbin, Marchi. converging evidence. Soc Sci Med.1997;44:901-910. sex, or geographic residence. Our aim Obtained funding: Braveman, Cubbin, Egerter, Marchi. 10. Williams DR. Race, socioeconomic status, and Administrative, technical, or material support: Cubbin, health: the added effects of racism and discrimination. has been to provide convincing evi- Chideya, Marchi, Metzler, Posner. Ann N Y Acad Sci.1999;896:173-188. dence for a wide audience of health re- Study supervision: Braveman, Cubbin. 11. Avendano M, Kunst AE, van Lenthe F, et al. Trends Financial Disclosures: None reported. in socioeconomic disparities in stroke mortality in six searchers and users of health research Funding/Support: The research for this article was sup- European countries between 1981-1985 and findings. While previous authors have ported by the Division of Reproductive Health, Co- 1991-1995. Am J Epidemiol.2005;161:52-61. written about many of these issues, the ordinating Center for Health Promotion, Centers for 12. Black D, Morris JN, Smith C, Townsend P. The Black Disease Control and Prevention (CDC-ATPM Coop- Report. In: Townsend P, Davidson N, Whitehead M, messages apparently have not reached erative Agreement TS-0842). One of the data sources eds. Inequalities in Health: The Black Report/The most clinical and public health re- (Maternal and Infant Health Assessment, MIHA) is a Health Divide.2nded.London,England:Penguin; collaborative effort of the authors (P.A.B., S.E., K.S.M.) 1999:29-213. searchers and policy makers. In par- with the California Department of Health Services Ma- 13. Smith GD, Ben-Shlomo Y. Geographical and so- ticular, the evidence presented here, re- ternal and Child Health Branch (MCAH), supported cial class differentials in stroke mortality—the influ- by MCAH (see Acknowledgment). Dr Chideya’s par- ence of early-life factors: comment on papers by Ma- inforced by the discussion in a recent ticipation in this effort was funded by the Depart- heswaren and colleagues. JEpidemiolCommunity JAMA Editorial on measuring racial/ ment of Family and Community Medicine, Univer- Health.1997;51:134-137.

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

14. Stansfeld SA, Head J, Marmot MG. Explaining so- 33. Galobardes B, Lynch JW, Smith GD. Childhood 55. Winker MA. Measuring race and ethnicity: why cial class differences in depression and well-being. Soc socioeconomic circumstances and cause-specific mor- and how? JAMA.2004;292:1612-1614. Psychiatry Psychiatr Epidemiol.1998;33:1-9. tality in adulthood: systematic review and 56. Centers for Disease Control and Prevention. Be- 15. Mackenbach JP, Kunst AE, Groenhof F, et al. So- interpretation. Epidemiol Rev.2004;26:7-21. havior Risk Factor Surveillance System (BRFSS). 2000. cioeconomic inequalities in mortality among women 34. Yen IH, Syme SL. The social environment and Available at: http://www.cdc.gov/brfss. Accessibil- and among men: an international study. Am J Public health: a discussion of the epidemiologic literature. ity verified October 24, 2005. Health.1999;89:1800-1806. Annu Rev Public Health.1999;20:287-308. 57. Centers for Disease Control and Prevention. Na- 16. Kunst AE, Groenhof F, Mackenbach JP, Work- 35. Robert SA. Socioeconomic position and health: tional Health Interview Survey (NHIS). Available at: ing EU; Group on Socioeconomic Inequalities in Health. the independent contribution of community socio- http://www.cdc.gov/nchs/nhis.htm. Accessibility veri- Occupational class and cause specific mortality in economic context. Annu Rev Sociol.1999;25:489-516. fied October 24, 2005. middle aged men in 11 European countries: compari- 36. Pickett KE, Pearl M. Multilevel analyses of neigh- 58. Centers for Disease Control and Prevention. Third son of population-based studies. BMJ.1998;316:1636- bourhood socioeconomic context and health out- National Health and Nutrition Examination Survey 1642. comes: a critical review. JEpidemiolCommunity (NHANES III), 1988-1994, Public Use Data Files. Avail- 17. Marmot MG, Bosma H, Hemingway H, Brunner Health.2001;55:111-122. able at: http://www.cdc.gov/nchs/products E, Stansfeld S. Contribution of job control and other 37. Sampson RJ, Morenoff JD, Gannon-Rowley T. As- /elec_prods/subject/nhanes3.htm. Accessibility veri- risk factors to social variations in coronary heart dis- sessing “neighborhood effects”: social processes and fied October 24, 2005. ease incidence. Lancet.1997;350:235-239. new directions in research. Annu Rev Sociol.2002;28: 59. National Institutes of Health. National Longitu- 18. Benzeval M, Judge K, Shouls S. Understanding the 443-478. dinal Study of Adolescent Health (Add Health), relationship between income and health: how much 38. Marmot M. Inequalities in health. NEnglJMed. 1994-1995. Available at: http://www.nichd.nih.gov can be gleaned from cross-sectional data? Soc Policy 2001;345:134-136. /about/cpr/dbs/res_add.htm. Accessibility verified Oc- Adm.2001;35:376-396. 39. Lynch JW, Everson SA, Kaplan GA, Salonen R, Sa- tober 24, 2005. 19. Mackenbach JP, Martikainen P, Looman CW, Dal- lonen JT. Does low socioeconomic status potentiate 60. Maternal and Infant Health Assessment (MIHA). stra JA, Kunst AE, Lahelma E. The shape of the rela- the effects of heightened cardiovascular responses to Available at: http://www.ucsf.edu/csdh/reseach tionship between income and self-assessed health: stress on the progression of carotid atherosclerosis? _projects.htm. Accessibility verified October 24, an international study. Int J Epidemiol.2005;34:286- Am J Public Health.1998;88:389-394. 2005. 293. 40. Taylor SE, Repetti RL, Seeman T. Health psychol- 61. Cubbin C, Braveman P, Marchi K, Chavez G, San- 20. Lantz PM, House JS, Lepkowski JM, Williams DR, ogy: what is an unhealthy environment and how does telli J, Gilbert BJ. Socioeconomic and racial/ethnic dis- Mero RP, Chen J. Socioeconomic factors, health be- it get under the skin? Annu Rev Psychol.1997;48:411- parities in unintended pregnancy among postpartum haviors, and mortality: results from a nationally rep- 447. women in California. Matern Child Health J.2002;6: resentative prospective study of US adults. JAMA. 41. Krieger N. Theories for social epidemiology in the 237-246. 1998;279:1703-1708. 21st century: an ecosocial perspective. Int J Epidemiol. 62. Galbraith AA, Egerter SA, Marchi KS, Chavez G, 21. Lannin DR, Mathews HF, Mitchell J, Swanson MS, 2001;30:668-677. Braveman PA. Newborn early discharge revisited: are Swanson FH, Edwards MS. Influence of socioeco- 42. Singer B, Ryff C. Racial and ethnic inequalities in California newborns receiving recommended postna- nomic and cultural factors on racial differences in late- health: environmental, psychological, and physiologi- tal services? Pediatrics.2003;111:364-371. stage presentation of breast cancer. JAMA.1998;279: cal pathways. In: Devlin B, Fienberg SE, Resnick DP, 63. Chung E, Hung Y, Marchi K, Chavez G, Brave- 1801-1807. Roeder K, eds. Intelligence, Genes, and Success: Sci- man P. Infant sleep position: associated maternal and 22. Lowry R, Kann L, Collins JL, Kolbe LJ. The effect entists Respond to the Bell Curve.NewYork,NY:Co- infant factors. Ambul Pediatr.2003;3:234-239. of socioeconomic status on chronic disease risk be- pernicus; 1997:89-122. 64. Heck KE, Schoendorf KC, Chavez GF, Braveman haviors among US adolescents. JAMA.1996;276:792- 43. Wilkinson RG. Socioeconomic determinants of P. Does postpartum length of stay affect breastfeed- 797. health: health inequalities: relative or absolute mate- ing duration? a population-based study. Birth.2003; 23. Weber M. Classes, status groups and parties. In: rial standards? BMJ.1997;314:591-595. 30:153-159. Runciman WG, ed. Max Weber: Selections in Trans- 44. Steptoe A, Marmot M. The role of psychobio- 65. SAS.Version9.1.Cary,NC:SASInstituteInc;2002. lation.Cambridge,England:CambridgeUniversity logical pathways in socio-economic inequalities in car- 66. SUDAAN.Version9.0.1.ResearchTrianglePark, Press; 1978:43-61. diovascular disease risk. Eur Heart J.2002;23:13-25. NC: Research Triangle Institute; 2005. 24. Wright EO. Review of Coleman RP, Rainwater L, 45. McEwen BS. Protective and damaging effects of 67. McCullagh P, Nelder JA. Generalized Linear McClelland KA. Social standing in America—new di- stress mediators. NEnglJMed.1998;338:171-179. Models.London,England:Chapman&Hall;1989. mensions of class. Am J Sociol.1980;85:1443-1449. 46. Arber S, Ginn J. Gender and inequalities in health Monographs on Statistics and Applied Probability.2nd 25. Abramson JH, Gofin R, Habib J, Pridan H, Gofin in later life. Soc Sci Med.1993;36:33-46. ed. Vol 37. J. Indicators of social class: a comparative appraisal of 47. Nicolaidis C, Ko CW, Saha S, Koepsell TD. Racial 68. Ross CE, Mirowsky J. Refining the association be- measures for use in epidemiological studies. Soc Sci discrepancies in the association between paternal vs. tween education and health: the effects of quantity, Med.1982;16:1739-1746. maternal educational level and risk of low birth- credential, and selectivity. Demography.1999;36:445- 26. Berkman LF, Macintyre S. The measurement of weight in Washington State. BMC Pregnancy 460. social class in health studies: old measures and new Childbirth.2004;4:10. 69. Hadden WC. Annotation: the use of educational formulations. In: Kogevinas M, Pearce N, Susser M, 48. Barbeau EM, Krieger N, Soobader MJ. Working attainment as an indicator of socioeconomic position. Boffetta P, eds. Social Inequalities and Cancer.Lyon, class matters: socioeconomic disadvantage, race/ Am J Public Health.1996;86:1525-1526. France: International Agency for Research on Can- ethnicity, gender, and smoking in NHIS 2000. Am J 70. Braveman P, Cubbin C, Marchi K, Egerter S, Chavez cer; 1997:51-64. IARC Scientific Publications No. 138. Public Health.2004;94:269-278. G. Measuring socioeconomic status/position in stud- 27. Krieger N, Williams DR, Moss NE. Measuring so- 49. Pearl M, Braveman P, Abrams B. Relationship of ies of racial/ethnic disparities: maternal and infant cial class in US public health research: concepts, meth- neighborhood socioeconomic characteristics to birth health. Public Health Rep.2001;116:449-463. odologies, and guidelines. Annu Rev Public Health. weight among 5 ethnic groups in California. Am J Pub- 71. Winkleby MA, Jatulis DE, Frank E, Fortmann SP. 1997;18:341-378. lic Health.2001;91:1808-1814. Socioeconomic status and health: how education, in- 28. Liberatos P, Link BG, Kelsey JL. The measure- 50. Stafford M, Cummins S, Macintyre S, Ellaway A, come, and occupation contribute to risk factors for car- ment of social class in epidemiology. Epidemiol Rev. Marmot M. Gender differences in the associations be- diovascular disease. Am J Public Health.1992;82:816- 1988;10:87-121. tween health and neighbourhood environment. Soc 820. 29. Sundquist J, Johansson SE. Indicators of socio- Sci Med.2005;60:1681-1692. 72. Winch C. The economic aims of education. JPhi- economic position and their relation to mortality in 51. Gazmararian JA, Adams MM, Pamuk ER. Asso- los Educ.2002;36:101-117. Sweden. Soc Sci Med.1997;45:1757-1766. ciations between measures of socioeconomic status 73. Mirowsky J, Ross CE. Education, personal con- 30. Diez Roux AV, Merkin SS, Arnett D, et al. Neigh- and maternal health behavior. Am J Prev Med.1996; trol, lifestyle and health: a human capital hypothesis. borhood of residence and incidence of coronary heart 12:108-115. Res Aging.1998;20:415-449. disease. NEnglJMed.2001;345:99-106. 52. Griffin JM, Fuhrer R, Stansfeld SA, Marmot M. 74. Reynolds JR, Ross CE. Social stratification and 31. Smith GD, Hart C, Blane D, Hole D. Adverse so- The importance of low control at work and home on health: education’s benefit beyond economic status cioeconomic conditions in childhood and cause spe- depression and anxiety: do these effects vary by gen- and social origins. Soc Probl.1998;45:221-247. cific adult mortality: prospective observational study. der and social class? Soc Sci Med.2002;54:783-798. 75. Ross CE, Van Willigen M. Education and the sub- BMJ.1998;316:1631-1635. 53. Krieger N, Fee E. Social class: the missing link in jective quality of life. JHealthSocBehav.1997;38:275- 32. Rahkonen O, Lahelma E, Huuhka M. Past or pre- U.S. health data. Int J Health Serv.1994;24:25-44. 297. sent? childhood living conditions and current socio- 54. Williams DR. Race/ethnicity and socioeconomic 76. Ross CE, Wu C. The links between education and economic status as determinants of adult health. Soc status: measurement and methodological issues. Int health. Am Sociol Rev.1995;60:719-745. Sci Med.1997;44:327-336. JHealthServ.1996;26:483-505. 77. Brenner MH. Health, productivity, and the eco-

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014 SOCIOECONOMIC STATUS IN HEALTH RESEARCH

nomic environment: dynamic role of socioeconomic 94. US Census Bureau. 2000 Standard Occupational 112. Macintyre S, Ellaway A, Cummins S. Place ef- status. In: Green GM, Baker F, eds. Work, Health and Classification System. Available at: http://www.census fects on health: how can we conceptualise, opera- Productivity.NewYork,NY:OxfordUniversityPress; .gov/hhes/www/ioindex/view.html. Accessibility veri- tionalise and measure them? Soc Sci Med.2002;55: 1991:241-255. fied October 24, 2005. 125-139. 78. US Census Bureau. Net Worth and Asset Own- 95. Blane D, Hart CL, Smith GD, et al. Association of 113. Morrison RS, Wallenstein S, Natale DK, Senzel ership of Households: 1998 and 2000.Washington, cardiovascular disease risk factors with socioeco- RS, Huang L-L. “We don’t carry that”—failure of phar- DC: US Census Bureau; May 2003. Current Popula- nomic position during childhood and during adulthood. macies in predominantly nonwhite neighborhoods to tion Reports P70-88. BMJ.1996;313:1434-1438. stock opioid analgesics. NEnglJMed.2000;342:1023- 79. Cohen S, Doyle WJ, Turner RB, Alper CM, Skoner 96. Kuh D, Ben-Shlomo Y, eds. ALifeCourseAp- 1026. DP. Childhood socioeconomic status and host resis- proach to Chronic Disease Epidemiology.NewYork, 114. Massey DS. Segregation and stratification: a bio- tance to infectious illness in adulthood. Psychosom NY: Oxford University Press; 1997. social perspective. The DuBois Review: Social Sci- Med.2004;66:553-558. 97. van de Mheen H, Stronks K, Looman CWN, Mack- ence Research on Race.2004;1:1-19. 80. Macintyre S, McKay L, Der G, Hiscock R. Socio- enbach JP. Role of childhood health in the explana- 115. Cubbin C, Hadden WC, Winkleby MA. Neigh- economic position and health: what you observe de- tion of socioeconomic inequalities in early adult health. borhood context and cardiovascular disease risk fac- pends on how you measure it. JPublicHealthMed. JEpidemiolCommunityHealth.1998;52:15-19. tors: the contribution of material deprivation. Ethn Dis. 2003;25:288-294. 98. Bosma H, van de Mheen HD, Mackenbach JP. So- 2001;11:687-700. 81. Smith GD, Shipley MJ, Rose G. Magnitude and cial class in childhood and general health in adult- 116. Brewster KL. Race differences in sexual activity causes of socioeconomic differentials in mortality: fur- hood: questionnaire study of contribution of psycho- among adolescent women: the role of neighborhood ther evidence from the Whitehall study. JEpidemiol logical attributes. BMJ.1999;318:18-22. characteristics. Am Sociol Rev.1994;59:408-424. Community Health.1990;44:265-270. 99. Power C, Matthews S, Manor O. Inequalities in 117. Diez Roux AV. Invited commentary: places, 82. Von dem Knesebeck O, Luschen G, Cockerham self-rated health: explanations from different stages people, and health. Am J Epidemiol.2002;155:516- WC, Siegrist J. Socioeconomic status and health among of life. Lancet.1998;351:1009-1014. 519. the aged in the United States and Germany: a com- 100. Frankel S, Elwood P, Sweetnam P, Yarnell J, Smith 118. Diez-Roux AV. Bringing context back into epi- parative cross-sectional study. Soc Sci Med.2003;57: GD. Birthweight, body-mass index in middle age, and demiology: variables and fallacies in multilevel analysis. 1643-1652. incident coronary heart disease. Lancet.1996;348: Am J Public Health.1998;88:216-222. 83. Robert S, House JS. SES differentials in health by 1478-1480. 119. Tienda M. Poor people, poor places: decipher- age and alternative indicators of SES. JAgingHealth. 101. Duncan G. Income dynamics and health. Int J ing neighborhood effects on poverty outcomes. In: 1996;8:359-388. Health Serv.1996;26:419-444. Huber J, ed. Macro-micro Linkages in Sociology.New- 84. Smith GD, Bartley M, Blane D. The Black Report 102. McDonough P, Duncan G, Williams D, House bury Park, Calif: Sage; 1991:244-262. on socioeconomic inequalities in health 10 years on. J. Income dynamics and adult mortality in the United 120. Singh-Manoux A, Adler NE, Marmot MG. Sub- BMJ.1990;301:373-377. States, 1972 through 1989. Am J Public Health.1997; jective social status: its determinants and its associa- 85. Chandola T. Social inequality in coronary heart 87:1476-1483. tion with measures of ill-health in the Whitehall II study. disease: a comparison of occupational classifications. 103. Korenman S, Miller JE, Sjaastad JE. Long-term Soc Sci Med.2003;56:1321-1333. Soc Sci Med.1998;47:525-533. poverty and child development in the United States: 121. Mayer SE, Jencks C. Poverty and the distribu- 86. Diderichsen F, Hallqvist J. Trends in occupa- results from the NLSY. Child Youth Serv Rev.1995;17: tion of material hardship. JHumResour.1989;24:88- tional mortality among middle-aged men in Sweden 127-155. 114. 1961-1990. Int J Epidemiol.1997;26:782-787. 104. Miller JE, Korenman S. Poverty and children’s nu- 122. Kaufman JS, Cooper RS, McGee DL. Socioeco- 87. Vagero D, Erikson R. Socioeconomic inequalities tritional status in the United States. Am J Epidemiol. nomic status and health in blacks and whites: the prob- in morbidity and mortality in western Europe. Lancet. 1994;140:233-242. lem of residual confounding and the resiliency of race. 1997;350:516. 105. Duncan GJ, Yeung WJ, Brooks-Gunn J, Smith Epidemiology.1997;8:621-628. 88. Macintyre S. The patterning of health by social JR. How much does childhood poverty affect the life 123. Williams DR, Collins C. Racial residential segre- position in contemporary Britain: directions for socio- chances of children? Am Sociol Rev.1998;63:406-423. gation: a fundamental cause of racial disparities in logical research. Soc Sci Med.1986;23:393-415. 106. Warren JR, Hauser RM. Social stratification across health. Public Health Rep.2001;116:404-416. 89. Pamuk ER. Social-class inequality in infant mor- three generations: new evidence from the Wisconsin 124. Muntaner C, Nieto FJ, O’Campo P. Additional tality in England and Wales from 1921 to 1980. Eur J longitudinal study. Am Sociol Rev.1997;62:561-572. clarification re: “The Bell curve: on race, social class, Popul.1988;4:1-21. 107. Rodgers JR. Intergenerational transmission of and epidemiologic research.” Am J Epidemiol.1997; 90. Basso O, Olsen J, Johansen AMT, Christensen K. poverty in the United States. Soc Sci Q.1995;76:178- 146:607-608. Change in social status and risk of low birth weight in 194. 125. Bamshad M. Genetic influences on health: does Denmark: population based cohort study. BMJ.1997; 108. US Department of Labor. NLSY79: National Lon- race matter [published correction appears in JAMA. 315:1498-1502. gitudinal Survey of Youth, 1979. Available at: http: 2005;294:1620]? JAMA.2005;294:937-946. 91. Theorell T, Tsutsumi A, Hallquist J. Decision lati- //www.bls.gov/nls/nlsy79.htm. Accessibility veri- 126. Sankar P, Cho MK, Condit CM, et al. Genetic tude, job strain, and myocardial infarction: a study of fied October 24, 2005. research and health disparities. JAMA.2004;291:2985- working men in Stockholm. Am J Public Health.1998; 109. Morland K, Wing S, Diez Roux A. The contex- 2989. 88:382-388. tual effect of the local food environment on resi- 127. American Academy of Pediatrics; Committee on 92. Karasek R, Brisson C, Kawakami N, Houtman I, dents’ diets: the atherosclerosis risk in communities Pediatric Research. Race/ethnicity, gender, socioeco- Bongers P, Amick B. The Job Content Questionnaire study. Am J Public Health.2002;92:1761-1767. nomic status—research exploring their effects on child (JCQ): an instrument for internationally comparative 110. Macintyre S. The social patterning of exercise health: a subject review. Pediatrics.2000;105:1349- assessments of psychosocial job characteristics. JOc- behaviours: the role of personal and local resources. 1351. cup Health Psychol.1998;3:322-355. Br J Sports Med.2000;34:6. 128. Gilthorpe MS, Wilson RC. Rural/urban differ- 93. Karasek R. Lower health risk with increased job 111. Sampson RJ, Raudenbush SW, Earls F. Neigh- ences in the association between deprivation and control among white collar workers. JOrganBehav. borhoods and violent crime: a multilevel study of col- healthcare utilisation. Soc Sci Med.2003;57:2055- 1990;11:171-185. lective efficacy. Science.1997;277:918-924. 2063.

Downloaded From: http://jama.jamanetwork.com/ by a Portland State University User on 10/30/2014