⏐ RESEARCH AND PRACTICE ⏐

Hispanic Paradox in Biological Risk Profiles

| Eileen M. Crimmins, PhD, Jung Ki Kim, PhD, Dawn E. Alley, PhD, Arun Karlamangla, MD, PhD, Teresa Seeman, PhD

Many studies report that in the Objectives. We examined biological risk profiles by race, ethnicity, and nativ- have better or similar health to ity to evaluate evidence for a paradox in measured health indicators. that of non-Hispanic Whites (hereafter referred Methods. We used data on adults aged 40 years and older (n=4206) from the to as Whites), despite Hispanics having lower National Health and Examination Surveys (1999–2002) to compare blood incomes and less education.1,2 Most studies pressure, metabolic, and inflammatory risk profiles for Whites, Blacks, US-born that examine differences in adult mortality find and foreign-born Hispanics, and Hispanics of Mexican origin. We controlled for that Hispanics have relatively lower mortality age, gender, and . rates compared with Whites.2–6 This better- Results. Hispanics have more risk factors above clinical risk levels than do than-expected health and mortality of Hispan- Whites but fewer than Blacks. Differences between Hispanics and Whites disap- ics, given their lower socioeconomic status peared after we controlled for socioeconomic status, but results differed by na- tivity. After we controlled for socioeconomic status, the differences between (SES), has been called the .6,7 foreign-born Hispanics and Whites were eliminated, but US-born Mexican Amer- Not all empirical findings support the exis- icans still had higher biological risk scores than did both Whites and foreign- tence of a Hispanic paradox. Differentials de- born . pend on the domain of health and the popula- Conclusions. There is no Hispanic paradox in biological risk profiles. However, tion investigated. Evidence for a Hispanic our finding that foreign-born Hispanics and Whites had similar biological risk mortality advantage is strongest among men, profiles, but US-born Mexican Americans had higher risk, was consistent with hy- persons of advanced age, and those born in pothesized effects of migrant health selectivity (healthy people in-migrating and Mexico.6 However, some studies have found unhealthy people out-migrating) as well as some differences in health behaviors no difference in mortality between Hispanics between US-born and foreign-born Hispanics. (Am J Public Health. 2007;97: and Whites,8 and others have questioned the 1305–1310. doi:10.2105/AJPH.2006.091892) data quality in estimates of mortality among Hispanics.9,10 “salmon hypothesis,” which suggests that sick which physiological systems accounted for Ethnic differences are even less clear-cut in persons return to their place of origin.16 ,17 overall differentials by race, ethnicity, and na- analyses of function, disability,11,12 and morbid- Palloni and Arias16 emphasized this second tivity. If differentials were concentrated among ity.3 Self-reports of health status may be influ- explanation for their finding that the mortality 1 or 2 sets of indicators rather than spread enced by cultural differences in reporting or dif- advantage is limited to foreign-born Hispanics, across categories, this would provide another ferences in health knowledge acquired through particularly those who were born in Mexico. indicator of how health differences arise. interaction with the medical system.13 Analyses Both of these explanations for the Hispanic Risk profiles based on multiple factors are of self-reported health status usually find that paradox imply that the Hispanic health advan- useful in the analyses of a variety of health Hispanics report worse health than Whites.1,14 tage is a feature exclusive to foreign-born His- outcomes.25–27 The total number of indicators Researchers have argued recently that the panics, rather than US-born Hispanics. of physiological status outside the normal oper- migrants who immigrate are different from Measurements of biological risk factors for ating range has been shown to be a better pre- persons from the same country of origin who poor health (e.g., blood pressure, blood glucose, dictor of health outcomes than individual do not migrate may also play a large role in and cholesterol) should provide objective indi- markers,21,2 4 but differences in risk by type of observed Hispanic health advantages, suggest- cators of health status that are related to subse- marker may be informative for the analysis of ing that the “paradox” may not be so paradoxi- quent onset of disease, loss of function, and racial and ethnic differences. It is possible that cal in a population that is heavily weighted mortality.19–21 We examined differences in 10 differentials in risk factors by race, ethnicity, with immigrants.15 ,16 The healthy migrant hy- physiological indicators by race, ethnicity, and and nativity may be more concentrated in pothesis provides 1 explanation for better- nativity.21 These indicators represent multiple some physiological systems than in others. For than-expected health outcomes among Hispan- physiological processes and have individually instance, Blacks have been shown to have a ics.17 ,18 It has been suggested that Hispanics and cumulatively been linked to important age- higher prevalence of hypertension, a cardiovas- who immigrate to the United States are health- related health outcomes, including cardiovascu- cular risk factor, than either Whites or Hispan- ier than Hispanics who remain in their country lar disease, cognitive decline, physical disability, ics of Mexican origin (Mexican Americans).28 of origin; this selection of healthy persons from and death.22–24 We examined both a summary Conversely, metabolic syndrome is more prev- the sending population can improve the level indicator of risk as well as blood pressure, alent in Mexican Americans than it is in of health in the receiving population. Another metabolism, and inflammation risk profiles to Blacks.29 Blacks also have been shown to have explanation of the Hispanic paradox is the investigate whether it was possible to identify higher levels of inflammatory markers such

July 2007, Vol 97, No. 7 | American Journal of Public Health Crimmins et al. | Peer Reviewed | Research and Practice | 1305 ⏐ RESEARCH AND PRACTICE ⏐

as C-reactive protein30,31 and fibrinogen.32 being of Mexican origin and 508 persons TABLE 1—Clinical High-Risk Criteria for Differences in Hispanics are more complicated who indicated that they were born in Mexico. Risk Factors: National Health and and vary by subgroup. One study of all sub- For the analysis of differences between the Nutrition Examination Survey, 1999–2002 groups of Hispanic women found that total Hispanic population and others, the data Biological Risk Indicators High-Risk Cutpoint C-reactive protein levels were similar to those were weighted to represent the total US His- 31 in White women, but an analysis of Mexican panic population, about half of which is indi- Blood pressure risk factors American women found them to have higher viduals of Mexican origin (46.7 %). Systolic blood pressurea ≥140 mm Hg33 30 levels of C-reactive protein than Whites. Respondents excluded from the survey be- Diastolic blood pressurea ≥90 mm Hg33 We used measured indicators of physiologi- cause of missing data in all race and ethnic Pulse rate at 60 s ≥90 cal status to determine whether Hispanics had groups were more likely to be women, older, Metabolic risk factors biological risk profiles similar to those of and to have lower education levels than those Total cholesterolb ≥240 mg/dL34 Whites. We further examined how biological cases included in the analysis. Because of miss- HDL cholesterolb <40 mg/dL34 risk profiles vary by nativity in the total His- ing biological data, 16% of Whites, 16% of Body mass indexc ≥30 kg/m35 panic population and in Hispanics of Mexican Hispanics, and 27% of Blacks were excluded. Glycated hemoglobind ≥6.4 %36 origin. If there is a Hispanic paradox in biologi- Whites and Hispanics did not differ signifi- Inflammation risk factors cal profiles, Hispanics would be expected to cantly from each other in the percentage miss- C-reactive proteine >3.0 mg/L37 have better risk profiles compared with Whites ing data for any of the 10 individual indicators; Fibrinogen >400 mg/dL38 after control for SES. nor did foreign-born and US-born Hispanics Albumin <3.8 g/dL39 differ from each other in the percentage of METHODS missing data for any of the indicators. How- Note. HDL=high-density lipoprotein. aAn average of 2 or 3 sets of seated blood pressure ever, foreign-born Hispanics were more likely measurements. Data than were Whites to lack values for blood bMeasured with Boehringer Mannheim/Hitachi 737 Analyzer (Roche, Basel, Switzerland). We obtained data from the National Health pressure (P=.003) and were less likely to be c Weight in kilograms divided by height in meters squared. and Nutrition Examination Survey (NHANES) missing data for body mass index (weight in dMeasured by boronate affinity chromotagraphy. e 1999–2002, a representative sample of the kilograms divided by height in meters squared; C-reactive protein analyzed by high-sensitivity latex- enhanced nephelometry on a BNII nethelometer. US civilian, noninstitutionalized population col- P=.005). Blacks were more likely than other lected by the National Center for Health Statis- groups to have missing data for all indicators tics. NHANES included information from a except body mass index. Self-reported health questionnaire, laboratory analysis, and clinical status for individuals who were missing data (total cholesterol, high-density lipoprotein cho- exams. We used data for Whites, Blacks, and and for cases used in the analysis, when com- lesterol, glycated hemoglobin, and body mass Hispanics aged 40 years and older (N=5912) pared within racial and ethnic groups, indi- index); and 3 indicators of inflammation (C-re- who participated in the laboratory analysis and cated no difference in self-reported health for active protein, fibrinogen, and albumin). Risk physical exams and who had information on Blacks and Mexican Americans, but Whites was determined by levels measured by clinical 10 biological markers used in our study and all Hispanics who are missing from the and laboratory tests without consideration of (n=4855) and additional independent vari- analysis reported somewhat worse health. prescription drug usage. Although drugs can ables (n=4206). Some respondents had missing data for 1 or be used to control hypertension and choles- The analytic sample (n=4206) was ethni- more indicators of SES, health behaviors, or ac- terol levels, many people who take them do cally diverse, consisting of Whites (n=2338), cess to care (n=649). Information about in- not achieve levels below the cutoff of what is Blacks (n=717), and US-born (n=505) and come was most likely to be missing (n=449). considered high, particularly for hyperten- 40 foreign-born Hispanics (n=646). Hispanics We tested the sensitivity of our results to the in- sion. We developed a summary measure that were self-identified as born in or with ances- clusion of people who did not report income by indicated the number of elevated risk factors tors from Spain or other Spanish-speaking coding them to the mean of the poverty ratio present in total (range 0 to 10) and in each of countries in Central America, South America, and including a dummy variable to indicate the 3 systems: blood pressure (0–3), metabolic or both, including the Caribbean basin. His- their missing status. This did not change results. (0–4), and inflammatory (0–3). panics of Mexican birth or descent who self- We examined associations between race, identified as Mexican, Mexican American, Measures ethnicity, nativity, and biological risk scores 41–4 4 Chicano, or Tex-Mex were classified as Mexi- We created risk scores by summing the after we controlled for SES. We used 2 can Americans. Because of the sampling de- number of biological risk factors that met clini- indicators of SES: (1) education (less than sign, the Hispanic population in NHANES cal high-risk criteria. Details about measures 12 years of schooling), which would reflect 1999–2002 is primarily Mexican American are provided in Table 1. Our measurements in- lifelong conditions, and (2) household income (84.2%). The sample of 4206 persons who cluded 3 indicators related to blood pressure as a percentage of poverty level, which had complete data included 455 US-born (systolic and diastolic blood pressure and would reflect current conditions. The defini- participants who identified themselves as pulse); 4 indicators of metabolic functioning tion of education was limited by the detail on

1306 | Research and Practice | Peer Reviewed | Crimmins et al. American Journal of Public Health | July 2007, Vol 97, No. 7 ⏐ RESEARCH AND PRACTICE ⏐

education provided in NHANES 1999–2002 access to care to determine whether race, Biological Risk Including All Hispanics as a result of privacy concerns. ethnicity, and nativity differences were inde- Table 3 shows mean numbers of high-risk To better account for sources of differences pendent of these factors. Because there is sig- biological factors that have been adjusted for in biological risk by race, ethnicity, and nativ- nificant variation in health, SES, and migra- age and gender. Differences in risk profiles by ity, we controlled for some of the mechanisms tion history1 within the Hispanic population, race and ethnicity, and when controlled for by which demographic and social factors we also separately examined biological risk age and gender, indicated that Hispanics had might be associated with biological risk pro- scores in the largest Hispanic group, Mexican a higher average biological risk score than did files, including health behaviors and availabil- Americans. Whites. This was true for total risk as well as ity of medical care. Health behaviors included for the metabolic and inflammatory subcate- an indicator of lack of exercise (no vigorous RESULTS gories of risk factors, which indicated that in- or moderate activity during the last 30 days), creased risk was spread over multiple physio- current , and poor diet (defined as In health-related surveys, socioeconomic logical systems (Table 3, model 1). Hispanics more than 30% of energy consumption com- characteristics of ethnic groups have been also had significantly fewer biological risk fac- ing from fat).45–48 Because health care use shown to differ in different national tors at high-risk levels in total and for blood may also affect biological risk profiles and be samples49; therefore, it was important to clar- pressure and inflammatory scales than did related to SES, we controlled for current ify how the race, ethnic, and nativity groups Blacks. Hispanics and Blacks had similar health insurance availability, either public or in our analysis differed in education and in- numbers of metabolic factors at high-risk lev- private, which we used as a proxy for access come (Table 2). Hispanics had the lowest SES els even though previous research has em- to health care. of the 3 major groups in our sample. They phasized high metabolic risk among Hispanics had the largest percentage of individuals who compared with Blacks. Analysis had low levels of education (49.8%), the low- Both US-born and foreign-born Hispanics We determined the mean number of bio- est income-to-poverty ratio (2.19), and the had higher numbers of high-risk biological logical risk factors at high-risk levels by race, largest proportion of individuals living below factors than did Whites. In addition, US-born ethnicity, and nativity groups and adjusted for or near the federal poverty level (37.8%), all Hispanics had more inflammatory factors at covariates using Stata 8.2 (Stata Corp, College significantly larger than those for Whites or high-risk levels than did Whites. A compari- Station, Tex) to account for the complex sam- Blacks. Foreign-born Hispanics had signifi- son of biological risk factors among all His- ple design. We first examined differences by cantly lower levels of education and higher panics who were US born and those who race, ethnicity, and nativity; we controlled for levels of poverty than US-born Hispanics. were foreign born indicated that the number age and gender because they both vary across Among Mexican Americans, those who were of biological risk factors at high-risk levels was these population groups and both are related born in Mexico had very low levels of educa- not significantly different between all US-born to these risk factors.49,50 We then controlled tion and income, lower than those of Whites, and foreign-born Hispanics. for SES to examine how the means and differ- Blacks, and US-born Mexican Americans. US- The question of the Hispanic paradox ences in the risk profile score by race, ethnic- born Mexican Americans were less likely than deals with differences relative to SES, so it ity, and nativity would change. Finally, we other Hispanic groups or Blacks to live below was appropriate to examine differences in bi- added a control for health behaviors and or near the federal poverty level. ological risk factors while controlling for edu- cation and income level. After we controlled for SES (Table 3, model 2), we found no sig- nificant difference in the number of risk fac- TABLE 2—Socioeconomic Characteristics of Respondents (N=4206), by Race, Ethnicity, and tors at high-risk levels between Whites and Nativity: National Health and Nutrition Examination Survey, 1999–2002 all Hispanics in any of the measured biologi- All Hispanic Mexican American cal systems, which indicated that if the distri- US Foreign US Mexican bution of low education and income or pov- White Black All born born All born born erty were the same, biologically estimated (n=2338) (n=717) (n=1151) (n=505) (n=646) (n=963) (n=455) (n=508) risk would not be significantly different Less than 12 years education (%) 15.4 39.3a 49.8a,b 36.6a,b 58.1a,b,c 59.1a,b 40.0a,b 75.9a,b,c between these 2 groups. Another interpreta- Household income less than 125% 12.2 27.3a 37.8a,b 30.7a,b 42.3a,b,c 32.8a,b 21.0a,b 43.2a,b,c tion is that the differences are “explained” by of federal poverty level (%) SES. Results were the same for foreign-born Poverty income ratiod (mean) 3.50 2.55a 2.19a,b 2.60a 1.92a,b,c 2.27a,b 2.84a,b 1.77a,b,c and US-born Hispanics. Blacks had a higher number of biological a Significantly different from White. risk factors at high-risk levels than did His- bSignificantly different from Black. cSignificantly different from US-born at the P=.05 level. panics or Whites, even after we controlled dThe poverty income ratio was the ratio of the family income in a given year to the federally defined poverty level. for SES. Although Blacks no longer differed in metabolic risk from the other groups, the

July 2007, Vol 97, No. 7 | American Journal of Public Health Crimmins et al. | Peer Reviewed | Research and Practice | 1307 ⏐ RESEARCH AND PRACTICE ⏐

TABLE 3—Mean Biological Risk Score, by Race, Ethnicity, and Nativity: National Health and biological risk than did Mexican-born Mexi- Nutrition Examination Survey, 1999–2002 can Americans. Also after we controlled for SES, those born in Mexico had levels of bio- All Hispanic Mexican American logical risk factors similar to Whites, in total US Foreign US Mexican and in the 3 subsystems. White Black All born born All born born (n=2338) (n=717) (n=1151) (n=505) (n=646) (n=963) (n=455) (n=508) Differences in Biological Risk and Model 1 Health Behavior and Access to Care Total risk (0–10) 1.87 2.58a 2.19 a,b 2.23 a,b 2.16 a,b 2.26 a,b 2.41 a 2.14 a,b,c To determine whether differences in health Blood pressure risk (0–3) 0.36 0.61 a 0.43b 0.41b 0.45b 0.45 a,b 0.51 a 0.40b behaviors or access to health care explained Metabolic risk (0–4) 0.82 0.96 a 0.96 a 0.99 0.94 1.00 a 1.03 a 0.98 a the lower numbers of high-risk biological fac- Inflammation risk (0–3) 0.69 1.01 a 0.79 a,b 0.83 a,b 0.77b 0.81 a,b 0.87 a,b 0.76b tors for foreign-born Hispanics and the higher Model 2 numbers for Blacks and US-born Mexican Total risk (0–10) 1.92 2.40 a 1.98b 2.08 1.85b 2.00b 2.28 a 1.72b,c Americans, we estimated mean differences Blood pressure risk (0–3) 0.37 0.58 a 0.39b 0.38b 0.40b 0.41b 0.49 0.33b,c and controlled for lack of exercise, current Metabolic risk (0–4) 0.85 0.89 0.85 0.92 0.80 0.87 0.97 0.78c smoking status, percentage of diet from fat, Inflammation risk (0–3) 0.70 0.95 a 0.71b 0.78b 0.66b 0.72b 0.82b 0.62b,c and health insurance status (Table 3, model Model 3d,e 3). Results did not change much relative to Total Risk (0–10) 1.92 2.40 a 1.98b 2.08 1.90b 2.03b 2.28 a 1.79b,c the model that controlled only for SES, sug- Blood pressure risk (0–3) 0.37 0.59 a 0.39b 0.38b 0.40b 0.37b 0.49b 0.33b,c gesting that differences in health behaviors Metabolic risk (0–4) 0.85 0.88 0.86 0.92 0.81 0.88 0.96 0.80c and access to health care cannot explain the Inflammation risk (0–3) 0.70 0.93 a 0.73b 0.78b 0.69b 0.74b 0.83 0.65b,c differences in biological risk. Mexican-born persons were much less likely to have health Note. Model 1 adjusted for age and gender. Model 2 adjusted for age, gender, low education, and poverty score. Model 3 adjusted for age, gender, low education, poverty score, health behaviors, and access to health care. insurance than were US-born Mexican Ameri- aSignificantly different from White. cans (17% vs 45%) and were less likely to ex- b Significantly different from Black. ercise (38% vs 58%); by contrast, they were cSignificantly different from US born at the .05 level. dHealth behaviors included current smoking status, poor diet (defined as more than 30% of energy coming from fat; from less likely to have a high-fat diet (52% vs self-reports of all foods and beverages consumed in a 24-h period before interview); and no exercise (i.e., no vigorous or 73%) and did not differ in the likelihood of moderate exercise for at least 10 minutes over the past 30 days; vigorous activities defined as those that cause heavy smoking. Our control for the combination of sweating or large increases in breathing or heart rate [e.g., running, lap swimming, aerobics classes, or fast bicycling], and moderate activities defined as those that cause only light sweating or a slight to moderate increase in breathing or heart rate these factors did not affect differences be- [e.g., brisk walking, bicycling for pleasure, golf, and dancing]). tween US-born and foreign-born Mexican e Measured as currently having insurance (private or public, Medicare, Medicaid/Children’s Health Insurance Program, or Americans. other government insurance). DISCUSSION differences between Blacks and Hispanics higher among Mexican Americans than and between Blacks and Whites hardly among Whites. When we examined Mexican Our goal was to determine whether objec- changed after SES was controlled, which in- Americans by nativity, US-born Hispanics of tive biological measures of health risk profiles dicated that Blacks had a higher risk for poor Mexican origin had poorer biological risk pro- would provide evidence of the existence of a health than predicted by their level of SES. files in total and within all systems than did Hispanic paradox. However, we did not find a The disappearance of the difference between Whites; Mexican-born Hispanics had higher Hispanic paradox in biological risk profiles. Blacks and others in metabolic risk indicates total and metabolic risk levels than did The Hispanic population had higher biological that the difference was related more to SES Whites. The US-born population of Mexican risk than did the White population on 3 of 4 disparities than to disparities in other areas origin did not differ from Blacks in their total scores (total, metabolic, and inflammation risk of biological risk. biological and metabolic risk profiles, but they profiles), which indicated worse biological did have lower inflammatory risk than did profiles across a range of physiological sys- Biological Risk Among Only Mexican Blacks. The Mexican-born population had tems. The differences existed between Whites Americans lower mean numbers of total risk factors as and all Hispanics and specifically for those of When we limited our analysis of Hispanics well as blood pressure and inflammatory risk Mexican heritage. This finding is in line with to Mexican Americans, the results were gen- factors than did Blacks. numerous studies in the literature that have erally similar to the results for all Hispanics. When we controlled for SES, US-born not found lower levels of risk in individual When we controlled for age and gender, the Mexican Americans still had worse total bio- risk factors for Hispanics.30,51–53 number of risk factors at high-risk levels in logical risk profiles than did Whites but not Our control for the low SES of the His- total and within the 3 subcategories was Blacks; they also had significantly higher total panic population eliminated the differences in

1308 | Research and Practice | Peer Reviewed | Crimmins et al. American Journal of Public Health | July 2007, Vol 97, No. 7 ⏐ RESEARCH AND PRACTICE ⏐

biological risk between the White, total His- Our study and others51,55,56 have found that American, our ability to look at other Hispanic panic, Mexican-origin, and foreign-born (His- Blacks have higher levels of biologically esti- subgroups was limited. Despite these limita- panic and Mexican) populations. When we mated risk than do Whites, even after we con- tions, this study provides evidence that His- controlled for low SES, we found that Mexi- trolled for SES and health behaviors. We also panics are not advantaged in their measured can immigrants had a total biological risk pro- found this to be true for the US-born Mexican biological risk profiles. file similar to Whites and better than US-born American population. The disproportionate Mexican Americans. This is not a Hispanic burden of biological risk in the Black and US- About the Authors paradox but rather an indication that the high born Hispanic population may arise from gen- Eileen Crimmins and Jung Ki Kim are with the Andrus levels of risk in the foreign-born Mexican pop- erations of social inequality, which was not Gerontology Center, University of Southern California, Los ulation relative to Whites are related to low captured in our measure of SES. Angeles. Dawn Alley is with the Robert Wood Johnson Health and Society Scholars Program, University of Penn- SES. However, our control for SES did not Overall, our results demonstrate the useful- sylvania, Philadelphia. Arun Karlamangla and Teresa eliminate differences between Whites and ness of examining objective measures of risk Seeman are with the Division of Geriatrics, School of Med- US-born Hispanics of Mexican origin. We for mortality and other poor outcomes and icine, University of California, Los Angeles. Requests for reprints should be sent to Eileen Crimmins, were not able to “explain” this higher level of the importance of addressing differences be- Andrus Gerontology Center, University of Southern risk among US-born Mexican Americans rela- tween subgroups of the Hispanic popula- California, Los Angeles, CA 90089-0191 (e-mail: tive to the White population. It may have tion.1,57 Objective measures of risk for mortal- [email protected]). arisen from the same social inequities that are ity and other poor outcomes show that at the basis of the differences between Blacks ethnicity, nativity, and SES contribute to dif- Contributors E. Crimmins designed the overall study and supervised all and others. This higher level of risk among ferences in clinically relevant risk factor pro- aspects of its implementation. J.K. Kim and D. Alley com- US-born Mexican Americans relative to files, which point to the potential for appropri- pleted the analyses, interpretted findings, and reviewed Whites, although not significant in the subcat- ate interventions. In addition, these biological drafts of the article. A. Karlamangla and T. Seeman inter- pretted findings and reviewed drafts of the article. egories, appears to be spread across all 3 cat- measures may help researchers avoid some of egories. When we controlled for SES, health the interpretation issues they face with self- 9,10,13 Human Participant Protection behaviors, and access to care, metabolic dif- reported and mortality data. This study was approved by the University of Southern ferences between Hispanics and Whites dis- Our study had limitations. First, although California’s University Park Institutional Review Board. appeared, which suggests that SES factors this analysis adds to the literature by provid- played a role in metabolic risk differences. ing information about racial, ethnic, and nativ- Acknowledgments Our findings of higher metabolic risk for Mex- ity differences in risk factors for major dis- This research was supported by the National Institute on Aging (grants R01 AG023347, T32 AG00037, ican Americans are consistent with other eases, disability, and death, we do not know K12AG01004, and P30 AG17265). studies that found this group to have a higher that these risk factors are equally important prevalence of mellitus. Because dia- when it comes to producing these outcomes in References betes is primarily related to abdominal obe- all groups. This is an important area for future 1. Cho Y, Frisbie WP, Hummer RA, Rogers R. Nativ- sity, public health measures to improve diet research. Second, only respondents who had ity, duration of residence, and the health of Hispanic adults in the United States. Int Migration Rev. 2004; and exercise in this group would mean reduc- complete data for all 10 biomarkers were in- 38:184–211. tion in the morbidity associated with diabetes. cluded in the study. Those missing data were 2. Elo IT, Turra CM, Kestenbaum B, Ferguson BR. Our study supports other studies that have more likely to be women, older, Black, have Mortality among elderly Hispanics in the United found that the US-born Mexican American lower SES, and have somewhat worse self-re- States: past evidence and new results. Demography. 2004;41:109–128. population has higher levels of cardiovascular ported health than those respondents with 54 3. Franzini L, Ribble J, Keddie A. Understanding the risk than the foreign-born population. In complete data. If the sickest individuals were Hispanic paradox. Ethn Dis. 2001;11:496–518. our analysis, the difference persisted after we less likely to provide complete data, it is prob- 4. Hummer RA, Benjamins MR, Rogers RG. Racial controlled for SES and health behaviors and able that we underestimated the poorer health and ethnic disparities in health and mortality among access to care. The higher level of risk in the status of the Black and low-SES populations; the US elderly population. In: Anderson N, Bulato R, Cohen B, eds. Critical Perspectives on Racial and Ethnic US-born population relative to the foreign- therefore, nonresponse is unlikely to change Differences in Health in Late Life. Washington, DC: The born population appears to be unexplained our conclusions. Third, educational informa- National Academies Press; 2004:53–94. by behavioral and care differences and may tion in this sample was limited to 2 categories, 5. Liao Y, Cooper RS, Cao G, et al. Mortality pat- be related to initially healthy immigrants or rather than a continuous measure of years of terns among adult Hispanics: findings from the NHIS, 1986 to 1990. Am J Public Health. 1998;88:227–232. the return migration of the unhealthy. The education. Having limited categories may 6. Markides KS, Eschbach K. Aging, migration, and fact that differences in biological risk profiles have obscured differences in education levels, mortality: current status of research on the Hispanic par- by nativity are not as strong as differences in particularly at low levels of education. Hispan- adox. J Gerontol B Psychol Sci Soc Sci. 2005;60:68–75. mortality may mean that return migration is ics, particularly foreign-born Hispanics, have 7. Adler NE, Boyce T, Chesney MA, et al. Socioeco- nomic status and health: the challenge of the gradient. more strongly related to serious illness than lower levels of education relative to both Am Psychol. 1994;49:15–24. to the earlier-occurring changes in biological White and Black populations. Finally, because 8. Hunt KJ, Resendez RG, Williams K, Haffner SM, risk factors.16 the sample was predominantly Mexican Stern MP, Hazuda HP. All-cause and cardiovascular

July 2007, Vol 97, No. 7 | American Journal of Public Health Crimmins et al. | Peer Reviewed | Research and Practice | 1309 ⏐ RESEARCH AND PRACTICE ⏐

mortality among Mexican American and non-Hispanic 26. Reynes JF, Lasater TM, Feldman H, Assaf AR, 42. Winkleby MA, Jatulis DE, Frank E, Fortmann SP. White older participants in the San Antonio Heart Carleton RA. Education and risk factors for coronary Socioeconomic status and health: how education, income, Study—evidence against the “Hispanic paradox.” Am J heart disease: results from a New England community. and occupation contribute to risk factors for cardiovascu- Epidemiol. 2003;158:1048–1057. Am J Prev Med. 1993;9:365–371. lar disease. Am J Public Health. 1992;82:816–820. 9. Patel KV, Eschbach K, Ray LA, Markides KS. 27. Winkleby MA, Fortmann SP, Barrett DC. Social 43. Winkleby MA, Cubbin C, Ahn DK, Kraemer HC. Evaluation of mortality data for older Mexican Ameri- class disparities in risk factors for disease: eight-year Pathways by which SES and ethnicity influence cardio- cans: implications for the Hispanic paradox. Am J Epi- prevalence patterns by level of education. Prev Med. vascular disease risk factors. Ann NY Acad Sci. 1999; demiol. 2004;159:707–715. 1990;19:1–12. 896:191–209. 10. Smith DP, Bradshaw BS. Rethinking the Hispanic 28. Burt VL, Welton P, Roccella EJ, et al. Prevalence 44. Mirowsky J, Ross CE. Education, personal control, paradox: death rates and for US non- of hypertension in the US adult population: results from lifestyle and health: a human capital hypothesis. Res Hispanic White and Hispanic populations. Am J Public the Third National Health and Nutrition Examination Aging. 1998;20:415–449. Health. 2005;96:1686–1692. Survey, 1988–1991. Hypertension. 1995;25:305–313. 45. Hellenius ML, de Faire U, Berglund B, Hamsten A, 11. Hajat A, Lucas JB, Kington R. Health outcomes 29. Park YW, Zhu S, Palaniappan L, Heshka S, Krakau I. Diet and exercise are equally effective in re- among Hispanic subgroups: data from the National Carnethon MR, Heymsfield SB. The metabolic syn- ducing risk for cardiovascular disease. Results of a Health Interview Survey, 1992–1995. Adv Data. drome: prevalence and associated risk factor findings randomized controlled study in men with slightly to 2000;310:1–14. in the US population from the Third National Health moderately raised cardiovascular risk factors. Athero- 12. Markides KS, Coreil J. The health of Hispanics in and Nutrition Examination Survey, 1988–1994. Arch sclerosis. 1993;103:81–91. the southwestern United States: an epidemiologic para- Intern Med. 2003;163:427–436. 46. Koster A, Bosma H, Penninx B, et al. for the dox. Public Health Rep. 1986;101:253–265. 30. Ford ES, Giles WH, Mokdad AH, Myers GL. Health ABC Study. Association of inflammatory mark- 13. Angel R, Thoits P. The impact of culture on the Distribution and correlates of C-reactive protein con- ers with socioeconomic status. J Gerontol A Bio Sci Med cognitive structure of illness. Cult Med Psychiatry. centrations among adult US women. Clin Chem. Sci. 2006;61:284–290. 1987;11:465–494. 2004;50:574–581. 47. Lynch JW, Kaplan GA, Salonen JT. Why do poor 14. Ren X, Amick B. Racial and ethnic disparities in 31. Albert MA, Glynn RJ, Buring J, Ridker PM. people behave poorly? Variation in adult health behav- self-assessed health status: evidence from the National C-reactive protein levels among women of various eth- iours and psychosocial characteristics by stages of the so- Survey of Families and Households. Ethn Health. nic groups living in the United States (from the Women’s cioeconomic lifecourse. Soc Sci Med. 19 97;4 4:8 0 9–819. 1996;1:293–303. Health Study). Am J Cardiol. 2004;93:1238–1242. 48. Ockene IS, Miller NH for the American Heart 32. Tracy RP, Bovill EG, Fried LP, et al. The distribu- 15. Jasso G, Massey DS, Rosenzweig MR, Smith JP. Association Task Force on Risk Reduction. Cigarette tion of coagulation factors VII and VIII and fibrinogen Immigrant health: selectivity and . In: smoking, cardiovascular disease, and . Circulation. in adults over 65 years: results from the Cardiovascu- Anderson N, Bulato R, Cohen B, eds. Critical Perspec- 1997;96:3243–3247. tives on Racial and Ethnic Differences in Health in Late lar Health Study. Ann Epidem. 1992;2:509–519. 49. Crimmins EM, Hayward MD, Seeman TE. Race/ Life. Washington, DC: The National Academies Press; 33. Chobanian AV, Bakris GL, Black HR, et al. Sev- ethnicity, socioeconomic status and health. In: Anderson 2004:227–268. enth report of the Joint National Committee on Preven- N, Bulato R, Cohen B, eds. Critical Perspectives on Ra- 16. Palloni A, Arias E. Paradox lost: explaining the tion, Detection, Evaluation, and Treatment of High cial and Ethnic Differences in Health in Late Life. Wash- Hispanic adult mortality advantage. Demography. Blood Pressure. Hypertension. 2003;42:1206–1252. ington, DC: The National Academies Press; 2004: 2004;41:385–415. 34. National Cholesterol Education Program (NCEP). 310–352. 17. Abraido-Lanza AF, Dohrenwend BP, Ng-Mak DS, Detection, evaluation, and treatment of high blood 50. Seeman TE, Singer BH, Ryff CD, Love GD, Turner JB. The Latino mortality paradox: a test of the cholesterol in adults. Available at: http://www.nhlbi. Levy-Storms L. Social relationships, gender, and allo- “salmon bias” and healthy migrant hypotheses. Am J nih.gov/guidelines/cholesterol/atp3xsum.pdf. Accessed static load across two age cohorts. Psychosom Med. Public Health. 1999;89:1543–1548. March 4, 2004. 2002;64:395–406. 18. Sorlie PD, Backlund E, Johnson NJ, Rogot E. 35. World Health Organization (WHO) expert consul- 51. Mensah GA, Mokdad AH, Ford ES, Greenlund KJ, Mortality by Hispanic status in the United States. tation. Appropriate body-mass index for Asian popula- Croft JB. State of disparities in cardiovascular health in JAMA. 1993;270:2464–2468. tions and its implications for policy and intervention the United States. Circulation. 2005;15:1233–1241. 19. Brunner EJ. Toward a new social biology. In: strategies. Lancet. 2004;363:157–163. Berkman LF, Kawachi I, eds. Social Biology. New York, 36. United States Preventive Services Task Force 52. Mitchell B, Stern M, Haffner S, Hazuda H, Patterson NY: Oxford University Press; 2000:306–311. (USPSTF). Use of glycated hemoglobin and microalbu- J. Risk factors for cardiovascular mortality in Mexican Americans and non-Hispanic Whites. San Antonio 20. Crimmins EM, Seeman T. Integrating biology minuria in the monitoring of diabetes mellitus. Avail- Heart Study. Am J Epidemiol. 1990;131:423–433. into the study of health disparities. Popul Dev Rev. able at: http://www.ahrq.gov/clinic/epcsums/glycasum. 2004;30:89–107. pdf. Accessed March 4, 2004. 53. Ogden C, Carroll M, Curtin L, McDowell M, Tabak C, Flegal K. Prevalence of overweight and obe- 21. Seeman TE, Crimmins E, Huang M, et al. Cumula- 37. Ridker PM. C-reactive protein: a simple test to sity in the United States, 1999–2004. JAMA. 2006; tive biological risk and socio-economic differences in help predict risk of heart attack and stroke. Circulation. 295:1549–1555. mortality: MacArthur Studies of Successful Aging. Soc 2003;108:e81–e85. Sci Med. 2004;58:1965–1997. 38. Palmieri V, Celentano A, Roman M, et al. Relation 54. Sundquist J, Winkleby M. Cardiovascular risk fac- 22. Fried LP, Kronmal RA, Newman AB, et al. Risk fac- of fibrinogen to cardiovascular events is independent tors in Mexican American adults: a transcultural analy- tors for 5-year mortality in older adults: the Cardiovascu- of preclinical cardiovascular disease: the strong heart sis of NHANES III, 1988–1994. Am J Public Health. lar Health Study. JAMA. 1998;279:585–592. study. Am Heart J. 2003;145:467–474. 1999;89:723–730. 23. Ridker PM, Rifai N, Rose L, Buring JE, Cook NR. 39. Corti M, Guralnik JM, Salive ME, Sorkin JD. Serum 55. Geronimus AT, Hicken M, Keen D, Bound J. Comparison of C-Reactive Protein and low-density albumin level and physical disability as predictors of mor- “Weathering” and age patterns of allostatic load scores lipoprotein cholesterol levels in the prediction of first car- tality in older persons. JAMA. l994;272:1036–1042. among Blacks and Whites in the United States. Am J Public Health. 2006;96:826–833. diovascular events. N Engl J Med. 2002;347:1557–1565. 40. Hajjar I, Kotchen TA. Trends in prevalence, aware- 24. Seeman TE, Singer B, Rowe J. Exploring a new ness, treatment, and control of hypertension in the United 56. Winkleby M, Kraemer H, Ahn D, Varady A. concept of cumulative biological risk—allostatic load and States, 1988–2000. JAMA. 2003;290:199–206. Ethnic and socioeconomic differences in cardiovascular disease risk factors. JAMA. 1998;280:356–362. its health consequences: MacArthur Studies of Successful 41. Brunner EJ, Marmot MG, Nanchahal K, et al. Aging. Proc Natl Acad Sci USA. 2001;98:4770–4775. Social inequality in coronary risk: central obesity and 57. Borrell L. Racial identity among Hispanics: impli- 25. Reaven GM. Role of insulin resistance in human the metabolic syndrome. Evidence from the Whitehall cations for health and well-being. Am J Public Health. disease. Diabetes. 1998;37:1595–1607. II Study. Diabetologia. 1997;40:1341–1349. 2005;3:379–381.

1310 | Research and Practice | Peer Reviewed | Crimmins et al. American Journal of Public Health | July 2007, Vol 97, No. 7