Introduction

With implications for chronic disease and mortality, alcohol-related problems represent a threat to population health. For Hispanics in the US, drinking outcomes are often tied to acculturation [1,2,3], a process which involves reconciling two cultures: a culture of heritage and the culture of the US. Prior literature has shown support for a statistically-significant relationship between acculturation measures and a range of drinking outcomes for US Hispanic groups [4,5,6], especially women, and this relationship has been documented with language-based measures, acculturation scales, and demographic variables (e.g., nativity, years lived in the US) utilized as proxies of acculturation.

Results from two meta-analyses of acculturation and drinking in US Hispanics indicate that studies with comprehensive measures of acculturation (which incorporate elements of cultural identification and values) report smaller effects on drinking outcomes, compared to studies that utilize measures centered around language [5, 6]. Although acculturation encompasses changes in practices, identifications, and values in both receiving and heritage cultures [7], the link with drinking outcomes for US Hispanics is most prominent for language-based acculturation measures and is observed primarily with measures of cultural acquisition, as opposed to cultural retention [6]. English language acquisition may facilitate access to the “mainstream” US culture [6], potentially leading to greater adoption of US practices, identifications, and values. In addition to its role in the process of acculturation, English use/proficiency may also serve as a gauge of exposure to and immersion in English-speaking settings in the US.

Despite well-documented relationships between acculturation-related measures and health outcomes, it is less clear how or why these relationships exist [8]. Adopting specific beliefs, values, or behaviors of US society, insofar as they promote drinking, has been presumed to influence the drinking patterns of Hispanics with immigrant backgrounds [9,10,11,12,13]. This explanation may be more relevant for Hispanic women who experience changes in “gender-related drinking norms” [4]. Emerging hypotheses assert that, beyond acculturation and culture, other factors in the social environment are responsible for the observed changes in drinking and mental health among Hispanics with the passage of time in the US [14]. One key factor in the social environment is ethnic discrimination.

Findings from a landmark qualitative investigation identify “‘othering’ and discrimination as potential pathways through which the health of Hispanic immigrants and their descendants erodes” [15]. Among US Hispanics, the prevalence of experiences of ethnic discrimination has been estimated at 30.0% [16], or as high as 79.5% [17], with variation based on heritage country, sex, age, geographic region, and acculturation profile [17, 18]. A systematic review of 97 studies reported that experiencing racial/ethnic discrimination is positively associated with alcohol-use related problems [19]. In US Hispanics specifically, research also links experiences of racial/ethnic discrimination with alcohol use disorder and alcohol-related problems, with variation based on characteristics such as sex, nativity, and heritage [20, 21].

Acculturation-related measures, including English use, are positively associated with reports of ethnic discrimination in US Hispanic adults [17]. For many US Hispanics, the process of acculturation entails making sense of their status as a member of an ethnic minority in US society, often involving experiences of othering, exclusion, or discrimination [15]. As such, a measure of “acculturation” may simultaneously capture exposure to, or perception of, discrimination, and “exposure to discrimination, not acculturation per se, may account for the observed declines in health or increases in risky health behaviors among Hispanics” [14]. The present study aims to test this hypothesis in the context of alcohol use disorder in US Hispanic immigrants, evaluating perceived ethnic discrimination as a potential mediator in the relationship between English use/proficiency and alcohol use disorder.

Built on frameworks connecting discrimination and health outcomes [22, 23] for immigrant populations [15], the conceptual model for the study is depicted in Fig. 1. This model rests on the notion that identifying and making sense of discriminatory experiences is a “learned process” [15] occurring with exposure to life in the US, with English language acquisition accompanied by experiences of, and perception of, ethnic discrimination. The model is also based on documented associations between discrimination and alcohol use disorder among US Hispanic groups [18, 20], as well as a vast body of research reporting associations between English use/proficiency and drinking outcomes in US Hispanics [4,5,6].

Fig. 1
figure 1

Conceptual model depicting the hypotheses in the study. The “+” symbol indicates a positive association

Full size image

All hypotheses are tested separately for men and women, in light of well-documented sex differences in drinking and acculturation profiles [4,5,6]. Overall, past research has documented a modest relationship between acculturation-related measures and drinking outcomes in Hispanic men, with some studies reporting non-significant findings; for women, however, past research has supported a robust link between acculturation-related measures and drinking outcomes [4]. The present study’s hypotheses are depicted in Fig. 1 and presented below.

Hypotheses

Among non-US-born Hispanic adult men and women who reported past-year alcohol consumption (after adjusting for age, education, family income, marital status, and family history of alcohol-related problems):

  • Hypothesis 1 (H1) Higher BAS-English (Bidimensional Acculturation Scale-English use/proficiency) scores will be associated with significantly higher odds of self-reported perceived ethnic discrimination.

  • Hypothesis 2 (H2) Self-reported perceived ethnic discrimination will be associated with significantly higher odds of meeting criteria for past-year DSM-5 alcohol use disorder.

  • Hypothesis 3 (H3) Higher BAS-English scores will be associated with significantly higher odds of meeting criteria for past-year DSM-5 alcohol use disorder.

  • Hypothesis 4 (H4) Higher BAS-English scores will be associated with significantly higher odds of meeting criteria for past-year DSM-5 alcohol use disorder, in a model that includes self-reported perceived ethnic discrimination.

  • Hypothesis 5 (H5) Self-reported perceived ethnic discrimination will act as a mediator in the relationship between BAS-English and past-year DSM-5 alcohol use disorder.

Methods

Data source and sample

This study utilizes data from the National Epidemiologic Survey on Alcohol and Related Conditions-III (NESARC-III, 2012–2013), a cross-sectional, nationally-representative study with a target population of the civilian, noninstitutionalized adult population in the 50 US states and District of Columbia [24]. Participants in NESARC-III were identified via multistage probability sampling, with individual counties serving as the primary sampling units, followed by census-defined blocks, dwelling units, and individuals, with oversampling in areas with moderate or high representation of ethnic minorities [25].

Data were collected through in-person interviews (with Spanish questionnaires available), using computer-assisted interviewing and automated questionnaires. Details about the methodology of NESARC-III have been previously published [25]. Access to NESARC-III is restricted and controlled by the National Institute on Alcohol Abuse and Alcoholism (NIAAA). Approval to use the limited-access data from NESARC-III was received after submitting a data user agreement and a human subjects research exemption issued by an Institutional Review Board.

The present study encompassed the non-US-born Hispanic participants in NESARC-III (n = 2222) who self-reported any consumption of alcohol within the past year. The study was restricted to individuals reporting past-year alcohol consumption to focus on the association between acculturation and problematic drinking, as opposed to drinking status [4]. Hispanic/Latino ethnicity was assessed by participants’ self-report (i.e., “Are you of Hispanic or Latino origin?”), and self-reported place of birth was used to identify the non-US-born participants for the study’s analytic sample. Participants born in a US territory (e.g., Puerto Rico) were classified in NESARC-III as not born in the US; as such, any Hispanic participant not born in the 50 US or District of Columbia was considered non-US-born. Listwise deletion was employed due to the negligible proportion of missing data (less than one percent of missing data across the variables in the study).

Measures

Outcome variable

The outcome variable in the present study was past-year DSM-5 alcohol use disorder. This dichotomous variable indicated whether a participant met criteria for DSM-5 alcohol use disorder in the past year, as assessed through the Alcohol Use Disorder and Associated Disabilities Interview Schedule-5 (AUDADIS-5). The procedural validity of the diagnoses obtained with the AUDADIS-5 in NESARC-III is reported elsewhere [26, 27].

Predictor variable: English use/proficiency

NESARC-III utilized the Bidimensional Acculturation Scale (BAS) [28], a validated 24-item instrument for measuring language use, proficiency, and media consumption, with each item scored on a four-point Likert scale. Twelve questions are asked regarding English, and the same 12 questions are asked regarding Spanish. Taking the mean of the items yields two sets of scores (one score for English and one for Spanish), each with a range of 1–4.

While the score on the English portion of the BAS is often conceptualized as a measure of acculturation, the score on the Spanish portion is often considered a measure of enculturation [29]. The present study’s sample of Hispanic immigrants exhibited relatively low variation on the Spanish portion of the BAS, with approximately 93.7% of the sample at or above the cut-off score for a “high level of adherence” to the Spanish orientation [28]. In the present study, only the English BAS score (hereafter referred to as BAS-English) was utilized. In these data, internal consistency was 0.98 for the BAS-English scale. Results of principal component factor analysis indicated that a one-factor solution explained 79.8% of the variance in the set of 12 items (with factor loadings ranging from 0.81 to 0.94), suggesting a unidimensional scale. The 12-item BAS-English score was accordingly treated as a unidimensional continuous variable.

Variable modeled as a mediator

Self-reported perceived ethnic discrimination was measured by NESARC-III using a modified version of the Experiences of Discrimination (EOD) questionnaire, “a valid and reliable self-report measure of racial discrimination” [30]. The EOD questionnaire asked about the frequency of experiences of ethnic discrimination with respect to six different situations: (1) accessing healthcare/health insurance; (2) receiving treatment or care; (3) in public, on the streets, in stores, or in restaurants; (4) obtaining a job or housing, in admission to a school or training program, or in the courts or with police; (5) being called a racist name; (6) experiencing verbal or physical abuse or threats. It is important to note that the EOD questionnaire in NESARC-III asks Hispanic participants to report discrimination experienced specifically “because you are Hispanic or Latino.” This reduces the possibility of conflating pre-immigration discrimination (in immigrants’ countries of origin) with the post-immigration experiences specifically related to the status ascribed to respondents’ Hispanic or Latino identity in the US. The time-frame for the measure of ethnic discrimination utilized in the present study was prior to the past year, in consideration of the temporal order between predictor (i.e., discrimination before past year) and outcome [31].

Each item in the questionnaire was dichotomized in the present study (i.e., “never” responses were coded with zero, while any other response, such as “almost never”, “sometimes”, “fairly often”, or “very often” was coded with one), consistent with prior literature [32, 33]. The six dichotomous variables were then used to compute an overall dichotomous variable indicating whether or not the participant reported experiencing discrimination “because you are Hispanic or Latino” at any frequency, in any setting [34]. As part of sensitivity analyses, however, a count variable was also created for the number of settings in which discrimination was reported (score range 0–6) [34], and another variable was created to sum the frequency of experiences reported in all settings (score range 0–24). Finally, a variable for discrimination in lifetime (as opposed to prior to the past year only) was also utilized for sensitivity analyses.

Other measures

Sociodemographic controls

Consistent with prior studies involving alcohol or drug use in the adult US population [35, 36], five sociodemographic variables were utilized: age (in years); sex (male/female); family income; educational attainment; and marital status. While age, family income, and educational attainment were utilized as continuous variables in multivariable analyses, they were presented in categories for the purpose of descriptive statistics. Family income (a variable with 21 categories) represents the total, combined family income self-reported in the past year, including income from social service programs. Educational attainment (consisting of 14 categories) represents the highest grade or year of school completed. Participants’ current marital status includes the following designations: married or cohabitating; widowed, divorced, or separated; and never married.

Family history of alcohol-related problems

Family history of alcohol-related problems was included as a covariate, in consideration of genetic and familial influences in alcohol use disorder [37]. NESARC-III asked participants whether any of their parents (blood/natural or adoptive), siblings (blood/natural) or partners (through marriage or cohabitation) was an “alcoholic or problem drinker” or had ever experienced problems as a result of alcohol use. Family history of alcohol-related problems was measured dichotomously (yes vs. no), similar to other studies [38,39,40].

Statistical methods

All analyses were stratified by sex. For sample characteristics (with unweighted data), cross-tabulations were used to compute percentages for categorical variables, while summary statistics were used to compute means and standard deviations for continuous variables. Multivariable binomial logistic regression analyses were utilized to test hypotheses one through four and to compute adjusted odds ratios (AOR) with 95% confidence intervals (CI). Analyses were computed with Stata 15.1 SE, and all models accounted for the complex survey design in NESARC-III.

Hypothesis five was tested through the KHB method for statistical mediation [41,42,43], in a single-mediator model [44]. The KHB method is appropriate for tests of mediation in models with binary outcomes, such as in the present study. The KHB method decomposes the effect of the predictor on the outcome, considering the mediating variable; that is, the total effect is separated into a direct and indirect effect. The proportion of confounding is also presented, as an estimation of the mediating effect. The efficacy of the KHB method has been supported through Monte Carlo simulations [42], and the method has been regarded as robust against unobserved heterogeneity.

The mediation analyses in the present study utilized the user-written KHB command [41] with the survey weight and statistical controls (age, educational attainment, family income, marital status, and family history of alcohol-related problems). The 95% confidence intervals were estimated with bootstrapping (20,000 replications), utilizing a user-written program to enable bootstrapping on weighted data. As part of sensitivity analyses, estimates of the total, direct, and indirect effects, with 95% confidence intervals, were also computed in three alternative manners: (1) using Stata’s GSEM (generalized structural equation modeling) and NLCOM (nonlinear combination of estimators) commands (Stata’s GSEM with weighted data does not support bootstrapping); (b) using two count, rather than binary, self-reported ethnic discrimination variables and a binary lifetime discrimination variable; (c) using three additional covariates (lifetime DSM-5 diagnoses of anxiety, depressive, and antisocial personality disorder), implicated in comorbidity theory [45].

Results

Sample characteristics (unweighted) are presented in Table 1. Men and women were represented nearly equally in the study’s sample of non-US-born Hispanic adults reporting any past-year alcohol consumption. Self-reported perceived ethnic discrimination was comparable between men and women, as was BAS-English (English use/proficiency). The mean BAS-English score was 2.6 (on a scale of 1–4) for both men and women, with a standard deviation of approximately one. While past-year DSM-5 alcohol use disorder was more prevalent in men (17.0%) than women (7.8%), family history of alcohol-related problems was more frequently reported by the women (46.6%) than the men (38.4%) in the sample.

Table 1 Sample characteristics of the study participants (non-US-born Hispanic adults reporting past-year alcohol consumption; n = 2222), stratified by sex
Full size table

Table 2 presents the results of multivariable binomial logistic regression of self-reported perceived ethnic discrimination on BAS-English, for men and women. In both men and women, each one-unit increase in BAS-English score was associated with significantly higher adjusted odds of self-reported perceived ethnic discrimination, providing support for hypothesis 1. For the outcome of past-year DSM-5 alcohol use disorder (presented in Table 3), however, results differed for men and women.

Table 2 Results of multivariable binomial logistic regression predicting perceived ethnic discrimination from BAS-English score (H1), stratified by sex
Full size table
Table 3 Results of multivariable binomial logistic regression predicting past-year DSM-5 alcohol use disorder from selected characteristics, stratified by sex
Full size table

For men, yet not women, perceived ethnic discrimination was associated with significantly higher adjusted odds of past-year DSM-5 alcohol use disorder (Model 1), providing only partial support for hypothesis 2. Each one-unit increase in BAS-English score was associated with significantly higher adjusted odds of past-year DSM-5 alcohol use disorder (Model 2), for both men and women, supporting hypothesis 3. Hypothesis 4 proposed that higher BAS-English scores would also be associated with significantly higher adjusted odds of meeting criteria for past-year DSM-5 alcohol use disorder, in a model that included self-reported perceived ethnic discrimination. Results supported this hypothesis (Model 3). In Model 3, with the inclusion of BAS-English in the regression model, men and women differed in that perceived ethnic discrimination was associated with significantly higher adjusted odds of alcohol use disorder for men, yet not women.

Table 4 presents the results of the mediation analyses, including the logit coefficients for total, direct, and indirect effects, as well as the percentage of the total effect due to mediation. The total effect represents the overall “effect” of BAS-English on DSM-5 alcohol use disorder, and the indirect effect represents the mediated effect, or the “effect” of BAS-English on DSM-5 alcohol use disorder via the mediator (perceived ethnic discrimination). Adjusted for age, educational attainment, family income, marital status, and family history of alcohol-related problems, results for men indicated that perceived ethnic discrimination acted as a mediator (logit coefficient 0.05; bootstrapped 95% CI, 0.01–0.10; p < 0.05) in the relationship between BAS-English and past-year DSM-5 alcohol use disorder, with mediation representing 17.1% of the total effect. Results for women are also presented in Table 4 for consistency and as part of hypothesis testing; however, the lack of a significant association between the mediator and outcome for the women in the study precluded the possibility of finding a significant indirect effect via ethnic discrimination among women.

Table 4 Decomposition of the total effect (into direct and indirect effects) of BAS-English on past-year DSM-5 alcohol use disorder via ethnic discrimination (H5), stratified by sex
Full size table

Results of sensitivity analyses showed that the point estimates computed with the KHB method and Stata’s GSEM command were identical (e.g., 0.05, men’s logit coefficient of indirect effect), and the 95% confidence intervals obtained with these two estimation methods were relatively similar (e.g., bootstrapped 95% CI, 0.011–0.098 for men’s indirect effect using the KHB method; 95% CI, 0.015–0.094 for men’s indirect effect using Stata’s GSEM). The analyses using the count (rather than binary) variable capturing the number of settings in which discrimination occurred provided relatively similar results (e.g., men’s logit coefficient of the indirect effect via ethnic discrimination was 0.06 [bootstrapped 95% CI, 0.01–0.11, p < 0.05], with 19.6% of the total effect due to mediation). For the discrimination variable measured on a scale of 0–24, results were also relatively similarly (e.g., men’s logit coefficient for the indirect effect via ethnic discrimination was 0.05 [bootstrapped 95% CI, 0.01–0.09, p < 0.05] with 15.9% of the total effect due to mediation). Results utilizing the lifetime (as opposed to before past year only) discrimination variable were similar (e.g., men’s logit coefficient for the indirect effect via ethnic discrimination was 0.06 [bootstrapped 95% CI 0.02–0.11, p < 0.05] with 20.2% of the total effect due to mediation). Lastly, results did not substantially change after adding three covariates (lifetime DSM-5 diagnoses of anxiety, depressive, and antisocial personality disorder) associated with alcohol use disorder (e.g., men’s logit coefficient of the indirect effect via ethnic discrimination was 0.04 [bootstrapped 95% CI, 0.01–0.08, p < 0.05], with 14.7% of the total effect due to mediation).

Discussion

Results of this study supported a positive and significant relationship between English use/proficiency (BAS-English) and DSM-5 alcohol use disorder in non-US-born Hispanic men and women. These results are consistent with prior studies with US Hispanics which indicate a positive relationship between language-based measures of acculturation and drinking outcomes [5, 6]. Beyond documenting the relationship between BAS-English and alcohol use disorder, however, the present study evaluated a hypothesized explanation—perceived ethnic discrimination—using a population-based probability sample.

For non-US-born Hispanics who reported past-year alcohol consumption, English use/proficiency (BAS-English) was identified as a predictor of self-reported perceived ethnic discrimination. Several explanations are possible. First, individuals with greater English proficiency may be better able to understand discriminatory speech and label it as such. Second, such individuals may spend more time with English-speakers and be more aware of the status ascribed to their ethnic group in a racialized society. Third, individuals with higher English utilization may also be more likely to have lived in the US for a greater part of their lives, potentially with a history of exposure to discrimination beginning at an earlier age, in multiple settings. Some individuals with low English utilization, in contrast, might be recent arrivals to the US or reside in ethnic enclaves with potentially less perception of both their minority status and ethnic-based discrimination [15].

Although English use/proficiency was associated with ethnic discrimination for both men and women in the sample, ethnic discrimination emerged as a predictor of alcohol use disorder for men only. Accordingly, ethnic discrimination acted as a partial mediator in the relationship between English use/proficiency and alcohol use disorder for men, but not women. The sex differences in the relationship between discrimination and alcohol use disorder are consistent with past research [20, 21]. Stress (which may stem from ethnic discrimination) is believed to affect men and women differently, such that internalizing problems may be more common among women while externalizing problems (e.g., related to alcohol and drug consumption) may be more common among men [46, 47]. It may also be important to note that the present study focused on Hispanic immigrants who reported alcohol use within the past year, and Hispanic immigrant women are more likely than men to abstain from alcohol use [48].

Several possibilities explain how “othering,” racial discrimination, and other manifestations of racism—which appear to often accompany the process of acculturation—may influence problematic drinking in non-US-born Hispanic men. Racist acts can take a toll on the body’s biochemical processes, increasing anxiety and causing rumination and defensive or aggressive responses, altering physiological arousal involved in stress-related diseases [49]. Cognition may also play a role in this relationship. Negative coping [50, 51] may explain heavy drinking in response to experiencing ethnic discrimination. Based on social learning theory [52], repeatedly experiencing the immediate and relieving effects of drinking may become a powerful mechanism to develop a maladaptive pattern of drinking. At the same time, experiences of discrimination may impair self-regulatory mechanisms, weakening individuals’ ability to resist or control alcohol use [53], diminishing individuals’ sense of self-efficacy and belief in their ability to abstain from drinking, resist pressures to drink, or not engage in heavy drinking [54].

Results of mediation analyses are consistent with findings of a prior study with adolescents of primarily Mexican heritage that reported an indirect effect from an acculturation-related measure to severity of alcohol use, via ethnic discrimination [55]. Certainly, the mediation effect in the present study accounted for a relatively small proportion (17.1%) of the relationship between English use/proficiency and alcohol use disorder in Hispanic immigrant men. Although the present study conceptualized a single-mediator model, it is likely that the association between English use/proficiency and alcohol use disorder, as well as between ethnic discrimination and alcohol use disorder, is mediated by a collection of factors. Potential mediators such as social support, social networks, cognitive factors, and context of reception were outside of the scope of the present study’s conceptual model. Previously-identified mediators in the relationship between racial discrimination and health include stress, self-esteem, psychological distress, depression [56], and sleep [57], and mediators specific to the relationship between racial or ethnic discrimination and drinking include anger, post-traumatic stress symptoms, and depressive symptoms [19].

Consequential to the analysis of secondary data, the present study utilized “mediation for explanation,” which “is probably more susceptible to chance findings” than “mediation for design” (wherein mediators are specified at the outset of the study design) [58]. The statistical mediation utilized in this study was unable to randomize the predictor or mediator. Not all assumptions of the single-mediator model, such as the temporal order of variables [58], could be verified; that is, although perceived ethnic discrimination was measured prior to the past year and the outcome was for the past year, temporal precedence could not be ensured in these cross-sectional data.

Analyses were unable to account for variation due to country of origin, due to the low representation of several Hispanic heritage groups, or differences in the contexts of reception experienced by immigrants settling in different regions in the US. It is possible that for some Hispanic immigrants (potentially depending on phenotype), greater English proficiency could lead to greater acceptance by the mainstream society and potentially fewer experiences of othering or discrimination. NESARC-III’s epidemiological data did not include questions about cognition (e.g., drinking motives), limiting the predictive power of the statistical models; other unobserved variables may also account for the observed associations. All measures utilized in the study were based on self-report and are therefore subject to biases such as recall bias and social desirability bias. The ethnic discrimination measure was not able to capture discrimination at a structural or societal level, nor discrimination that was not reported or consciously perceived.

Exposure to, and life in, a racialized society can influence minority health in a multitude of ways, many of which may not be identified as ethnic discrimination. Many factors lie at a systemic level, ingrained in society. Racial or ethnic discrimination represents only one type of discrimination; discrimination may be based on various characteristics and compounded by the intersection of multiple identities. Future studies should explore measures of discrimination that transcend the individual domain [19], or measures with the potential to identify experiences that may uniquely affect some Hispanic populations, such as language-based discrimination [59, 60]. Furthermore, research should consider the interplay of various factors, including skin tone and immigration status. Finally, in contemporary American society, research on discrimination should examine the mental and behavioral health impacts of policy and political rhetoric that vilifies, stigmatizes, or restricts immigrants or Hispanics.

Because acculturation is often considered an inevitable process for immigrants and their descendants, attributing negative mental health outcomes to acculturation diminishes the sense of possibility for prevention and intervention, fostering resignation to the status quo. In contrast, modifiable explanatory factors (mediators) offer targets for intervention. One of these factors appears to be perceived ethnic discrimination, particularly in the context of alcohol use disorder among Hispanic immigrant men. Therefore, culturally-competent behavioral health services (including screening, assessment, and treatment) need not be limited to linguistic adaptation or cultural awareness; the concept of cultural competence may be expanded to include sensitivity to, and validation of, the experiences of othering, discrimination, and exclusion that affect minority mental health.