Ethnic variation in health and the determinants of health among Latinos

Social Science & Medicine 61 (2005) 53 63 www.elsevier.com/locate/socscimed Ethnic variation in health and the determinants of health among Latinos Barbara A. Zsembik, Dana Fennell Department of Sociology, University of Florida, P.O. Box 117330, Gainesville, FL 32611-7330, USA Available online 18 January 2005 Abstract The purpose of this research is to document and explain ethnic variation in health among Latino adults in the United States. Results of analyses of data pooled from the 1997 2001 National Health Interview surveys reveal that health patterns are clearly different among Latino ethnic groups. Mexicans have health advantages, whereas Puerto Ricans experience health disparities. Cubans and Dominicans reveal a mix of health disparities and advantages, depending on the health outcome. The effects of social determinants of health are also contingent upon ethnicity. For example, worse health is associated with higher levels of socioeconomic status (SES) and acculturation among Mexicans, but with lower levels of SES and acculturation among Latinos whose origins are from Caribbean islands. We conclude that racial/ ethnic comparative health research should avoid pan-ethnic groupings, and explicitly acknowledge ethnic group distinctiveness. r 2004 Elsevier Ltd. All rights reserved. Keywords: Latinos; Health disparities; Determinants; USA Introduction The extent and nature of Latino health disparities in the US is inconclusive, with an especially thin knowledge of the health of Puerto Ricans, Cubans, and other non- Mexican-origin Latinos. Latinos, as a group, face sociocultural and economic barriers to health care, yet paradoxically they have lower levels of infant and adult mortality (Abraido-Lanza, Dohrenwend, Ng-Mak, & Turner, 1999; Cobas, Balcazar, Benin, Keith, & Chong, 1996; Falcon & Tucker, 2000; Hummer, Rogers, Amir, Forbes, & Frisbie, 2000; Idler & Angel, 1990; Landale, Oropesa, & Gorman, 2000; Padilla, Boardman, Hummer, & Espitia, 2002; Sorlie, Backlund, Johnson, & Hogat, 1993). We know much less about the major Corresponding author. Tel.: +1 352 392 0251x226; fax: +1 352 392 6568. E-mail address: zsembik@soc.ufl.edu (B.A. Zsembik). health statuses and processes that characterize the epidemiological reality of contemporary US adulthood, and are the focus of current social and behavioral science health research. The scanty evidence available suggests relatively good health among Latinos, given their fewer economic resources (Hayes-Bautista, 1992; Markides & Coreil, 1986). A more recent analysis, however, describes significant ethnic heterogeneity among Latinos, finding better health among Cubans and Mexicans, and a discernible health disparity among Puerto Ricans (Hajat, Lucas, & Kington, 2000). The purpose of this research is to document and explain ethnic variation in health among Latino adults. National-level data permit us to distinguish among the three largest Latino groups in the US, Mexicans, Puerto Ricans, and Cubans, as well as among Dominicans, a rapidly growing ethnic group. We describe Latino ethnic variation evident in different levels of health, yielding health disparities or health advantages, relative 0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2004.11.040

54 B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 to non-latino whites. We explore ethnic variation that may also appear in how social determinants of health shape health outcomes and differentials. In the next section, we delineate the sets of health determinants prevalent in social science research on health disparities. As we discuss each set of determinants, we evaluate the extent to which they differ among the Latino ethnic groups, and thus pose mechanisms that yield ethnic differentials in health. Social determinants of health Much current US research and policy interest focuses on health disparities in order to reduce societal inequalities and to improve overall population health. This large and burgeoning literature establishes that socioeconomic status (SES) has a strong impact on health, revealing at all points along the SES gradient that poorer economic conditions are correlated with poorer health (Feinstein, 1993; House et al., 1996; Marmot & Wilkinson, 1999). Recognizing the substantial and longstanding association of racial/ethnic status and SES, a great deal of health disparities among African Americans, Latinos, and Native Americans may be attributed to their disproportionate concentration in the lower segments of the SES distribution (House & Williams, 2000; Kington & Nickens, 2001; Williams, 2001). More recent research recognizes the distinctive and independent effect of racial/ethnic status on health, typically attributed to the psychological and physiological toll of stress produced by racism and discrimination (House & Williams, 2000; Kington & Nickens, 2001; Krieger, 2000; Williams, 2001) or to acquisition of negative health behaviors with acculturative processes (Amaro & de la Torre, 2002; Finch, Hummer, Reindl, & Vega, 2002). Finally, racial/ethnic status and SES independently shape exposure to negative health behaviors and the impact of negative behaviors on health outcomes, further contributing to ethnic variation in morbidity. Socioeconomic factors Better health is associated with higher levels of SES, whether measured as higher levels of education, occupational prestige, income and net worth, or insurance coverage (Lynch & Kaplan, 2000). Levels of SES vary by immigrant generation, postulated by the classic assimilation model to be lowest among the immigrant generation, then rising with each subsequent generation. SES levels are also sensitive to opportunities for prosperity in the local labor market, especially the presence of an ethnic economic enclave. Heterogeneity in SES derives from ethnic differences in the premigration human and business capital of migrants, geographic distribution and concentration of ethnic groups across distinct US regional economies, and group-specific experiences of economic and social assimilation. Observed SES heterogeneity among Latinos is likely to yield a similar pattern of health heterogeneity. Initial Mexican and Puerto Rican migration streams drew more heavily from the lower segments of the socioeconomic distribution, whereas post-wwii Cuban immigration drew from the upper middle classes. More recent migrations of upper middle class Puerto Ricans into Florida and other parts of the US and of lower SES Cubans into southern Florida introduces greater SES heterogeneity into current immigrant generations. The socioeconomic success of and their children and grandchildren is different among Latino ethnic groups (Portes & Rumbaut, 1990). The vital Cuban enclave provided opportunities to the immigrant generation and their children, and the youngest Cuban generations are now successfully transitioning into the professional classes (Zsembik, 2000). Social mobility appears highest for children of Mexican, but the third generation s successes are lower than those of their parents (Zsembik & Llanes, 1996). The Puerto Rican experience better fits a segmented assimilation model, wherein social mobility is generally constrained across all generations (Landale, Oropesa, Llanes, & Gorman, 1999; Massey, 1993; Tienda, 1989; Torres & Rodriguez, 1991), partially due to their concentration in the economically struggling northeastern metropolises. Cultural factors Latino ethnicity may directly affect health through other social mechanisms such as discrimination, migration, and culture. Individual and institutional racism and discrimination affect health outcomes among Latinos through residential segregation, the quality and quantity of goods (e.g., fresh produce) and services (e.g., education and medical care), and subjectively experienced stress (Amaro, Russo, & Johnson, 1987; Salgado de Snyder, 1987). The concept of an epidemiological paradox arises among studies of Mexican because of their good health relative to their poor economic profile, and prompts speculation on the relative roles of migration and culture in producing health. Migration selectivity is a potential explanation for the observed Mexican paradox, where migration to the US selects healthier and return migration to Mexico repatriates who become ill or acquire impairments and disability (Abraido-Lanza et al., 1999; Vega & Amaro, 1994; Weeks, Rumbaut, & Ojeda, 1999). A complementary explanation contends that cultural factors account for the strikingly good health among Mexicans with a poor socioeconomic profile, buffers

B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 55 which erode with acculturation (Cobas et al., 1996; Guendelman & Abrams, 1994; Scribner, 1996). Specifically, cultural buffers are evident in health histories of minimal alcohol, drug, and tobacco use. Acculturation into mainstream US health habits implies an increasing prevalence of negative health behaviors. It remains an empirical question whether explanations derived from studies of the Mexican origin population also extend to other Latino ethnic groups. Together these explanations, and assuming they extend to other Latino ethnicities, suggest better health among culturally endowed relative to US-born Latinos and whites. The health advantage of the Latino population is decremented by the selective return migration of ailing and the erosion of cultural buffers with longer duration of residence among and subsequent generations (Black & Markides, 1993; Finch, Hummer, Kolody, & Vega, 2001; Himmelgreen, Perez-Escamilla, Peng, Martinez, & Wright, 2001; Scribner, 1996; Wolff & Portis, 1996). The definitions of culture and acculturation, and the processes of acculturation, are rarely well specified, and typically used to explain observed health differences between and non-latinos, and the health differences between and subsequent generations (Arcia, Skinner, Bailey, & Correa, 2001; Hunt, Schneider, & Comer, 2004). Neither are the complex interrelationships of culture, SES and health behaviors fully examined. Consequently there are appreciable limitations in the number and variety of measures of culture in demographic and health surveys. Cultural factors are most commonly measured by proxy variables of language and immigrant generation. Typically cultural effects are inferred from significant ethnic group differences that persist after statistical adjustment for economic and demographic variables, and significant language and nativity coefficients. Heterogeneity in language and nativity should yield heterogeneity in health outcomes. Mexicans have a greater variability than Puerto Ricans and Cubans in the sizes and numbers of immigrant generations (del Pinal & Singer, 1997), and have larger migrant flows between the US and Mexico than Cubans do with Cuba. Puerto Ricans US citizenship and circular migration patterns provides for greater acculturation across all migrant generations and a minor role of migrant selectivity. On the other hand, Cubans concentration in the immigrant generation, and geographic and social concentration in an ethnic enclave, promotes cultural retention and prevents migration selectivity. Health behavior risks Genetic risk factors vary across race and ethnic groups, risk factors that interact with health behaviors to determine whether risk is activated and to what extent pathology progresses (Singer & Ryff, 2001). For example, the indigenous heritage of Mexicans has been implicated in their stronger genetic predisposition for adult onset diabetes, a risk enhanced by lower levels of physical activity and healthy body weight (Gardner et al., 1984; Lorenzo et al., 2001; Sundquist & Winkleby, 2000). Health behavior risks consist of tobacco use, alcohol and substance abuse, physical inactivity and excess body weight. The research literatures yield few studies that compare the health behaviors of Latino ethnic groups, most focusing either on a pan-ethnic Latino/Hispanic group or a nonprobabilistic community sample of members of an individual ethnic group, typically Mexican Americans. Findings from these studies indicate that the use of tobacco, alcohol, and illicit substances is higher among the more-acculturated native-born than immigrant Latinos (Amaro & de la Torre, 2002; Black & Markides, 1993; Gundelman, 1998; Guendelman & Abrams, 1994). Available evidence also shows that Latinos are disproportionately overweight and less physically active than non-latino whites (CDC, 1997; Zambrana & Logie, 2000). Obesity appears particularly prevalent in the Mexican-origin population (Randolph, Stroup-Benham, Black, & Markides, 1998; Sundquist & Winkleby, 2000; Winkleby, Robinson, Sundquist, & Kraemer, 1999). The analytical focus of this research centers on addressing gaps remaining in our understanding of health outcomes among Latinos in the US. What are the health profiles of non-mexican Latinos? To what extent do health advantages and health disparities vary among Latino ethnic groups? Do the conventional determinants of health vary among Latino national origin groups? The answers to these questions will result in a more comprehensive understanding of health patterns of Latinos in the US by providing baseline health profiles of Latinos of Caribbean origin and identifying advantages and disparities in several health outcomes in adulthood. Consequently the analyses will provide insight into the utility of a pan-ethnic Latino group in health research and the generalizability of inferences based on health studies of the Mexican-origin population to other Latino ethnic groups. Research design and methods Data and sample Data for this research are pooled from 5 years (1997 2001) of the National Health Interview Survey (NHIS). The NHIS, conducted by the National Center of Health Statistics and the Centers for Disease Prevention and Control, is an annual multipurpose

56 B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 health survey of a nationally representative sample of households. Face-to-face interviews collect three data components: a family core, sample adult core, and sample child core. This analysis uses the sample adult core because of its detailed information on health outcomes. The final response rate of the sample adult cores ranges from a low of 69.6% in 1999 to 80.4% in 1997. Each year in the sample adult core component, approximately 30,000 randomly selected adults completed interviews. We restrict the sample to adults aged 25 and older to more accurately measure individual SES characteristics because the transition to economically independent adulthood has extended into the early 20s in the US. Educational careers are likely to be completed, and income and insurance access are less likely to reflect parental characteristics. A new survey design beginning in 1997 allows multiple surveys to be pooled to increase the sample sizes of Latino sub-groups. Our analytical sample is comprised of Latinos, and for comparative purposes, non-latino whites (n ¼ 94; 733; 73.1%) and non-latino Blacks (n ¼ 18; 747; 14.5%). We disaggregate Latinos into national origin sub-groups that are relatively homogeneous, yet large enough to sustain analysis. Specific national origins can be determined in all five surveys for the three largest Latino groups in the US: Mexicans (n ¼ 12; 028; 9.3%), Puerto Ricans (n ¼ 2268; 1.8%), and Cubans (n ¼ 1432; 1.1%). Dominicans can be identified in the 1999, 2000 and 2001 surveys, a numerically small category (n ¼ 388; 0.3%) that we preserve in order to provide the first national estimates of a rapidly growing Caribbean Latino sub-population. National origin detail is unavailable for the remainder of Latinos, who are excluded from analysis. The resultant pooled sample size equals 129,596. NHIS data limitations include its cross-sectional nature and the relative absence of high quality measures of SES and culture variables. The unique character of the NHIS to address the proposed research objectives far outweighs these limitations, however. The data provide sufficient numbers of adults in each homogeneous Latino sub-group. The health items offer standard measures used to assess the health of the US population. The national probability sample permits results to be generalized to the population. Measures of health outcomes Three health outcomes are used: the number of chronic medical conditions, the number of physical functioning impairments, and self-rated health. The chronic medical conditions index cumulates the number of diagnoses a respondent has received for the following major medical conditions: hypertension, heart disease (inclusive of coronary heart disease, angina pectoris, and heart attack), stroke, emphysema, cancer, and diabetes. The functional limitation index cumulates the number of functional activities a respondent cannot perform or performs only with great difficulty: walk 1/4 mile, walk up 10 steps, stand for 2 hours, sit for 2 hours, stoop/ bend/kneel, reach over head, grasp objects, lift/carry 10 pounds, and push a large object. Self-perceived health is the standard question that asks for a subjective health rating of excellent, very good, good, fair, or poor. Independent variables Education, income, and health insurance coverage serve as socioeconomic indicators of health. Education is the number of years completed. Income is measured as the ratio of family income to poverty threshold, wherein higher values reflect incomes farther from poverty. Persons without income data are retained in the analysis as a separate category. Persons who have no health insurance are compared to persons who have some form of health insurance. Cultural determinants of health include nativity, immigrant s length of residence in the US, and language of interview. The length of US residence among is categorized into: less than 10 years, 10 14 years, and 15 or more years. Respondents who did not complete the interview in English, or did so in a mixture of Spanish and English, are compared with those who replied only in English. Health risk factors include substance use histories, activity level, and weight. The smoking and drinking histories organize people into current users, former users, and lifetime abstainers. The physical activity variable describes whether respondents are inactive (do not participate in moderate or vigorous activities at least 3 times per week). The body mass index is used to categorize people as overweight (25.0+). We control for the effects of sex and age. Procedures Data are weighted and scaled to approximate the sample size in order to facilitate evaluation of the relative effects of variable sets across models. Because of the skewed distributions of the health outcome variables, they are truncated and estimated in STATA with an ordinal logit model. Standard errors are adjusted for the effects of the complex sampling design. Results Descriptive statistics The ethnoracial group distributions and descriptive statistics of variables in the models are provided in

B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 57 Table 1. Among Latinos, Mexicans have fewer medical conditions, and Mexicans, Cubans, and Dominicans have comparably low levels of functional impairment. Mexicans in particular present a profile of relative health advantage relative to non-latino whites and blacks. Puerto Ricans have a more negative health profile, a pattern of health disparities, especially evident in strikingly higher levels of comorbid medical conditions and multiple impairment. Higher proportions of Puerto Ricans, Cubans and Dominicans report fair or poor health, whereas Mexicans subjective health ratings are more comparable to whites. Table 1 Descriptive statistics, National Center for Health Statistics (2004) (n ¼ 129; 596): proportions and means (Standard errors) Mexican Puerto Rican Cuban Dominican Black White Medical conditions a 0 0.76 0.65 0.66 0.68 0.57 0.61 1 0.18 0.24 0.26 0.26 0.31 0.28 2+ 0.06 0.11 0.08 0.06 0.12 0.11 Functional impairments a 0 0.78 0.67 0.77 0.82 0.66 0.64 1 3 0.08 0.10 0.08 0.05 0.10 0.15 4+ 0.13 0.23 0.15 0.12 0.24 0.21 Health a Good excellent 0.87 0.80 0.82 0.84 0.80 0.89 Fair 0.10 0.15 0.14 0.12 0.14 0.08 Poor 0.03 0.05 0.04 0.04 0.05 0.03 Education b 11 (0.08) 13 (0.18) 13 (0.18) 12 (0.12) 14 (0.05) 15 (0.02) Income-poverty ratio (IPR) b 5 (0.07) 6 (0.14) 7 (0.23) 5 (0.22) 6 (0.09) 8 (0.4) IPR unknown 0.21 0.21 0.21 0.18 0.22 0.21 Has health insurance a 0.51 0.67 0.65 0.67 0.65 0.72 Nativity a Native-born 0.42 0.44 0.17 0.11 0.91 0.95 Foreign-born in US: o10 years 0.14 0.06 0.14 0.25 0.02 0.003 10 14 years 0.18 0.09 0.14 0.22 0.01 0.004 15+years 0.26 0.41 0.55 0.42 0.06 0.04 Spanish interview a,c 0.43 0.21 0.64 0.52 0.003 0.003 Smoking a Nonsmoker 0.67 0.58 0.63 0.75 0.57 0.49 Former 0.16 0.17 0.17 0.10 0.17 0.28 Current 0.17 0.25 0.20 0.15 0.26 0.23 Drinking a Abstainer 0.32 0.28 0.32 0.38 0.30 0.17 Former 0.12 0.15 0.09 0.10 0.20 0.17 Current 0.54 0.54 0.59 0.51 0.49 0.66 Activity a Inactive 0.72 0.70 0.75 0.72 0.69 0.58 Weight a Overweight 0.71 0.65 0.59 0.57 0.70 0.58 Men a 0.46 0.41 0.47 0.39 0.40 0.46 Age b 41 (0.18) 44 (0.89) 51 (0.62) 42 (0.62) 47 (0.20) 50 (0.08) Weighted data, standard errors adjusted for sampling design. a Significantly different by ethnoracial group. b Significantly different from whites. c Non-English interview.

58 B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 Table 2 Summary of F tests for addition of interaction terms to additive ordinal logit models (degrees of freedom) Adjusted Wald test for interaction terms ]Medical conditions ]Impairments Self-rated health F p F p F p Socioeconomic status (15,325) 3.95 0.000 6.66 0.000 3.00 0.0002 Culture (20,320) 3.40 0.000 3.78 0.000 2.53 0.0004 Health risks (30,310) 1.20 0.220 2.88 0.000 3.98 0.000 National Center for Health Statistics (2004) (n ¼ 129; 596). Cubans and Dominicans are more likely than Mexicans or Puerto Ricans to be, though the majority of all have resided in the U.S for 15 or more years. Dominicans are most likely to be recent, followed by Mexicans and Cubans. Cubans are most likely and Puerto Ricans least likely to have completed the interview completely or partly in Spanish. All Latino groups have lower SES profiles than whites, yet Mexicans stand out with the lowest levels of income, education, and health insurance. Latinos are less likely than whites to incur health risks due to smoking or drinking habits, but have appreciably higher levels of excess weight and physical inactivity. Dominican and Puerto Rican sub-groups are proportionately more female than whites and other Latino groups. Cuban and white sub-groups have older age structures, whereas Mexican and Dominican populations tend to be younger. Multivariate analyses We examine whether health differences among Latino sub-groups arise from sub-group differences in health determinants. We estimate three ordinal logit models in which the distributions of each health outcome are predicted by ethnic group membership, cultural, socioeconomic, and health risk determinants, and the interaction of ethnic group membership with one of the three sets of health determinants. Interaction terms are formed as the products of the relevant variables. All models control for the effects of age and sex. The additive models include all determinants and control variables. The additive and interaction models predicting self-rated health also include the numbers of medical conditions and functional impairments as control variables. The additive models (not shown) reveal the same health patterns observed in the bivariate analysis with two exceptions. The Cuban and Dominican advantage in having fewer medical conditions disappears. Controlling for the effects of all of the variables, reveals a Dominican disparity and a Cuban equivalence in the number of medical conditions compared to whites. Table 2 presents the significance tests of each interaction model relative to the baseline additive model, wherein each test statistic evaluates the contribution of the set of interaction terms to the additive model. The top two rows show that the interactions of ethnicity with socioeconomic determinants and with cultural determinants are significant, net of ethnicity, socioeconomic, cultural, health behavior, and control variables. The bottom row shows that the interaction of ethnicity with health risk determinants is significant in predicting variation in functional impairments and self-rated health, but not in predicting variation in the number of medical conditions. The interaction of ethnicity and socioeconomic status is generally stronger than the interactions of ethnicity and cultural determinants, and ethnicity and health risks. Compared to the interaction of ethnicity and health risks, the interaction of ethnicity and culture is stronger in predicting ethnic variation in medical conditions and functional impairments, but weaker in predicting self-rated health. The main effects and interaction terms for each of the three interaction models predicting number of medical conditions, number of impairments and self-rated health are presented in Tables 3 5 respectively. Socioeconomic determinants The effects of socioeconomic factors on health vary among Latino sub-groups (see Table 3). The contrasting effects of education and income indicate a complex set of relationships among ethnicity, socioeconomic status and health. The majority of the ethnicity and education interaction terms and the ethnicity and insurance interaction terms are positive coefficients, indicating higher educated Latinos and insured Latinos report poorer health than their less educated and uninsured counterparts. The majority of the ethnicity and income interaction terms, however, are negative coefficients, indicating that higher income Latinos report better health than lower income Latinos.

B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 59 Table 3 Interaction coefficients of ethnicity and socioeconomic determinants of health and main effects coefficients from ordinal logit regressions (other coefficients not shown) Table 4 Interaction coefficients of ethnicity and cultural determinants of health and main effects coefficients from ordinal logit regressions (other coefficients not shown) Medical conditions Functional impairments Self-rated health Medical conditions Functional impairments Self-rated health Main effects Ethnicity Mexican 0.71* 1.03* 0.22* Puerto Rican 0.42* 0.26 0.03 Cuban 0.63* 0.84* 0.19 Dominican 0.46 0.81 0.45 Black 0.16(*) 0.25* 0.40* Education 0.02* 0.04* 0.09* Poverty-income ratio (PIR) 0.03* 0.07* 0.10* PIR unknown 0.53* 1.11* 0.69* Has health insurance 0.01 0.11* 0.07* Interaction effects Mexican Education 0.03* 0.05* 0.03* PIR 0.002 0.01(*) 0.001 Insured 0.21* 0.20* 0.17* Puerto Rican Education 0.01 0.02 0.02 PIR 0.03* 0.04* 0.01 Insured 0.18 0.41* 0.21 Cuban Education 0.04(*) 0.02 0.02 PIR 0.03(*) 0.01 0.03 Insured 0.14 0.10 0.22 Dominican Education 0.04 0.01 0.06 PIR 0.02 0.11* 0.10(*) Insured 0.63(*) 0.97* 0.33 Black Education 0.01 0.01 0.01 PIR 0.01* 0.02* 0.02* Insured 0.05 0.06 0.02 National Center for Health Statistics (2004). pp0:05; ð Þ0:054pp0:10: Attending to only statistically significant coefficients reveals how the association of socioeconomic status with health outcomes varies among Latino sub-groups. Among Mexicans, the socioeconomic effects are primarily driven by education and insurance coverage, and suggest poorer health among higher SES Mexicans than among lower SES Mexicans. In contrast, the socioeconomic effects among Puerto Ricans are primarily driven by income and reveal better health among higher Main effects Ethnicity Mexican 0.19* 0.30* 0.41* Puerto Rican 0.06 0.19(*) 0.44* Cuban 0.45(*) 0.07 0.67(*) Dominican 0.42 0.03 0.08 Black 0.28* 0.18* 0.48* Spanish 0.22 0.12 0.27 Nativity: o10 years 0.43* 0.29* 0.07 10 14 years 0.36* 0.25 0.38 15+years 0.21* 0.31* 0.06 Interaction effects Mexican Spanish 0.32 0.20 0.25 o10 years 0.52* 0.68* 0.56* 10 14 years 0.37* 0.51* 0.89* 15+years 0.27* 0.25* 0.23(*) Puerto Rican Spanish 0.51* 0.13 0.09 o10 years 0.60* 0.65* 0.33 10 14 years 0.27 0.20 0.35 15+years 0.19 0.34* 0.12 Cuban Spanish 0.24 0.40 0.23 o10 years 0.66* 0.37 0.78(*) 10 14 years 0.68(*) 0.52 0.59 15+years 0.56* 0.33 0.12 Dominican Spanish 0.34 0.33 0.68 o10 years 0.44 1.48* 0.63 10 14 years 0.09 1.02 1.14 15+years 0.06 0.80 0.46 Black Non-English 0.21 0.57 1.18 o10 years 0.23 0.58* 0.57(*) 10 14 years 0.20 0.68* 0.53 15+years 0.24* 0.35* 0.17 National Center for Health Statistics (2004). pp0:05; ð Þ0:054pp0:10: SES Puerto Ricans. Among Cubans, socioeconomic status only marginally affects sub-group variation in the number of medical conditions, and education and income operate in contrasting directions. Better health

60 B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 Table 5 Interaction coefficients of ethnicity and health risk determinants of health and main effects coefficients from ordinal logit regressions (other coefficients not shown) Table 5 (continued ) Medical conditions Functional impairments Self-rated health Medical conditions Functional impairments Self-rated health Main effects Ethnicity Mexican 0.06 0.27* 0.38* Puerto Rican 0.32(*) 0.11 1.03* Cuban 0.004 0.65* 1.08* Dominican 0.86* 0.30 1.11(*) Black 0.26* 0.20* 0.50* Smoking Former 0.27* 0.23* 0.21* Current 0.16* 0.38* 0.53* Drinking Former 0.28* 0.45* 0.05 Current 0.03 0.10* 0.60* Inactive 0.09* 0.25* 0.50* Overweight 0.61* 0.52* 0.02 Interaction effects Mexican Smoking: Former 0.04 0.02 Current 0.16* 0.39* Drinking: Former 0.12 0.05 Current 0.12 0.35* Inactive 0.22* 0.43* Overweight 0.07 0.13 Puerto Rican Smoking: 0.10 0.39 Former Current 0.11 0.32 Drinking: Former 0.16 0.53* Current 0.36* 0.22 Inactive 0.22* 0.16 Overweight 0.17 0.06 Cuban Smoking: Former 0.003 0.04 Current 0.03 0.04 Drinking: Former 0.52* 0.002 Current 0.39(*) 0.01 Inactive 0.05 0.15 Overweight 0.12 0.15 Dominican Smoking: Former 0.66 0.15 Current 0.19 1.56* Drinking: Former 0.11 0.14 Current 0.73* 0.41 Inactive 0.29 0.23 Overweight 0.42 0.28 Black Smoking: Former 0.09(*) 0.27* Current 0.08(*) 0.25* Drinking: Former 0.09(*) 0.07 Current 0.04 0.39* Inactive 0.04 0.17* Overweight 0.10* 0.05 National Center for Health Statistics (2004). pp0:05; ð Þ0:054pp0:10: is reported among Cubans with lower levels of education and among Cubans with higher incomes. Both income and insurance coverage play stronger roles than education among Dominicans. Higher income Dominicans report fewer impairments, but are marginally more likely to rate their health more negatively than lower income Dominicans. Cultural determinants The results presented in Table 4 reveal that the effects of culture on health also vary among Latino sub-groups. Spanish-language generally does not distinguish among members of each Latino ethnic group. The one exception reveals that Spanish-speaking Puerto Ricans have more medical conditions than English-speaking Puerto Ricans, evidence against the cultural buffer hypothesis. In contrast to the effect of language, immigrant status and duration of residence in the US indicate significant differences among Latinos. The coefficients among Mexicans are significant, negative and monotonically decline with length of residence. The coefficients among Dominicans are also negative and decline monotonically with duration, however, only the coefficient for the most recent immigrant cohort is statistically significant. Consistent with the cultural buffer hypothesis, Mexican, and perhaps Dominican, immigrant health advantages decline with acculturation. Health disparities among, however, appear among Puerto Ricans and Cubans. Puerto Rican, especially the most recent cohort, have more medical conditions and functional impairments than their native-born counterparts. Similarly, Cuban have more medical conditions than Cubans

B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 61 born in the US. The negative interaction terms in the final column suggest that Latino tend to rate their health status more positively than US-born Latinos. This effect is significant among all cohorts of Mexican, but only significant among the most recent Cuban immigrant cohort. Health risk determinants The results of the final interaction model, interacting ethnicity with health risks, are presented in Table 5. The interaction model predicting ethnic variation in the effects of health risks upon number of medical conditions was not a significant improvement over the main effects model. Only the interaction models of functional impairment and self-rated health are presented. Compared to nonsmokers, smoking carries a health effect only among currently smoking Mexicans and Dominicans. Mexicans and Dominicans who smoke view their health status more positively, and Mexicans who smoke also report less functional impairment. Among Latinos who currently drink, compared to lifetime abstainers, Mexican rate their health as poorer, Cubans report marginally more impairments, and Dominicans and Puerto Ricans report fewer impairments. Cubans who no longer drink also report more functional impairment, whereas Puerto Ricans who no longer drink rate their health more positively. Inactive lifestyles are associated with better health among Mexicans, but more impairment among Puerto Ricans. The health effect of excess weight does not vary among Latino sub-groups. Conclusions Extent of Latino health disparities There is a mix of health advantages and disparities among Latinos, although this varies by health measure and national origin. Mexicans have health advantages over whites, indicating healthier chronic disease profiles. Puerto Ricans, in contrast, reveal health disparities across all health outcomes. Cubans and Dominicans reveal a mix of health disparities and health advantages, both reporting poorer subjective health ratings and each indicating a health advantage on one of the objective health measures. Health patterns clearly are different among Latino ethnic groups, a finding that challenges the utility of a pan-ethnic Latino category in health research. Ethnic health differences and their mechanisms Ethnic variation in the social determinants of health is observed and thus helps account for the ethnic variation in health outcomes. Each ethnic group is a unique composition of SES, cultural, and lifestyle characteristics. Moreover, the interaction models reveal that the relationships of social characteristics and health outcomes are contingent upon ethnicity. Mexicans present the most distinctive profile, the evidence consistent with the hypothesized paradox i.e., worse health among those with higher levels of SES and acculturation. The profile suggests that cultural factors operate as a health buffer, a health advantage moderated with acculturation. The evidence also suggests indirectly that migration selectivity may shape the health, acculturation and economic composition of the Mexican ethnic group. The health profile of adult Puerto Ricans is situated in clear contrast. Specifically, Puerto Ricans who have worse health tend to be concentrated among those who have lower levels of SES and acculturation. Although acculturation and economic assimilation are associated with better health among Puerto Ricans, migration selectivity may contribute unique impact. But selectivity may take a different form than among Mexicans. Spanish-speaking and lower SES Puerto Ricans who experience health problems may come to the US for medical care. As well, poor health and the need for medical care may keep the lower SES and less acculturated segments of the US Puerto Rican population from participating in any economic return migrations to Puerto Rico. Cubans and Dominicans provide less crisp profiles, but ones that look more similar to the Puerto Rican profile than the Mexican one. Better health outcomes are associated with higher levels of SES and acculturation. For Cubans, the effects of acculturation and SES may be more precisely measured because migration selectivity is constrained by US Cuba migration policy. In all three Caribbean island Latino groups, the positive effects of acculturation on health may be moderated through a concomitantly increasing exposure to negative health effects of lifestyle behaviors. Indeed, the primarily cameo appearance of health behaviors in shaping health outcomes is striking, especially given the current public health focus on promoting healthier individual lifestyle behaviors relative to other determinants. The primary goal of this research is to document and explain ethnic variation in health among Latino adults. The rationale underlying the research question rests on the arguments that ethnic heterogeneity is not fully appreciated in health research, non-mexican ethnic groups are inadequately examined, and thus conclusions regarding Hispanic disparities or health paradoxes may be overly simplistic. We show the complexity of health patterns in the Latino-origin population varies clearly with ethnicity. We further demonstrate that ethnic variation in social determinants of health significantly contributes to observed ethnic group differences in number of medical conditions, functional physical impairment, and self-rated health. An ob vious recommendation to the

62 B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 health research community is to improve research designs of racial/ethnic comparative health studies to more closely approximate the complexities of multiple race/ethnic groups and the effects of social determinants of health. Improved research designs presumably yield a greater opportunity to enhance the success of health intervention programs by a more efficient and effective delivery of group-specific health interventions. Conceptually, this research demonstrates that ethnicity is as critical as racial group or minority group membership in understanding andequalizingthedistributionofillnessandhealthinthe US population. References Abraido-Lanza, A. F., Dohrenwend, B. P., Ng-Mak, D. S., & Turner, J. B. (1999). The Latino mortality paradox: a test of the salmon bias and healthy migrant hypotheses. American Journal of Public Health, 89(10), 1543 1548. Amaro, H., & de la Torre, A. (2002). Public health needs and scientific opportunities in research on Latinas. American Journal of Public Health, 92(4), 525 529. Amaro, H., Russo, N. F., & Johnson, J. (1987). Family and work predictors of psychological well-being among Hispanic women professionals. Psychology of Women Quarterly, 11, 505 521. Arcia, E., Skinner, M., Bailey, D., & Correa, V. (2001). Models of acculturation and health behaviors among Latino to the US. Social Science & Medicine, 53, 41 53. Black, S. A., & Markides, K. S. (1993). Acculturation and alcohol consumption in Puerto Rican, Cuban-American, and Mexican-American women in the United States. American Journal of Public Health, 83(6), 890 893. Centers for Disease Control and Prevention. (1997). Update: prevalence of overweight among children, adolescents, and adults United States, 1988 1994. Morbidity and Mortality Weekly Report, 46, 198 202. Cobas, J. A., Balcazar, H., Benin, M. B., Keith, V. M., & Chong, Y. (1996). Acculturation and low-birthweight infants among Latino women. American Journal of Public Health, 86(3), 394 396. Falcon, L. M., & Tucker, K. L. (2000). Prevalence and correlates of depressive symptoms among Hispanic elders in Massachusetts. Journals of Gerontology, Series B, 55(2), s108 s117. Feinstein, J. S. (1993). The relationship between socioeconomic status and health: a review of the literature. Milbank Memorial Fund Quarterly, 71, 279 322. Finch, B. K., Hummer, R. A., Kolody, B., & Vega, W. A. (2001). The role of discrimination and acculturative stress in the physical health of Mexican-origin adults. Hispanic Journal of Behavioral Sciences, 23(4), 399 421. Finch, B. K., Hummer, R. A., Reindl, M., & Vega, W. A. (2002). Validity of self-rated health among Latino(a)s. American Journal of Epidemiology, 155(8), 755 759. Gardner, L. I., Stern, M. P., Haffner, S. M., Gaskill, S. P., Hazuda, H. P., Relethoford, J. H., & Eifler, C. W. (1984). Prevalence of diabetes in Mexican Americans: relationship to percent of gene pool derived from Native American sources. Diabetes, 33, 86 92. Guendelman, S., & Abrams (1994). Dietary intake among Mexican-American women: generational differences and a comparison with white non-hispanic women. American Journal of Public Health, 85, 20 25. Gundelman, S. (1998). Health and disease among Hispanics. In S. Loue (Ed.), Handbook of immigrant health (pp. 277 302). New York: Plenum Press. Hajat, A., Lucas, J. B., & Kington, R. (2000). Health outcomes among Hispanic subgroups: data from the National Health Interview Survey, 1992 95. Advance Data from Vital and Health Statistics 310. Hyattsville, MD: National Center for Health Statistics. Hayes-Bautista, D. E. (1992). Latino health indicators and the underclass model: from paradox to new policy models. In A. Furino (Ed.), Health policy and the Hispanic. Boulder, CO: Westview Press. Himmelgreen, D. A., Perez-Escamilla, R., Peng, Y., Martinez, D., & Wright, A. (2001). Length of time in US, acculturation status, and overweight and obesity among Latinos in two urban settings. American Journal of Physical Anthropology Annual, 2001, 81. House, J. S., Lepkowski, J. M., Kinney, A. M., Mero, R. P., Kessler, R. C., & Herzog, A. R. (1996). The social stratification of aging and health. Journal of Health and Social Behavior, 35, 213 234. House, J. S., & Williams, D. R. (2000). Understanding and reducing socioeconomic and racial/ethnic disparities in health. In B. D. Smedley, & S. L. Syme (Eds.), Promoting health: intervention strategies from social and behavioral research (pp. 81 124). New York: National Academy Press. Hummer, R. A., Rogers, R. G., Amir, S. H., Forbes, D., & Frisbie, W. P. (2000). Adult mortality differentials among Hispanic subgroups and non-hispanic whites. Social Science Quarterly, 81(1), 459 477. Hunt, L. M., Schneider, S., & Comer, B. (2004). Should acculturation be a variable in health research? A critical review of research on US Hispanics. Social Science & Medicine, 59, 973 986. Idler, E., & Angel, R. (1990). Self-rated health and mortality in the NHANES-I epidemiological follow-up. American Journal of Public Health, 80, 446 452. Kington, R. S., & Nickens, H. W. (2001). Racial and ethnic differences in health: recent trends, current patterns, future directions. In N. J. Smelser, W. J. Wilson, & F. Mitchell (Eds.), America becoming: racial trends and their consequences, vol. 1 (pp. 253 310). New York: National Academy Press. Krieger, N. (2000). Discrimination and health. In L. F. Berkman, & I. Kawachi (Eds.), Social epidemiology (pp. 36 75). New York: Oxford University Press. Landale, N. S., Oropesa, R. S., & Gorman, B. K. (2000). Migration and infant death: assimilation or selective migration among Puerto Ricans. American Sociological Review, 65, 888 909. Landale, N. S., Oropesa, R. S., Llanes, D., & Gorman, B. K. (1999). Does Americanization have adverse effects on health? Stress, health habits, and infant health outcomes among Puerto Ricans. Social Forces, 78(2), 613 630. Lorenzo, C., Serrano-Rios, M., Martinez-Larrad, M., Gabriel, R., Williams, K., Gonzalez-Villalpando, C., Stern, M., Hazuda, H., & Haffner, S. (2001). Was the historic

B.A. Zsembik, D. Fennell / Social Science & Medicine 61 (2005) 53 63 63 contribution of Spain to the Mexican gene pool partially responsible for the higher prevalence of Type 2 diabetes in Mexican-origin populations? Diabetes Care, 24, 2059 2065. Lynch, J., & Kaplan, G. (2000). Socioeconomic position. In L. F. Berkman, & I. Kawachi (Eds.), Social epidemiology (pp. 13 35). New York: Oxford University Press. Markides, K. S., & Coreil, J. (1986). The health of Hispanics in the southwestern United States. Public Health Reports, 101, 253 265. Marmot, M., & Wilkinson, R. G. (1999). Social determinants of health. New York: Oxford University Press. Massey, D. S. (1993). Latinos, poverty, and the underclass: a new agenda for research. Hispanic Journal of Behavioral Sciences, 15, 449 475. National Center for Health Statistics. (2004). National health interview survey, 1997 2001. Computer files and data documentation. Padilla, Y. C., Boardman, J. D., Hummer, R. A., & Espitia, M. (2002). Is the Mexican American epidemiologic paradox advantage at birth maintained through early childhood? Social Forces, 80(3), 1101 1124. Portes, A., & Rumbaut, R. G. (1990). Immigrant America: a portrait. Berkeley, CA: University of California Press. Randolph, W. M., Stroup-Benham, C., Black, S. A., & Markides, K. S. (1998). Alcohol use among Cuban- Americans, Mexican-Americans, and Puerto Ricans. Alcohol Health & Research World, 22(4), 265 270. Salgado de Snyder, V. N. (1987). Factors associated with acculturative stress and depressive symptomatology among married immigrant Mexican women. Psychology of Women Quarterly, 11, 475 488. Scribner, R. (1996). Paradox as paradigm: the health outcomes of Mexican Americans. American Journal of Public Health, 86(3), 303 305. Singer, B. H., & Ryff, C. D. (2001). New horizons in health: an integrative approach. New York: National Academy Press. Sorlie, P. D., Backlund, E., Johnson, N. J., & Hogat, E. (1993). Mortality by hispanic status in the United States. Journal of the American Medical Association, 270(20), 2464 2468. Sundquist, J., & Winkleby, M. A. (2000). Cardiovascular risk factors in Mexican American adults: a transcultural analysis of NHANES III, 1988 1994. American Journal of Public Health, 89(5), 723 730. Tienda, M. (1989). Puerto Ricans and the underclass debate: evidence for structural explanations of labor market performance. Annals of the American Academy of Political and Social Science, 501, 105 119. Torres, A., & Rodriguez, C. E. (1991). Latino research and policy: the Puerto Rican case. In E. Melendez, C. B. Rodriguez, & J. B. Figeroa (Eds.), Hispanic in the labor force: issues and policy. New York: Plenum. Vega, W. A., & Amaro, H. (1994). Latino outlook: good health, uncertain prognosis. Annual Review of Public Health, 15, 39 67. Weeks, J. R., Rumbaut, R. G., & Ojeda, N. (1999). Reproductive outcomes among Mexican-born women in San Diego and Tijuana: testing the migration selectivity hypothesis. Journal of Immigrant Health, 1(2), 77 90. Williams, D. R. (2001). Racial variations in adult health status: patterns, paradoxes, and prospects. In N. J. Smelser, W. J. Wilson, & F. Mitchell (Eds.), America becoming: racial trends and their consequences, vol. 1 (pp. 371 410). New York: National Academy Press. Winkleby, M. A., Robinson, T. N., Sundquist, J., & Kraemer, H. C. (1999). Ethnic variation in cardiovascular disease risk factors among children and young adults: findings from the third national health and nutrition survey, 1988 1994. Journal of the American Medical Association, 281(11), 1006 1007. Wolff, C. B., & Portis, M. (1996). Smoking, acculturation, and pregnancy outcome among Mexican Americans. Health Care for Women International, 17(6), 563 573. Zambrana, R. E., & Logie, L. A. (2000). Latino child health: need for inclusion in the US national discourse. American Journal of Public Health, 90(12), 1827 1833. Zsembik, B. A. (2000). The Cuban ethnic economy and labor market outcomes of Latinos in metropolitan Florida. Hispanic Journal of Behavioral Sciences, 22, 223 236. Zsembik, B. A., & Llanes, D. (1996). Generational differences in educational achievement among Mexican Americans. Social Science Quarterly, 77, 363 375.