| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Demography and Cedeplar, Universidade Federal de Minas Gerais, Brazil.
2 Office of Population Research, Princeton University, NJ.
Address correspondence to Cassio M. Turra, PhD, Universidade Federal de Minas Gerais, Department of Demography, Rua Curitiba, 8 Andar, Belo Horizonte, Minas Gerais 30170-120, Brazil. E-mail: turra{at}cedeplar.ufmg.br
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Methods. We used Poisson regression models based on data from the 1989 to 1994 waves of the National Health Interview Survey, with linked mortality through 1997, to estimate death rates for Hispanics and non-Hispanic Whites by age, gender, and socioeconomic status.
Results. Deaths rates varied significantly (p <.05) by education and income for Whites and Hispanic subgroups defined by nativity (U.S. born and foreign born) and nationality (Mexican, Puerto Rican, and other Hispanic). However, with the exception of Puerto Ricans, the effects of education were significantly smaller for Hispanics than for Whites. The ethnic differences in mortality patterns by income were not statistically significant.
Discussion. The findings reveal that the mortality advantage for Hispanics is concentrated at lower levels of socioeconomic status, with little or no advantage at higher levels. We propose several mechanisms related to immigration and assimilation patterns that may underlie these patterns of mortality.
IN recent decades, epidemiologists and social scientists have devoted considerable attention to estimating and understanding disparities in health and survival in the United States. This research has often been bifurcated: Some scholars have focused on racial and ethnic differences, whereas others have been primarily interested in the social gradient (i.e., the seemingly ubiquitous associations between progressively higher levels of social status and better health and greater longevity). Separation of these two areas of research is artificial and unproductive. Interactions among race or ethnicity, socioeconomic status (SES), and health are complex, and researchers' understanding of racial differences in health status and health trajectories is likely to depend on their ability to explicate these linkages (see, for example, Williams, 2005
).
The potential interplay between ethnicity and SES comes to the forefront in analyses of a pattern of health termed the Hispanic paradox. Despite the unfavorable socioeconomic profile of U.S. Hispanics, most Hispanic groups, with the exception of Puerto Ricans, have levels of adult mortality similar to or better than those of non-Hispanic Whites (Elo, Turra, Kestenbaum, & Ferguson, 2004; Hummer, Rogers, Amir, Forbes, & Frisbie, 2000; Rosenwaike, 1987; Sorlie, Backlund, Johnson, & Rogot, 1993). This mortality advantage is most apparent among middle-aged and elderly Hispanics; for example, the ratio of death rates of Hispanics to non-Hispanic Whites exceeds 1.00 in the twenties, falls below unity at around age 45, decreases to between 0.75 and 0.90 in the sixties, and persists at these levels through old age (Elo & Preston, 1997; Elo et al., 2004; Hummer, Benjamins, & Rogers, 2004; Turra, 2004).
There are large variations in mortality within the Hispanic population, with foreign-born and certain regional subgroups—most notably Mexicans, Central Americans and South Americans—most likely to experience this mortality advantage (Hummer et al., 2000; Markides & Eschbach, 2005; Palloni & Arias, 2004). In the presence of controls for SES, the mortality advantage of persons of Mexican origin, the largest Hispanic group in the United States, widens notably and the disadvantage of Puerto Ricans largely disappears (Hummer et al., 2000). These findings are in sharp contrast to those for African Americans, who, on average, have a similar socioeconomic profile to Hispanics, yet have considerably higher mortality rates (Hoyert, Heron, Murphy, & Kung, 2006; National Center for Health Statistics [NCHS], 2005). Adjustment for the lower SES of African Americans typically reduces, but does not eliminate, their survival disadvantage vis-à-vis other groups (Williams & Collins, 1995).
Despite unabated interest among researchers in social inequalities in health, and continued efforts to understand the Hispanic paradox, social scientists know surprisingly little about social gradients in health among Hispanics. The outcome that has received the most attention in health disparity research is mortality, both because it poses fewer measurement problems and because the Hispanic paradox is most apparent with regard to survival. Many studies of racial and ethnic inequalities in mortality include measures of SES (most frequently education, income, or occupational status) in statistical models as control variables. However, because researchers typically assume SES to have constant effects across ethnic groups, these findings provide little information about the nature of SES disparities in survival within the Hispanic population (Hummer et al., 2000; Rogers, Hummer, Nam, & Peters, 1996; Singh & Siahpush, 2002). There have been several exceptions, however. One study (Lin, Rogot, Johnson, Sorlie, & Arias, 2003) noted that education differences in life expectancy at the lower levels of schooling and income appear to be smaller for Hispanic men than for non-Hispanic men. In contrast, the estimates presented in two analyses of death rates that used occupational class as a measure of SES (Muntaner, Hadden, & Kravets, 2004; Wei et al., 1996) and one that used income (Sorlie et al., 1993) suggested similar differences for Hispanics and non-Hispanic Whites. Unfortunately, none of these analyses provided statistical comparisons of the SES differentials across ethnic categories.
Estimates of social disparities with regard to health outcomes other than survival for the Hispanic population are also scarce. Based on data for African Americans, Whites, and Hispanics reported in Pamuk, Makuc, Heck, Reuben, and Lochner (1998), Williams (2005) argued that SES differences in health within the Hispanic population (and within the other two groups) were large and more substantial than the differences across racial categories. This assertion contrasts sharply with recent findings from three U.S. surveys (Goldman, Kimbro, Turra, & Pebley, 2006). In this study, education levels were only weakly related, or were not related at all, to several health behaviors (smoking and drinking) and health outcomes (obesity, work-related limitations, and depressive symptoms) for U.S. Hispanic adults overall (as well as for the subgroup of persons of Mexican origin). These shallow gradients for Hispanics differed significantly from the steeper ones characterizing non-Hispanic Whites, challenging the presumption that SES inequalities in health are universal and suggesting a second paradoxical health pattern for the Hispanic population.
In the present analysis, we attempted to broaden the field's understanding of social inequalities in the United States by assessing mortality differences by education and income within the Hispanic population. Our goal was twofold. The first was to evaluate whether the relatively flat education gradients in health-related measures identified by Goldman and colleagues (2006) extended to mortality. The second was to ascertain whether and how differences in social disparities in mortality between Hispanics and non-Hispanic Whites were related to the Hispanic paradox. Although we examined mortality patterns across the adult life cycle, we focused on the older ages, because the vast majority of deaths in the U.S. population occur after age 50 (Edwards & Tuljapurkar, 2005), and the Hispanic mortality advantage appears to be largest at these ages.
Extrapolation of findings regarding the magnitude of the social gradient from the measures of health reported by Goldman and colleagues (2006) to survival is problematic, because mortality risks reflect long-term exposure to a broad range of behaviors, acute and chronic health conditions, and access and use of health care. There are at least two reasons to speculate that SES differentials in mortality for Hispanics will be more prominent than those identified by Goldman and colleagues. First is the relatively high prevalence of certain conditions among Hispanics (such as diabetes, obesity, and liver diseases) that may be more frequent among persons of lower SES (Escarce, Morales, & Rumbaut, 2006; Markides, Rudkin, Angel, & Espino, 1997; NCHS, 2005). The second is the fact that many Hispanics, particularly those of low SES, are without health insurance and hence are less likely than others to use health services, have a regular place of care, or obtain adequate treatment. For example, estimates from the 2003 National Health Interview Survey (NHIS) revealed that about 35% of Hispanics younger than age 65 and about 5% of Hispanics aged 65 and older lack any type of health insurance. Moreover, within the Hispanic population, poor Hispanics are between two and three times as likely to lack health insurance coverage and about twice as likely not to have received medical care during the previous year because of cost considerations compared with their not-poor counterparts (Schiller, Adams, & Coriaty, 2005). Therefore, we hypothesized that mortality among Hispanic adults would vary significantly by education and income, albeit less than it would among non-Hispanic Whites.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
Since 1986, information for NHIS respondents aged 18 and older has been linked with the National Death Index to create the NHIS Multiple Cause of Death public use data files. The matching methodology uses a score constructed on the basis of 12 criteria, which combines several personal identifiers such as social security number, first name, last name, and father's surname. Potential matches between the National Death Index and the NHIS records classify into one of five mutually exclusive categories based on number and type of items matched. Ascertainment of the vital status of individuals who fall into classes that do not fulfill the requirements to be considered either true or false matches is made based on cutoff scores calculated from two independent calibration samples (NCHS, 2000). Biases in the matching procedure usually result from missing social security numbers, incorrect recording of ethnic names, emigration of foreign-born individuals, and changing surnames among women. The NCHS estimates that about 94% of the deaths of women and 97% of the deaths of men are correctly classified in the linked files. Among non-Whites, these percentages are about 85% and 88%, respectively (NCHS, 2000).
Of the 512,073 persons aged 18 and older interviewed in the 1989 to 1994 surveys, there were 30,266 presumed deaths from 1989 through 1997. We excluded about 4% of the original NHIS sample because of insufficient information to perform the linkage procedure. To avoid duplicates, we excluded another 7,712 Hispanics from a supplemental sample in 1992, most of whom were interviewed in 1991. As described below, we restricted the present study to Hispanics and non-Hispanic native-born Whites aged 25 and older. Among the persons in the sample satisfying these criteria, 331,079 remained after the exclusion of those with missing data on explanatory variables; this figure included 308,939 survivors and 22,140 deaths.
Explanatory Variables
Explanatory variables for the statistical models comprised age (single years), gender, race/ethnicity, educational attainment, and income. We defined all variables as of the NHIS interview date and, except for age, assumed them to remain constant throughout the follow-up period. Because this assumption is untenable for educational attainment among young adults, we restricted the analysis to persons aged 25 and older.
Information on race and ethnicity was based on self-identification: Researchers first asked respondents about their racial background and then if their national origin or ancestry could be described by any of eight Hispanic origin categories. Respondents who indicated any of these Hispanic designations were considered to be Hispanic. Additional information for the construction of nativity status was obtained from a question pertaining to number of years lived in the United States. Our analysis focused on two groups: Hispanics and native-born non-Hispanic Whites (hereafter, Whites). In later stages of the analysis, we considered two subdivisions of the Hispanic group: (a) U.S.-born versus foreign-born Hispanics; and (b) Mexicans, Puerto Ricans, and other Hispanics. We followed convention by classifying Puerto Ricans born outside the 50 states or the District of Columbia as foreign born. The sample size of deaths was not sufficiently large to permit further subdivision of these variables or joint classification of nativity and nationality.
SES comprises educational attainment and family income. Educational attainment was classified into five categories: 0 to 8 years (omitted category), 9 to 11 years, 12 years, 13 to 15 years, and 16 or more years. With regard to income, 86.5% of respondents in our analysis sample reported annual family income in categories ranging from less than $1,000 to $50,000 and more; the remainder reported annual family income as a dichotomy: less than $20,000 or $20,000 and more. Because of the large ranges within these categories, we imputed values for income based on data from the Current Population Surveys for the years 1989 to 1994. For each of the 5 years of the NHIS, we assigned respondents the mean family income of persons of the same age, gender, ethnic group, and income category who were interviewed in the Current Population Survey of the same year. We then reclassified this imputed income variable into quartiles based on the combined sample of Hispanics and Whites.
Analytic Strategy
We used Poisson regression models to estimate the number of deaths during the follow-up period as a function of person-years of exposure, SES, ethnicity, age, and gender. We based these estimates on survey commands in Stata (StataCorp, 2003) that adjusted for clustering and stratification of the NHIS in the estimation of the standard errors (NCHS, 2004). Each respondent's exposure began at the date of interview and terminated at the time of death or the end of 1997, whichever came first.
Age was a continuous variable. We estimated two sets of models, one that included education levels and the second that replaced education levels with income quartiles. Within each of these two sets, we fit three models that included different specifications of the Hispanic variable. The first model included Hispanics as a single group; the second distinguished between U.S.-born and foreign-born Hispanics; and the third identified Mexicans, Puerto Ricans, and other Hispanics. Thus, we estimated a total of six Poisson models.
The models included interaction terms as well as main effects of the explanatory variables. Because our objective was to explore differences in the SES gradient by ethnicity, all models included interaction terms between SES (education or income) and the relevant Hispanic variable(s). In exploratory analyses, we tested for the inclusion of all remaining two-way interaction terms involving age, gender, SES, and ethnicity, and the three-way interaction term among age, SES, and ethnicity. The three-way interaction term was not significant for either education or income. After excluding this term, we retained those two-way interaction terms with p values less than.05 (all p values are based on two-sided tests unless indicated otherwise). The interaction terms between age and ethnicity and between age and SES were significant in all models, so we included them in both the education and income models. The interaction terms between age and gender and between gender and SES were significant in the income but not the education models, and we thus included them in only the income models.
Because of the many interaction terms included in each of the models, it was not possible to present the coefficients here; however, they are available upon request. Instead, we present predicted death rates by education and by income. In order to examine the nature of the SES differentials over the life cycle, we calculated the predicted death rates separately for three ages—30, 50, and 70—selected to represent young, middle, and older ages. For a given model, we obtained the predicted death rates by considering all combinations of values for the categories of gender, ethnicity, SES, and the three selected ages; setting the interaction terms accordingly; and using the coefficients of the model to obtain the predicted number of deaths for the specified person-years of exposure. We summarized the magnitude of the education differentials within each ethnic group by calculating the ratio of the death rate for persons with 9 to 11 years of schooling to that for persons with 13 to 15 years of schooling; we avoided the extreme categories because of the relatively small number of Hispanics in the highest education level and Whites in the lowest. With regard to income, we present the ratio of the death rate associated with the bottom quartile to that for the top quartile. Note that because of the inclusion of interaction terms between gender and SES in the income but not the education models, the summary measure of education differentials was identical for men and women but the measure of income differentials for men consistently exceeded that for women. Other researchers have reported differences in the relationship of income and mortality between the sexes (Backlund, Sorlie, & Johnson, 1996; Miller & Paxson, 2006). Backlund and colleagues speculated that these patterns may reflect differences between men and women in their exposure to environmental hazards according to income level, in their psychosocial reactions to particular income levels, in the permanency of income, or in the distribution of death by cause. In order to provide further insights into ethnic differences in mortality gradients in the middle and older ages, we show the predicted death rates for men by level of education at ages 50 and 70; patterns for women were very similar to those for men.
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
; Beckett, 2000; House et al., 1994). Although mortality varied significantly by SES for most of the ethnic groups considered here, the SES patterns varied among groups. As shown in Figure 1 and Table 2, the differences in mortality by education were generally smaller for Hispanic groups than for Whites. One notable exception occurred for Puerto Ricans, for whom the education gradients in mortality were generally as steep as those for Whites. As shown in Table 3, the ethnic differences in mortality patterns by income were less notable than those for education. Statistical tests confirmed that, with the exception of Puerto Ricans, the interaction terms comparing the set of education coefficients for a given Hispanic subgroup with the corresponding coefficients for Whites were statistically significant (p <.05); that is, the education coefficients were significantly different between Whites and most Hispanic groups. In contrast, none of the interaction terms pertaining to ethnic differences in the income coefficients were significant.
|
|
|
; Rosenwaike, 1987) and that the Hispanic mortality advantage is larger at older ages than in midlife (Markides & Eschbach, 2005). Second, the mortality advantage was more notable for foreign-born Hispanics than for Hispanics born in the United States, particularly at older ages. Third, the mortality advantage was most prominent for Other Hispanics, whereas Puerto Ricans were generally disadvantaged relative to Whites, especially at the younger and middle ages. Our findings also underscored a pattern that has received little attention to date: The Hispanic mortality advantage pertains primarily to persons of lower SES. At middle and older ages, Hispanics with little schooling or low income experienced lower mortality than their White counterparts, whereas Hispanics in the highest education and income categories had generally similar or higher death rates than Whites. Indeed, although sample sizes were relatively small for Hispanics in the highest education category, the results suggest that U.S.-born Hispanics with 16 or more years of education have higher mortality at adult ages than similarly educated Whites (p <.05, one-sided test evaluated at the mean age of the sample).
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
), a negative association between income and mortality may be present even in the absence of mechanisms linking higher income to better health and survival.
The estimates provide an important insight into the relationship between observed SES patterns in mortality and the Hispanic mortality paradox. The concentration of the mortality advantage for Hispanics at lower levels of SES, with little or no advantage seen at higher levels, leads to flatter SES gradients for Hispanics than for Whites. Few researchers have paid attention to how the mortality advantage varies by level of SES. There are two noteworthy exceptions. The first used data from the San Antonio Heart Study to demonstrate that mortality differences between Mexican Americans and Whites are modest in the middle and higher SES categories but larger at lower levels of SES (Wei et al., 1996). A second analysis (McWilliams, Zaslavsky, Meara, & Ayanian, 2004) demonstrated that Hispanics with health insurance experience similar death rates to Whites, but that uninsured Hispanics have lower mortality than both uninsured and insured Whites.
Our statistical analysis underscores the importance not only of incorporating interactions between SES and ethnicity in analyses of mortality, but also of including interactions with age. In the present analysis, there were significant interactions between age and each of the remaining explanatory variables (SES, ethnicity, and gender). Unfortunately, some previous studies pertaining to ethnic differences in mortality either employed proportional hazards models, which are based on the assumption that the effects of the covariates are constant across age, or failed to consider important interaction terms (Rogers et al., 1996; Sorlie et al., 1993; Wei et al., 1996). Indeed, many restricted the statistical models to the main effects of the covariates. Thus, some earlier estimates of ethnic or SES differences in longevity may be misleading because they likely mask variations by age and SES.
What processes underlie the differential SES patterns in mortality? In particular, what leads Hispanics of low SES to have lower death rates than their White counterparts and Hispanics to have shallower SES gradients in mortality than Whites? We propose three sets of explanations: (a) immigration related, (b) assimilation, and (c) data errors.
Two immigration-related processes, which are similar to explanations proposed in an earlier study (Goldman et al., 2006), may be particularly important for the foreign-born population. The first is that SES differentials in Mexico and other parts of Latin America are weak or reversed for some health-related variables, such as smoking and obesity, in part because the poor are less able to afford such luxuries as cigarettes and high-calorie diets (Kain, Vio, & Albala, 2003; Rivera & Sepulveda, 2003; Vazquez-Segovia, Sesma-Vazquez, & Hernandez-Avila, 2002). Migrants from these areas are apt to bring these patterns with them when they move to the United States. The second pathway comprises two potential selection mechanisms that are related to what have been termed the healthy migrant and salmon bias effects: Migration to the United States may be selective of those in better health (or those with healthier behaviors), and return migration to Latin America may be selective of those in poorer health, especially for persons of low SES.
Although researchers know little about how these aspects of migration differ between Puerto Ricans and other Hispanics, it is possible that the steeper SES gradients found here for Puerto Ricans result both from Puerto Ricans experiencing weaker (or different) migration-related selection processes as a consequence of their status as American citizens (Landale, Oropesa, Llanes, & Gorman, 1999) and from stronger differentials in health and health-related behaviors being present in Puerto Rico as compared with Mexico and other Latin American countries. The potential uniqueness of the migration stream from Puerto Rico receives support from a recent analysis of education selectivity among immigrants to the United States from 32 countries. The results demonstrated that Puerto Ricans were negatively selected for schooling, whereas immigrants from all other countries were more educated than their nonmigrant counterparts (Feliciano, 2005). It is important to recognize that these migration-related pathways may affect mortality for U.S.- as well as foreign-born Hispanics (e.g., through intergenerational transmission of behaviors or of health status).
We refer broadly to the second set of explanations as assimilation. Immigrants from less favored ethnic groups often have little alternative but to assimilate into disadvantaged segments of U.S. society, increasing the likelihood that they and their children experience the restricted access to health services and poor health outcomes typical of residents in their neighborhoods. Moreover, whereas some immigrants may retain their healthy behaviors after arrival in the United States, discrimination and the lack of opportunity faced by other immigrants may lead them to adopt detrimental behaviors that are more common in the United States than in their home countries (e.g., smoking) and to experience the negative health consequences of stress. Indeed, there is evidence that Hispanic immigrants are more likely to engage in negative health behaviors with increasing time spent in the United States: Increasing acculturation among Hispanics has been associated with undesirable dietary behaviors and increased alcohol and drug use (Escarce et al., 2006; Lara, Gamboa, Kahramanian, Morales, & Bautista, 2005). Although previous studies provide no insights into whether these changes in behavior are related to SES, it is plausible that, if Hispanic immigrants of higher SES are as likely (or more likely) to experience increased stress and racism and adopt detrimental behaviors as Hispanic immigrants of lower social position, the assimilation process may weaken some of the pathways that link higher education and income to better health in other ethnic groups.
The third set of mechanisms pertains to errors in the data. Death rates for Hispanics, particularly at lower SES, may be biased downward because of both age misreporting (Preston, Elo, & Stewart, 1999) and errors related to matching death records to the NHIS (Elo et al., 2004). An earlier study (NCHS, 2000) suggested that the matching algorithm may be especially problematic for non-Whites because these individuals are more likely to have missing social security numbers and complex (e.g., hyphenated) surnames. Moreover, omission of deaths during the follow-up period may be more common among foreign-born individuals, because they are more apt to emigrate and die outside the United States than native groups (Turra, Elo, Kestenbaum, & Ferguson, 2005).
How do these various mechanisms account for the finding that the mortality advantages of lower SES Hispanics are apparent only at middle and older ages? An important part of the answer may pertain to the different causes of death that dominate at younger versus older ages. At younger ages, deaths result largely from external causes, such as homicide and accidents, which are strongly associated with environmental factors. In contrast, at middle and older ages, chronic illnesses, which are related to detrimental health-related behaviors as well as to health status at younger ages, are the major causes of death. Thus, the negative impact on mortality of assimilation into poor neighborhoods is likely proportionately larger at younger ages, and immigration-related processes likely offer survival protection primarily at middle and older ages. These suppositions are consistent with studies that have found excess external-cause mortality for Hispanics compared to Whites (Hummer et al., 2000; Rogers et al., 1996), as well as lower death rates for Hispanics as compared with Whites from some leading chronic diseases, including cancer and cardiovascular disease (Rogers et al., 1996; Sorlie et al., 1993). Puerto Ricans are an exception to the finding for chronic disease: Their risks of dying from heart disease (Hummer et al., 2000; Rosenwaike, 1987), liver diseases (Rosenwaike, 1987), and cancer (particularly among women; Hummer et al., 2000), are comparable to or higher than those of Whites, in part because they have less favorable health-related behaviors compared with other Hispanics (Perez-Stable et al., 2001; Rogers, 1991). In addition, Puerto Ricans report higher levels of psychological distress and mobility limitations than Mexicans and other Hispanics (Bratter & Eschbach, 2005; Cho, Frisbie, Hummer, & Rogers, 2004; Hummer et al., 2004). These unique patterns for Puerto Ricans may emanate from both distinct migration processes and different assimilation experiences.
This analysis provides new insights into two important and interrelated phenomena: the Hispanic mortality paradox and social inequalities in mortality. Our findings underscore previous results indicating that the Hispanic mortality advantage does not apply to all Hispanics; rather, the advantage is concentrated among the foreign-born from particular national origins and is present only at middle and older ages. Our estimates further suggest that efforts to understand this epidemiological paradox should focus on examining why Hispanics of lower SES experience most of the advantage and Hispanics with additional education benefit relatively little in comparison with Whites. Researchers also need to identify the health-related pathways that offer survival protection to many Hispanic groups of low SES (but apparently not to Puerto Ricans). For example, do most of the benefits stem from the adoption of healthier behaviors or rejection of harmful habits among Hispanics as compared with Whites? Are these patterns driven by immigration-related mechanisms or by behaviors adopted (or not adopted) by Hispanics living in the United States? Recent evidence suggests that social disparities in some health-related variables in Mexico and other immigrant-sending countries are changing in ways that resemble patterns in industrialized countries, whereby more educated and wealthier individuals have healthier behaviors (Bobak, Jha, Nguyen, & Jarvis, 2000; Filozof, Gonzalez, Sereday, Mazza, & Braguinsky, 2001). The likely consequence of these trends is a widening of SES differences in mortality in Latin America and subsequently for immigrants and their descendants, changes that will require researchers to pay increased attention to variations in social gradients in health and mortality among ethnic groups in the United States.
|
Footnotes |
|---|
Received for publication July 13, 2006. Accepted for publication September 12, 2006.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. D. Barger and L. C. Gallo Ability of Ethnic Self-Identification to Partition Modifiable Health Risk Among US Residents of Mexican Ancestry Am J Public Health, November 1, 2008; 98(11): 1971 - 1978. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
