Conceptualizing Ethnicity: How Dimensions of Ethnicity Affect Disparities in Health Outcomes Among Latinxs in the United States

Abstract

Methods

Using regression methods to analyze data from the 2006 Portraits of American Life Study, we examined how attributional and relational dimensions of ethnicity affect: 1) intragroup differences in Latinx mental and physical health status, as measured by feelings of worthlessness and self-rated health, respectively; and 2) intergroup differences between Latinxs and non-Hispanic Whites in these health outcomes.

Results

Latinxs have higher odds of feelings of worthlessness and lower odds of self-reporting good/excellent health compared with non-Hispanic Whites. Additionally, intragroup differences in health are observed among Latinxs, conditioned on attributional or relational dimensions of ethnicity.

Conclusion

Multidimensional measures of ethnicity that distinguish between characteristics associated with ethnicity (attributional) or race (relational) offer a nuanced explanation of health disparities by revealing aspects of ethnicity that shape health outcomes differently, contributing to the goals of health equity.

Background

Biomedical and public health research on health disparities often relies on large data sets that employ force-choice categories of ethnicity.1,2 Using an aggregate self-reported measure, however, masks important distinctions that may affect health differently between affiliated group members.3-5 The development and application of theoretically informed and empirically validated measurements that capture multiple dimensions of ethnicity is crucial for an accurate understanding of causal pathways that associate ethnicity with health. Improving measurements of ethnicity is one recognized goal of public health critical race praxis,6,7 which seeks to foster a more nuanced approach to the study of racial and ethnic health disparities for improved health and social equality.

At 18% of the total population, Latinxs make up the largest racial/ethnic minority group in the United States.8 Latinxs are also increasingly diverse, as the Mexican-origin share is declining while the proportion from Puerto Rico and Central America is on the rise.8 The heterogeneity of this group offers a unique opportunity to examine how variations in ethnic/racial characteristics, including national origin, language, and skin color, shape health outcomes.9 Research that examines how ethnicity affects Latinx health generally finds an association,10 but group heterogeneity and variation in operationalization and selection of such characteristics condition the direction and magnitude of the effect.11-14 As Cuevas and colleagues caution, the health implications of the relationship between ethnicity and Latinx health remain circumspect, due to the “inconsistent use of race and skin color measures, and omission of a wider range of immigration-related and contextual factors.”10,p213 Drawing from Ford and Harawa’s1 conceptualization of ethnicity, which distinguishes between attributional dimensions—which emphasize personal and group identities and sociocultural characteristics—and relational dimensions— which underscore social stratification and refers to the group’s social location in hierarchies (Table 1), we offer a comprehensive empirical examination of how multiple dimensions of ethnicity contribute to disparities in health among Latinxs in the United States.

Table 1. Conceptual and operational framework for the relationship between dimensions of ethnicity and health status^a.

Dimensions of Ethnicity	Brief Description	Adopted Measures	Predicted Relationship to Health
Attributional	Related to sociocultural characteristics; Emphasizes individual and group identities; Often associated with nativity and culture	Parental nativity status; Language use; Religious identification	The more respondents report association with Latinx cultures and place of origin (eg, foreign born parents, non-English language use, Catholic religion), the better their health will be, compared with non-Hispanic Whites
Relational	Emphasizes positioning within the US racial hierarchy; Often associated with phenotypical characteristics	Racial identification; Choice; Skin color; Racial/ethnic identity; Salience; Racial/ethnic misclassification	The more respondents report association with racialized status (eg, darker skin, racialized label choice (ie ‘other’), high racial/ethnic identity salience, racial/ethnic misclassification), the worse their health will be, compared with non-Hispanic Whites

a. Table informed by Ford and Harawa.¹

Our exploratory quantitative analysis of seven dimensions of ethnicity captured in the Portraits of American Life Study (PALS) highlights needed empirical evidence to clarify the theoretical assumptions and empirical implications of Ford and Harawa’s conceptualization of ethnicity1 and contributes to the research on Latinx health. Findings support Ford and Harawa’s call to consider both attributional and relational dimensions of ethnicity, thereby identifying the sociocultural and structural pathways that shape Latinx health. Moreover, observed divergent trajectories of attributional dimensions that are generally protective of Latinx mental health and relational dimensions that result in adverse outcomes are consistent with racialization, underscoring the salience of race within relational measures.

Theoretical Framework

Extant empirical research and systematic reviews collectively suggest multiple dimensions of ethnicity help explain intra-/inter-group variation in Latinx health, though the direction of the relationship varies by attributional or relational dimensions and type of health outcome.10 Attributional dimensions associated with nativity and culture generally tend to have a protective effect.6,15 For example, Vega and colleagues found Latinx immigrants report lower rates of psychiatric disorders than their US-born counterparts,11 a finding consistent with the immigrant paradox.14 Similar findings are observed for self-rated health, a global health indicator.16 Lommel and Chen revealed a strong and consistent association between acculturation, as measured by US citizenship and English language proficiency, and poor self-rated health for Latinxs17; other research has found reliance on Catholic religion linked to improved Latinx health.18

Relational dimensions of ethnicity1 associated with racial classification, including constructs such as skin color or racial self-identification, are generally often associated with adverse health effects.4,19 In their study of depressive symptoms among adolescent Latinxs, Ramos and colleagues found a significant, positive association among self-identified Black Afro-Latinas when compared with self-identified White Latinas.20 Likewise, non-Hispanic Whites reported better self-rated health outcomes than Latinxs who identify as White (Hispanic-Whites).16 A negative association between skin color and health has been found among Latinxs, such that the darker one’s skin color is, the worse their reported health.21 Research on intragroup differences revealed that having a strong racial/ethnic identity was associated with improved physical and mental health among Latinxs, although a negative racial group evaluation among Black Americans was associated with an increase in depressive symptoms.22,23 Recent studies on racial classification have demonstrated negative mental health outcomes for Latinxs whose race/ethnicity is misclassified by others,24,25 suggesting that inconsistent racial/ethnic classification may increase stress.26

Findings collectively demonstrate that indicators of relational ethnicity are generally associated with negative health outcomes, whereas indicators of attributional ethnicity are mixed but tend to point to better health outcomes. These effects, however, have been determined separately, limiting a comprehensive picture of their combined impact. Thus, we ask: How do attributional and relational dimensions of ethnicity explain inter/intragroup disparities in the health outcomes of Latinxs, when compared with non-Hispanic Whites? Informed by prior research, we hypothesize that the effect of attributional dimensions of ethnicity are protective, whereas relational dimensions of ethnicity will have adverse impacts (Figure 1). Our analysis brings to bear needed empirical evidence to clarify the theoretical assumptions undergirding Ford and Harawa’s conceptualization of multidimensional ethnicity1 and contributes to a better understanding of Latinx mental and physical health.

Figure 1. Predicted probability of feelings of worthlessness among Latinx for attributional and relational ideal types.

Methods

We used the 2006 Portraits of American Life Study (PALS), a nationally representative, face-to-face interview survey of 2,610 non-institutionalized, English- or Spanish-speaking civilian households in the contiguous United States. Participants were aged ≥18 years and survey questions focused on topics including racial and ethnic identity, religion, and health.27,28 Surveys were conducted in English or Spanish and had a response rate of 58%. Our use of PALS data sampling weights brings PALS data in line with the US Census Bureau’s American Community Survey (ACS) three-year average estimates for 2005-2007. The 2006 PALS data are ideal for this project because it oversamples Latinxs and includes multiple measures of race/ethnicity. All procedures were conducted in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000. The study was reviewed and approved by the institutional review board (#2016-99) at UC Merced.

Measures

Dependent Variables

Feelings of Worthlessness

We selected self-reported feelings of “worthlessness” within the last two weeks (1=yes) as an indicator of mental wellbeing, which is often used in scales to capture symptoms of psychological distress. This indicator is a useful measure of mental wellbeing given the underreporting of mental health constructs like depression among Latinxs due to stigma and cultural differences.29

Self-Rated Health

We examined self-rated health as a widely used indicator of overall health and wellbeing.30 Self-rated health is measured using a 3-point scale where 1=poor or fair 2=good 3=very good or excellent.

Independent Variables

Self-reported Race/ethnicity

Respondents were asked, “What race or ethnic group do you consider yourself?” We restricted our analytical sample to non-Hispanic Whites (herein Whites) and Hispanics (herein Latinxs).

Dimensions of Ethnicity

The PALS race/ethnicity module provided multiple dimensions of ethnicity. We limited our analysis of dimensions of ethnicity to the Latinx population as they compared with Whites for three reasons. First, in the United States, Latinxs are a racialized ethnic minority group whose ethnic identity is associated with structural exclusion,31 whereas Whites are characterized as the dominant cultural group whose ethnic identity is “symbolic” or “optional” and distinct from structural exclusion.32 White ethnic identity is symbolic because identifying as such does not require engaging in associated cultural or social practices, and optional because there is some choice in the matter.33 Second, not all dimensions were asked/reported for Whites. Finally, we observed a lack of variation on specific dimensions, such as nativity, asked/reported for Whites. For example, 90.74% of Whites were very light/light skinned, 83.21% non-Catholic, 99.10% spoke English, 90.82% had US-born parents, and 89.88% said they are consistently racially classified. Thus, for each dimension of ethnicity—both attributional and relational measures—the reference group is Whites (White=0).

Attributional measures1 were operationalized as dichotomous variables among the Latinx subsample consisting of: ancestry, ie, foreign-born or US-born parents14; English language use, ie, English as the primary language at home or not17; and, Catholic, ie, as self-reported religion vs all others.18

Relational measures among the Latinx subsample were operationalized to include: skin color, a categorical variable capturing interviewer coded skin color by matching perceived skin tone (1 “very light” to 4 “dark”) to the survey’s skin tone instrument13; racial/ethnic misclassification, a dichotomous variable of whether the respondent reported that other Americans believe their race/ethnicity is something other than their own self-report vs consistently classified24,25; racial/ethnic identity salience, a dichotomous variable of whether respondents answered “very” or “somewhat important” (vs “only a little” and “not important”) when asked how important their self-selected racial/ethnic identification is to their sense of who they are22,23; and racial choice, a dichotomous variable of self-reported census racial categories21 that we restricted to Latinxs who racially identify as “White” (41.7%) vs “other” (51.6%), given the small n’s in other responses (17 as American Indian, 11 as Black, 6 as don’t know, 6.7%).

Confounders

Full models are adjusted for age (continuous), sex (1=male), and socioeconomic status (SES, a standardized composite measure of income, educational attainment, and employment status). The composite measure allowed us to compare SES across two population groups with distinct socioeconomic distributions. SES scores ranged from 1.42 standard deviations below the mean to 1.56 standard deviations above the mean.

Analytical Strategy

Analyses were conducted in Stata 13 accounting for the PALS complex survey design. Observations with missing data for the outcomes or predictors (6% of sample) were excluded leaving 1,733 observations in the analytic sample (n=450 Latinxs, n=1,283 Whites). Using chi-square tests and independent t-tests, we first assessed bivariate relationships between self-reported ethnicity, social demographics, and outcome measures (Table 2), then analyzed bivariate relationships between dimensions of ethnicity to highlight intra-ethnic differences in health (Table 3). We next ran a series of binary logistic regression models for feeling worthless (Table 4) and ordered logistic regression models for self-rated health (Table 5) to examine multivariate relationships regressing each dimension of ethnicity on the health outcomes in separate models. We produced 19 models for each dimension on the two outcomes, first running unadjusted models and then adjusted regression models for key control variables: age, sex, and socioeconomic status. Finally, we ran two comprehensive models for each outcome including relational and attributional dimensions together. Results of the model are not shown but available upon request. The full models do not include “Latinx race” because that question was only asked of Latinx respondents. No dimensions were significant in these models, due in part to the large number of dimensions included and sample size.

Table 2. Social demographic and health status among non-Hispanic White and Latinx in PALS 2006^a.

Variable	Range	Full Sample, N=1733	White, n=1283	Latinx, n=450	Pb
Social Demographics
Male	0, 1	.485	.482	.507	.452
Age	18 – 80	46.416 (.695)	47.395 (.718)	39.940 (1.023)	<.001
Socioeconomic status	-1.266 – 1.557	.131 (.039)	.178 (.041)	-.180 (.056)	<.001
Health Status
Felt worthless	0, 1	.143	.136	.192	.041
Self-rated health	1-5				.002
Poor/fair		.208	.197	.280
Good		.247	.243	.278
Very good/excellent		.545	.561	.442

a. Means/proportions of variables are weighted; standard errors in parentheses.

b. P comparing Latinx with White study participants

Table 3. Proportion of Latinx respondents reporting feeling depressed and feelings of worthlessness and self-rated health by dimensions of ethnicity^a, n=450.

	Feelings of Worthlessnessb	Self-Rated Healthb
		Poor/Fair	Good	Very Good/Excellent
Attributional Variables
Ancestry
US born parents	.188	.316	.314	.370
Foreign-born parents	.193	.269	.267	.464
Language
English primary language	.202	.301	.272	.428
Non-English primary language	.152	.198	.305	.498
Religion
Catholic	.202	.297	.265	.439
Non-Catholic	.178	.255	.299	.447
Relational Variables
Skin color
Very light	.218	.236	.300	.463
Light	.189	.213	.272	.515
Medium	.283	.315	.193	.493
Dark	.129	.334	.307	.359a
Racial/ethnic misclassification
Consistent racial/ethnic classification	.161a	.301	.266	.433
Racial/ethnic misclassification	.298	.209	.322	.470
Racial/ethnic identity salience
Important	.205	.277	.250	.473
Not important	.157	.288	.354	.358
Latinx racial choice
White Latinx	.182	.262	.307	.431
Other Latinx	.201	.297	.252	.451

a. Statistically significant, P<.10.

b. Chi-squared tests performed to determine significant differences within dimension.

Table 4. Logistic regression model of feelings of worthlessness by dimensions of ethnicity, PALS 2016, N=1733.

	Worthlessnessa
	Unadjusted, OR (95%,CI)	Adjustedb, OR (95%,CI)
General race/ ethnic classification	Model 1d	Model 2
Latinx	1.508* (1.018, 2.234)	1.130 (.721, 1.771)
Attributional variables
Ancestry	Model 3	Model 4f
US-born parents Latinx	1.519c (.934, 2.470)	1.098 (.638, 1.890)
Foreign-born parents Latinx	1.471 (.765, 2.829)	1.244 (.622, 2.488)
Language	Model 5 c	Model 6f
English primary language Latinx	1.605d (1.047, 2.462)	1.196 (.737, 1.941)
Non-English language Latinx	1.136 (.545, 2.370)	.868 (.411, 1.834)
Religion	Model 7	Model 8f
Catholic Latinx	1.602c (.975, 2.633)	1.210 (.694, 2.108)
Non-Catholic Latinx	1.371 (.777, 2.418)	1.014 (.572, 1.800)
Relational variables
Skin color	Model 9 c	Model 10f
Very light Latinx	1.768 (.825, 3.788)	1.353 (.649, 2.820)
Light Latinx	1.475 (.758, 2.868)	1.204 (.575, 2.524)
Medium Latinx	2.509 d (1.214, 5.187)	1.904c (.897, 4.040)
Dark Latinx	1.025 (.578, 1.820)	.706 (.366, 1.363)
Racial/ethnic misclassification	Model 11e	Model 12f
Consistent racial/ethnic classification	1.218 (.741, 2.001)	.873 (.507, 1.502)
Racial/ethnic misclassification	2.691d (1.480, 4.892)	2.335e (1.240, 4.397)
Racial/ethnic identity salience	Model 13c	Model 14f
Important	1.639d (1.059, 2.537)	1.192 (.738, 1.926)
Not important	1.178 (.674, 2.059)	.956 (.486, 1.878)
Latinx racial choice	Model 15	Model 16f
White Latinx	1.414 (.835, 2.394)	1.108 (.634, 1.938)
Other Latinx	1.594 d (1.012, 2.511)	1.149 (.663, 1.994)

a. All models use non-Latinx Whites as the referent group.

b. Adjusted models (column 2) control for age, sex, and socioeconomic status.

c. p <.10, d. P<.05, e. P<.01, f. P<.001, significance of F-statistic reported for each model

Table 5. Ordered logistic regressions of self-reported health on dimensions of ethnicity, PALS 2016, N=1733.

	Self-Rated Healtha
	Unadjusted, OR (95%,CI)	Adjustedb, OR (95%, CI)
Race	Model 1e	Model 2f
Latinx	.622e (.461, .840)	.690d (.507, .938)
Attributional variables
Ancestry	Model 3f	Model 4f
US-born parents Latinx	.672d (.472, .956)	.784 (.555, 1.108)
Foreign-born parents Latinx	.490f (.338, .711)	.462e (.289, .738)
Language	Model 5e	Model 6f
English primary language Latinx	.578f (.432, .774)	.668d (.488, .915)
Non-English language Latinx	.832 (.424,1.636)	.784 (.379,1.621)
Religion	Model 7e	Model 8f
Catholic Latinx	.597e (.429, .830)	.699c (.489, 1.000)
Non-Catholic Latinx	.661d (.443, .986)	.675c (.440, 1.036)
Relational variables
Skin color	Model 09e	Model 10f
Very light Latinx	.712 (.437, 1.162)	.791 (.482, 1.296)
Light Latinx	.851 (.547, 1.325)	.887 (.578, 1.360)
Medium Latinx	.663 (.361, 1.218)	.706 (.390, 1.278)
Dark Latinx	.460f (.324, .653)	.536e (.350, .821)
Racial/ethnic misclassification	Model 11e	Model 12f
Consistent racial/ethnic classification	.585e (.412, .831)	.675d (.477, .956)
Racial/ethnic misclassification	.760 (.439, 1.316)	.738 (.436, 1.250)
Racial/ethnic identity salience	Model 13e	Model 18f
Important	.677d (.502, .913)	.802 (.596, 1.081)
Not important	.505 e (.329, .777)	.475e (.305, .739)
Latinx racial choice	Model 15e	Model 16f
White Latinx	.629d (.431, .917)	.678c (.458, 1.005)
Other Latinx	.616e (.440, .863)	.701d (.493, .996)

a. All models use non-Latinx Whites as the referent group.

b. Adjusted models (column 2) control for age, sex, and socioeconomic status.

c. P <.10, d. P<.05, e. P<.01, f. P<.001, significance of F-statistic reported for each model.

Following the estimation of these models, we used the post-estimation techniques (Stata’s margins command) to estimate predicted probabilities for likelihood of feeling worthlessness and excellent/very good health by holding constant levels of different combinations of dimensions in order to reveal how they interact together. We then plotted these estimates (using Stata’s marginsplot command). Informed by Table 1, our combinations of dimensions, or what we consider ideal types are: high relational/high attributional; high relational/low attributional; low relational/high attributional; low relational/low attributional. Latinx respondents with high relational dimension of ethnicity were those with dark skin, consistent racial classification (ie, self-identified and how others identify), and felt race was important to them. Those with high attributional dimensions of ethnicity were Latinx respondents who did not primarily use English at home, had foreign-born parents, and were Catholic.

Results

Table 2 presents descriptive statistics. Latinxs are significantly younger (mean age = 39.94, p<.001) and have significantly lower SES than Whites (.180 SD below mean, p<.001). For example, the average education in the sample was some college, consistent with the average education for Whites, but for Latinxs the average education was a high school diploma. The majority of the sample (64.1%) and Whites owned a home (67.2%), while the majority of Latinxs did not own a home (56.4%). Just over 14 percent of the total sample reported feeling worthless within the last two weeks. Latinxs were significantly more likely than Whites to report feeling worthless (19.2% compared with 13.6%, respectively). More than 50% of the sample self-rated their health as very good or excellent (54.5%), whereas more than one in four Latinxs rated their health as poor or fair (28.0%).

Table 3 shows dimensions of ethnicity and their relationship to health among Latinxs. Latinxs who experience racial/ethnic misclassification reported that they feel worthless at a marginally significantly higher rate (29.8%) than those whose racial/ethnic classification is confirmed by others (P<.10). Dark-skinned Latinxs reported poor or fair health at a marginally significantly higher rate (33.4%) than lighter skinned Latinxs (P<.10). On balance, however, dimensions do not differ markedly among Latinxs by dimensions of ethnicity.

Table 4 shows odds ratios and confidence intervals for separate logistic regressions of feelings of worthlessness for the nine dimensions of ethnicity, revealing numerous intergroup differences compared with Whites. Unadjusted models reveal that Latinxs reported odds of feeling worthless that are 1.5 times greater than Whites (P<.05), and greater within-group variation. Models relating to attributional ethnicity show that Latinxs whose primary language is English (OR=1.6 P<.05) and who identify as Catholic (OR=1.6 P<.10) were markedly more likely to report feeling worthless compared with Whites. In general, findings modestly suggest that Latinxs who are more acculturated or have a weaker connection to attributes of Latinx culture are at greater risk for mental health problems.

Table 4 also shows that medium-skinned Latinxs’ odds of having feelings of worthlessness were two and a half times greater than Whites (OR=2.5; P<.05). Latinxs who racially identify as other (OR=1.6; P<.05), who identify their race/ethnicity as salient (OR=1.6; P<.05), or whose race/ethnicity is misclassified by others (OR=2.7 P<.01), were significantly more likely to report feeling worthless compared with Whites. Findings suggest that Latinxs affiliated with a racialized-minority position were most at-risk for feelings of worthlessness. However, when adjusting for age, sex, and SES, the only significant association was that Latinxs whose racial/ethnic category was misclassified (OR=2.3, P<.01) or with a medium-skin tone (OR=1.9 P<.1) had higher odds than Whites of feeling worthless.

Table 5 displays odds ratios for self-rated health. Unadjusted models reveal the odds of reporting better health was lower for Latinxs than Whites (OR=.62; P<.01), and display greater within-group variation. Models pertaining to attributional dimensions of ethnicity show that Latinxs who identify as Catholic (OR=.60; P<.01) or non-Catholic (OR=.66; P<.05), whose primary language is English (OR=.58; P<.001), whose parents are foreign-born (OR=.49; P<.001) or not (OR=.67; P<.05) were all significantly less likely than Whites to report better health. These findings for self-rated health are more mixed than those predicting worthlessness.

The odds of dark-skinned Latinxs reporting better health were half that of Whites (OR=.46; P<.001). Latinxs who racially identify as other (OR=.62; P<.01) or White (OR=.63; P<.05) reported worse health compared with Whites, as do those who believe their race/ethnicity is not important (OR=.51; P<.01), and those who believe their race/ethnicity is important (OR=.68, P<.05). Additionally, those whose racial classification was confirmed by others (OR=.59 P<.01) had markedly lower odds than Whites of reporting better health. These findings suggest mixed results for relational (racialized) dimensions of ethnicity and self-rated health. Adjusted models indicate that having foreign-born parents (OR=.46; P<.01), speaking English as a primary language (OR=.67; P<.05), being dark-skinned (OR=.54; P<.001), being consistently racially classified (OR=.68; P<.05), indicating that one’s racial/ethnic identity is not important (OR=.48; P<.01), and racially identifying as other (OR=.70; P<.05) continued to be markedly and negatively associated with Latinxs’ self-reported health, when compared with Whites, net of the controls. These findings suggest a clear association between select relational and attributional dimensions of ethnicity and negative self-rated health.

Finally, we ran models complete with available dimensions asked of White and Latinx respondents to ascertain how different combinations matter for health. We then estimated predicted probabilities of feelings of worthlessness and excellent/very good health for four attributional and relational ideal types: 1) high relational (high racialization), high attributional (low acculturation); 2) high relational (high racialization), low attributional (high acculturation); 3) low relational (low racialization), high attributional (low acculturation); 4) low relational (low racialization), low attributional (high acculturation).

Figure 1 depicts the predicted probability of having feelings of worthlessness among Latinx for four attributional and relational ideal types. All were significant, which means the probabilities were significantly different than zero. As expected, Latinx respondents with low attributional dimensions (high acculturation) had high probabilities of reporting feelings of worthlessness, but the highest probability across all ideal types was among those who had low attributional and low relational scores (Pr=.23; P<.001). This suggests acculturation more strongly influences the relationship to feelings of worthlessness than relational dimensions when all attributes are used in the model.

Figure 2 depicts the predicted probability of reporting excellent or very good health among Latinx for attributional and relational ideal types. Latinx respondents with high relational but low attributional dimensions of ethnicity had the lowest probability of reporting excellent or very good health (Pr=.32, P<.001), while those with low relational but high attributional dimensions of ethnicity had the highest probability of reporting excellent or very good health (Pr=.57, P<.001).

Figure 2. Predicted probability of excellent/very good health among Latinx for attributional and relational ideal types.

Discussion

This study offers a more comprehensive understanding of intra/intergroup variations in Latinx health. Our analysis reveals that attributional and relational dimensions of ethnicity condition health differently, highlighting nuances that are not reflected in the broader definition of ethnicity.

Relational dimensions indexing “a groups location within the United States social hierarchy of socially constructed, groupings (ie, races)”1,p253 were generally predictive of feelings of worthlessness. Findings suggest that relational dimensions of ethnicity were associated with adverse mental health among Latinxs, whereas self-rated health was less straightforward. Overall, though, we contend that the process of racialization conditions negative health outcomes for the subgroup of Latinxs who share a minority status that reflects their disadvantaged placement within the US racial hierarchy.

In contrast, attributional dimensions that capture “the unique sociocultural characteristics of groups”1,p252 are associated with better health outcomes. Yet, Latinxs who were more acculturated (eg, English speaking, US citizenship) reported increased feelings of worthlessness. These findings are generally consistent with the Latinx health paradox.14 Though findings regarding self-rated health were less pronounced, as consistent with past research that questions the validity of the measure for Latinx,10,32 the results overall, and from the comprehensive models and predicted probabilities more specifically, emphasize how attributional and relational dimensions operate differently and give support for our study hypotheses. Collectively, this suggests Latinxs at most risk for health problems are those who score high in relational dimensions and low in attributional dimensions—or Latinx who are highly racialized and acculturated.

Study Limitations

While PALS data were beneficial for this exploratory analysis because it allowed us to examine multiple dimensions relating to race and ethnicity, we were limited in our analysis to a one-item measure of mental health (worthlessness), which is typically used in scales consisting of multiple items that measure depressive symptoms and psychological distress.34 Nevertheless, considering the under-reporting of depression among Latinxs,29 reporting feelings of worthlessness might actually capture a larger proportion of respondents with psychological distress. Additional objective mental and physical health indicators should also be explored, given validity issues related to self-rated health as well.32 Future research should additionally expand analyses to other populations, such as Asian Americans or White ethnic groups, to determine whether results are generalizable.

Conclusion

In conclusion, results suggest a cautionary note to public health scholars who use a broad definition of ethnicity35 and support Ford and Harawa’s appeal to distinguish attributional from relational dimensions of ethnicity.1 In keeping with critical race theory,36 we further suggest decoupling the concept of race from ethnicity to emphasize the structural underpinnings of the former and cultural underpinnings of the latter.36 As argued by critical race and ethnic scholars, Valdez and Golash-Boza, “it is incomplete to conclude that the positioning of groups along one structural dimension (ie, “race”) can reasonably predict their life chances, because those who identify as such are likely to vary across other salient dimensions as well (ie, ethnicity).”36,p24 Ultimately, a multidimensional approach has the potential to better illuminate why health disparities exist, which is especially needed to address immigrant health,37 and thereby contributes to critical race public health praxis’ goal of health equity. 4,5,7,37