Data source and strategy
This study uses the data from the National Latino Asian American Study (NLAAS) (Alegria et al., 2004), the first national population-based mental health study of Latino and Asian Americans. NLAAS was designed in coordination with and as a part of the Collaborative Psychiatric Epidemiology Studies (CPES), which includes NLAAS, the National Survey of American Life, and the National Comorbidity Survey Replication. This survey compared the association of immigration factors with mental health services across three major racial and ethnic categories: African Americans, Asian Americans, and Latino Americans. NLAAS collected extensive information on health and mental health status, behaviors, and health insurance coverage for all age groups. The questionnaire was available in six languages: English, Spanish, Cantonese, Mandarin, Tagalog, and Vietnamese. Trained bilingual interviewers interviewed all NLAAS participants. The study met the approval of three Internal Review Board Committees: Cambridge Health Alliance, the University of Washington, and the University of Michigan (Alegria et al., 2004). The total sample size of the NLAAS data was 4,649, including 2,554 Latinos and 2,095 Asian Americans. The survey included individuals 18 years or older of Latino or Spanish origin who speak English and Spanish. The survey also provided health-related information on the participants (Lewis-Fernández et al., 2009).
Our study sample comprised 264 Cuban-born women and a comparison group of 639 Latina immigrants. The Latina immigrants combined comparison groups of Latinas born in Puerto Rico, Cuba, and Mexico. We incorporated Cuban-born Latinas in the Latina immigrants combined group to mitigate a bias resulting from combining Latina subgroups into one category.
Measures
This study's two dependent variables are acculturative stress and psychological distress. In the past, discrimination has been treated as a part of the acculturative stress experience, with many acculturative stress measures incorporating discrimination (i.e., Hispanic Stress Inventory (HSI)) (Finch et al., 2000). While acculturative stress might correlate with discrimination, discrimination is a particular stressor that separately impacts one's psychological distress (Goodkind et al., 2008). Kulis, Marsiglia, and Nieri (2009) separated discrimination and acculturative stress into two indicators and found that discrimination emerged as a more vital indicator of poor mental health (Kulis et al., 2009). Concurring with Kulis et al. (2009), we posit that while acculturative stress is inherent to one's effort to adapt culturally to a new country, discrimination is related to one's ethnic background and socioeconomic status (Kulis et al., 2009). As such, we exclude discrimination from our acculturative stress measure.
Acculturative stress is measured by adults' self-rated level of agreement on six statements: (1) I felt guilty about leaving family or friends in the country of origin, (2) I have limited contact with family and friends, (3) I find interaction hard due to difficulty with the English language, (4) I was questioned about my legal status, (5) I think I might be deported if I go to a social or government agency, and (6) I avoid health service due to Immigration and Naturalization Services (INS). Possible scores ranged from 0–9, with higher scores indicating a higher level of acculturative stress. Items have been obtained from the Mexican-American Prevalence and Service Survey, which measures the experiences of psychosocial stress for Latinos. The scale has been tested repeatedly for appropriateness in general Latinos (Cervantes et al., 1991) and has high internal consistency (Cronbach is a = 0.88).
The second dependent variable, psychological distress, is measured by Kessler's Psychological Distress Scale consisting of 10 items surveying symptoms of depression and anxiety (K10). A validation study completed in Boston on a small convenience sample found evidence that the scales perform (Kessler et al., 2003). The K10 is also included in the National Comorbidity Survey Replication (NCS-R) as well as in all the national surveys in the World Health Organization (WHO) World Mental Health (WMH) Initiative (Furukawa et al., 2003). Items include responses to the following statements: (1) Past month I felt depressed, (2) Past month I felt hopeless, (3) Past month I felt restless/fidgety, and (4) Past month I felt tired for no good reason (5). The scale's internal consistency is high (Cronbach is a = 0.85).
External context. The first segment of the external context is represented by the U.S. climate, i.e., perceived daily and racial discrimination, visiting family abroad difficulties, contentment with the decision to move to the U.S., U.S. region of residence, and years in the U.S. Perceived discrimination is a two-dimensional construct consisting of perceived racial and daily discrimination. Perceived racial discrimination of adults over 18 is measured by questions such as (1) How often do people dislike you because of your race/ethnicity? (2) Have you seen friends mistreated because of their race/ethnicity? Moreover, (3) How often do people treat you unfairly because of your race/ethnicity? Perceived daily discrimination is measured by statements such as: (1) You are treated with less courtesy than other people, (2) You are treated with less respect than other people, (3) You receive poorer service than other people at restaurants or stores, (4) People act as if they think you are not smart, (5) People act as if they are afraid of you, (6) People act as if they think you are dishonest, (7) People act as if you are not as good as they are, (8) You are called names or insulted, and (9) You are threatened or harassed. The items were taken from the Detroit Area Study (DAS) (Williams, Yu et al., 1997), and the scales have been used in mental health research (Finch et al., 2002; Mays & Cochran, 2001). We reverse-coded the response categories such that higher scores would reflect a greater frequency of perceived discrimination. The internal consistencies are high (Cronbach's a = 0.88 and a = 0.92, respectively).
In assessing visiting family abroad difficulties, respondents were asked to rate their answers from relatives abroad (1) to Very difficult (5). Years in the U.S. is an ordinal variable, with responses ranging from Less than five years (1) to more than 20 years (4). The U.S. region of residence, which initially consisted of four categories, (1) West, (2) East, (3) North, and (4) South, was transformed into four dichotomous variables corresponding to the four regions in the U.S. Contentment with the decision to move to the U.S. is represented by the dichotomous question, Would you decide to move to the U.S.? The answers include (1) Yes and (0) No.
Development context, the second segment of the FSM's external context, consists of the following variables: Age, continuous variables with respondents ranging from 18–97 years old; Age at immigration, a categorical variable where participants were asked, How old were you when you first came to this country? Moreover, whose answer categories included (1) less than 12 years, (2) 13 to 17 years, (3) 18 to 34 years, and (4) older than 34, English skills represented with questions such as: (1) How well do you speak in English? (2) How well do you read in English? and (3) How well do you write in English? The answers range from 1 – Poor to 4, Excellent. The internal consistency is very high (Cronbach's a = 0.93). Other variables of development context include education, an ordinal variable coded as (1) 0–11, (2) 12 years of education, (3) 13–16 years, and (4) more than 16 years of education; and marital status, transformed into three dichotomous variables representing the following categories, (1) married/cohabiting, (2) divorced/separated/widowed, or (3) never married.
Financial context. The third segment of the FSM's external context consists of financial constraints, an ordinal variable represented by the question, Rate your difficulty in meeting the monthly financial obligations and answers –ranging from (1) Not at all difficult to (4) very difficult; employment, a dichotomous variable coded as (0) not employed and (1) fully employed; and the continuous variable household income, referred to as income.
Gender-based roles and family-culture conflict represent the cultural context. The gender-based role is measured by one item: Which of the partners has the final say in household decisions? Research indicates that this construct can be used to indicate whether Latina women live in a household where a traditional gender-based role is in place (Baker, 2004). Family-culture conflict is represented by five items drawn from a subscale of the Hispanic Stress Inventory (HIS), which measure the cultural and intergenerational conflict between Latina immigrant women and their families (Cervantes et al., 1991). Participants were presented with items such as: (1) Because of the lack of family unity, you have felt lonely and isolated, (2) Your personal goals have been in conflict with your family, (3) Being close to family interfered with goals, (4) You argue with family over different customs. Each question about family-culture conflict has three potential response categories (1) Hardly Ever or Never, (2) Sometimes (3) Often. Higher scores were indicative of higher levels of conflict. The scale's internal consistency for this population is high (Cronbach's a = 0.81).
The fifth FSM context, the global context, is represented by the impact of 9/11. To assess the impact of 9/11, respondents were asked, As a result of the attacks, how much has your life been affected in the areas such as (1) losing one's job, (2) reduction in family income, (3) becoming more patriotic, (3) feeling safe, (4) unfair treatments due to race/appearance, (5) optimism about future, and (6) not coping well, in general. The answers ranged from a lot (1) to none (4). All items were reverse coded so that higher scores corresponded to higher levels of 9/11 impact. The internal consistency among these items is high (Cronbach's a = .74).
The cultural value of familismo and religiosity represent FSM's Internal Context. Familismo incorporates opinions of shared family, cultural values, and a general orientation toward family. The current study incorporates opinions of shared familial, cultural values, a general orientation toward family, family cohesion, and the willingness of family members to spend time with each other. The scale has been used frequently with people of Latin origin (Felix-Ortiz et al., 1994). It is measured by asking adults to rate their level of agreement with statements such as (1) Family members respect each other, (2) The family shares values, (3) Things work well as a family, (4) The family trusts and confides in each other, and (5) Family is loyal to family. Possible scores ranged from 10–40. All items were reverse-coded so that higher scores correspond to higher family cohesion levels. The scale's internal consistency is high (Cronbach's a = 0.80). Religiosity was studied using a question about church attendance, which probed the respondent about how frequently they attend religious services. Response categories ranged from (1) Often to (4) Never. Responses were reverse-coded such that higher values reflected higher levels of religiosity.
Statistical methods
We used mediation path analysis in MPLUS (Muthen & Muthen, 2005) to estimate relationships between contextual factors, acculturative stress, and psychological distress among the subjects in this study. All ordinal variables with at least four levels are treated as continuous (Muthen & Muthen, 2005). Categorical variables were transformed into binary variables to facilitate the analysis, while the dichotomous variables were left in their original dichotomous form. Only those covariates based on theory (Berry, 1997) or literature related to acculturative stress were entered into the model. All regression assumptions were assessed, such as linearity, homoscedasticity, and multicollinearity. Before entering various variables into models, two bivariate statistical tools were first used: correlations and analysis of variance. An analysis of the correlation involved an examination of the Pearson correlation to check for statistical significance of covariates and the two dependent variables. Only those contextual factors determined to have a significant relationship with acculturative stress and psychological distress were kept in the analysis. For Latina immigrants combined, these variables included decision-making power, perceived daily and racial discrimination, years in the U.S., 9/11 impact, visiting family abroad difficulties, financial constraints and level of income, age at immigration, age, contentment with the decision to move to the U.S., and education level, family-culture conflict, English skills, religiosity, familismo, and U.S. region of residence. Non-significant contextual factors included marital status. For Cuban-born women, statistically significant variables included contentment with the decision to move to the U.S., age at immigration, visiting family abroad difficulties, financial constraints, years in the U.S. English skills, perceived racial discrimination, family-culture conflict, familismo, employment, and age.
The mean substitution was performed for measures where participants had at least 80% complete data. For the measures with remaining missing values after mean substitution was completed (acculturative stress, familismo, and family conflict), and for English language proficiency, the Markov Chain Monte Carlo (MCMC) multiple imputation procedure was performed to correct the missing values. Five imputations, Stress, Family, and Mental Health, were generated for measures with missing data. The values from the imputations were then aggregated to create a final estimate of the missing values upon which the final analyses were based.
Specification search (Schumacker & Lomax, 2004) permitted the search for a better-fitting model, while a Modification Index (MI) informed our inclusion of specific parameters. MPLUS automatically calculated the indirect effects of covariates estimated in this study. Comparative Fit Index (CF), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Residual (SRMR) were used to assess the fit of the models. The maximum likelihood method (MLR), an estimator robust to non-normality, was employed. Standard errors estimated through MLR are very close to those produced through "bootstrap" methods; nonetheless, all the estimated baseline models were cross-validated using the holdover cross-validation technique.
Interactions between the various contexts in predicting acculturative stress were estimated using Stata's fitting software. Before testing for interactions, continuous variables and the ordinal variables were treated as continuous and mean-centered, and categorical variables were transformed into dichotomous variables to avoid collinearity. No interactions were significant among Cuban-born women.