|Home | About | Journals | Submit | Contact Us | Français|
We examined potential pathways by which time in the U.S. may relate to differences in the predicted probability of last year psychiatric disorder among Latino immigrants as compared to U.S.-born Latinos.
We estimated predicted probabilities of psychiatric disorder for U.S.-born and immigrant groups with varying time in the U.S. adjusting for different combinations of covariates. We examined six “pathways” by which time in the U.S. could be associated with psychiatric disorder.
Increased time in the U.S. is associated with higher risk of psychiatric disorders among Latino immigrants. After adjustment for covariates, differences in psychiatric disorder rates disappear between U.S.-born and immigrant Latinos. Discrimination and family cultural conflict appear to have a significant role in the association between time in the U.S. and the likelihood of developing psychiatric disorders.
Increased perceived discrimination and family cultural conflict are pathways by which acculturation might relate to deterioration of mental health for immigrants. Future studies assessing how these implicit pathways evolve with increased contact with U.S. culture may help to identify strategies for ensuring maintenance of mental health for Latino immigrants.
Two common patterns have been typically described in the literature: Latino immigrants have better mental health than their U.S.-born counterparts and non-Latino whites despite having disadvantaged socioeconomic status1–3 (the “immigrant paradox”), and the mental health of immigrants declines over time in the host country3 (the “acculturation hypothesis”). Our findings from the National Latino and Asian-American Study (NLAAS) on the prevalence of psychiatric disorders among Latinos in the U.S. indicate that foreign nativity is protective for some Latino groups (e.g., Mexicans), but not others (e.g., Puerto Ricans),4 implying that other factors besides nativity play a part in their risk of psychiatric disorders. Also, there is evidence that risk of psychopathology is due to differences in immigrants’ length of residence in the U.S. and age of arrival. For example, Mexican immigrants in the U.S. for 13+ years had higher rates of any mental health disorder, any mood disorder, alcohol abuse and drug abuse than Mexican immigrants in the U.S. for <13 years.3 Findings from the NLAAS indicate that the longer Latino immigrants remain in their country of origin, the less cumulative risk of onset of psychiatric disorders they experience, resulting in lower lifetime rates of disorders.5
There is a lack of consensus about which aspects of U.S. exposure are relevant for mental health.5 A number of hypotheses have linked years in the U.S. and mental illness among Latinos.3 U.S.-born Latinos, in comparison to Latino immigrants, may have a weaker affiliation of traditional Latino values that buffer against mental illness. When individuals come into contact with U.S. culture, there may be negative outcomes such as increased intergenerational conflict,6 augmenting their risk for psychopathology. Although family factors have been hypothesized to be a protective factor for immigrant Latinos,7,8 few empirical studies have actually tested this hypothesis with regards to psychiatric disorders. Another hypothesis is that U.S.-born Latinos may have higher expectations for their quality of life than immigrant Latinos because of their citizenship status and acquisition of skills similar to non-Latino whites.9 However, over time, these expectations may be unfulfilled because of discrimination resulting in social stress and declining levels of mental health.1 Also, as time passes, immigrants may have perceptions of low social status10 that may be associated with higher risk of psychopathology.
Other immigration-related factors could also affect adaptation experiences when integrating into the U.S. Specifically, those arriving in the U.S. at early formative ages (0–10 years of age) may have weaker identification with Latino cultural values11 and confront significant pressure to acquire English as their dominant language.12 English language dominance represents a strong cultural anchor for socially-constructed meaning13 that may serve immigrants to join into certain peer networks and not others. Unsafe neighborhoods where Latinos typically live in comparison to non-Latino whites14 may increase the likelihood of psychiatric disorders15 by increasing ambient hazards. In addition, exposure to racial/ethnic-based discrimination16,17 has been associated with negative mental health outcomes. The NLAAS study provides a unique opportunity to explore these pathways since these domains have been assessed for both Latino immigrants and U.S.-born Latino respondents.
In this paper, we assess the association of time in the U.S. with last year risk for psychiatric disorder, with and without adjustment for potentially influential covariates. Next, we test different pathways explaining the link between time in U.S. and psychiatric disorders. Finally, we compare these analyses with previous research and provide recommendations for future studies.
The NLAAS is a nationally representative survey of English- and Spanish-speaking household residents ages 18 and older in the non-institutionalized population of the coterminous United States.18 2554 Latinos comprised the final sample with a response rate of 75.5%. Weighting provides coverage of the full national Latino population. The weighted sample is similar to the 2000 Census in sex, age, education, marital status and geographical distribution (data not shown) but includes more Latino immigrants and lower-income respondents. This is consistent with reports of the undercounting of immigrants in the Census.19
The development of the NLAAS instrument involved the adaptation of existing measures, and translation of most measures into Spanish. Key variables and scales with their psychometric properties are described in Alegría, Vila et al. (2004).20 Most measures were selected based on face validity, internal reliability, and use in other studies of Latino mental health. Data were collected by the Institute for Social Research at the University of Michigan between May 2002 and November 2003. Professional lay interviewers administered the NLAAS battery, averaging 2.6 hours.
The dependent variables of interest are self-reported diagnostic measures for last twelvemonth psychiatric disorders (any psychiatric disorder, any depressive disorder, and any anxiety disorder) using the diagnostic interview of the World Mental Health Survey Initiative version of the World Health Organization Composite International Diagnostic Interview (WMH-CIDI),21 a structured diagnostic instrument based on criteria of the DSM-IV. The independent variable of interest is time (years) in the U.S. since immigration (0–10, 11–20, 21+ years since arrival, or U.S.-born as reference category). Categorical variables were chosen instead of a continuous measure of time in the U.S. for model simplicity and ease of interpretation because nonlinear effects of time in U.S. on risk of psychiatric illness were observed. Each of the immigrant time in U.S. categories represents approximately one-third of the weighted Latino immigrant population, allowing for robust estimation for each of the groups. Sociodemographic variables are gender and age (18–24; 25–34; 35–49; 50–64;≥65), using comparable categories to those previously used in the literature.22 Other covariates include marital status, citizenship, employment status, respondent’s years of education (no high school (HS) degree [<12]; HS graduate ; some college [13–15]; college degree [≥16]), parent’s education (no HS [0–7]; some HS or HS degree [8–12] and any college [13+]), English language proficiency (excellent/very good/good or fair/poor) based on a scale (α=0.98) that assesses respondents’ ability to speak, read, and write in English (23), and health insurance (any or none) based on a series of questions that assess different types of insurance coverage.
We assessed six potential “pathways” that we hypothesized could explain how time in the U.S. relates to mental health outcomes. Exposure to discrimination is a nine-item measure assessing frequency of routine experiences of unfair treatment (e.g., being treated with less respect than others, having people act afraid of them; α=0.82). The family cultural conflict scale consists of five items measuring respondents’ frequency of intergenerational conflict with families (e.g., family interference with personal goals, arguments with family due to different belief systems; α=0.91). The ethnic identity scale (α=0.75) determines respondents’ identification, closeness of ideas, and shared time with members of their own ethnic group. Dissatisfaction with economic opportunities is based on a single question that asks the respondent how satisfied he/she feels about his/her economic opportunities in the U.S. Perceived social status is assessed by asking respondents to identify their social status relative to others in their U.S. community, based on money, education, and job respect.10 The neighborhood safety scale includes three items that measure respondents’ perceived level of neighborhood safety and lack of violence (α=0.72).
We first examined the unadjusted past year prevalence rates of any psychiatric disorder, any depressive disorder, and anxiety disorder, and selected risk factors among U.S.-born Latinos and three groups of Latino immigrants partitioned by time in the U.S. We analyzed psychiatric disorder in the year prior to interview (as opposed to lifetime prevalence) to ensure that disorder was experienced in the U.S. and not just in an individual’s home country. To assess correlates of the relationship between time in the U.S. and mental health, we used logistic regression to model the probability of last year psychiatric disorder as a function of time in the U.S, controlling for the above described list of covariates. Additionally, the six potential “pathways” described above were interacted with the time in U.S. categories in order to account for differential group effects. Because previous studies have found that the U.S.-born have a higher prevalence of psychiatric disorder than immigrant populations,4,5 we used the U.S.-born Latino population as the reference group. The fit of these logistic models was verified using Pregibon’s Link Test23 and the Hosmer-Lemeshow goodness of fit test modified to account for intra-cluster correlated data and design effects.24
Next, we estimated the residual direct effect (RDE) of each of the three immigrant time-in-U.S. groups on having any last year psychiatric disorder.25 This technique is also named predictive margins,26 or the recycled predictions approach,27,28 and has been used in previous health services studies.29–31 The RDE method is a generalization of adjusted treatment means to nonlinear models, allowing us to compare the prevalence of psychiatric disorders of immigrants in different time-in-US groups after setting the distribution of all of their other observable attributes equal. We balanced groups on observable variables by creating “counterfactual” comparison populations, groups that are hypothetical and not actually observed, but that can be created by adjusting selected individuals’ variable responses. For example, to assess the RDE of being in the 0–10 year immigrant group, we 1) modeled last year psychiatric disorder as a function of nativity and time in the U.S. and other covariates; 2) created a “counterfactual population” by selecting the U.S.-born individuals and changing their group membership from U.S.-born to the 0–10 year immigrant group while keeping other covariates at their original U.S.-born values; 3) generated predicted risk for the counterfactual population. This prediction represents the predicted probability of having a psychiatric disorder in the last year for a counterfactual immigrant population with 0–10 years in the U.S. but all other characteristics of the U.S.-born. We use the term prediction statistically. It is not our intention to predict the future mental health of these groups, but rather to produce model-based estimates of psychiatric illness prevalence for counterfactual groups. The method allows for direct comparisons of predicted prevalence rates between immigrant and U.S.-born groups, fully adjusting for all covariates.
Finally, we identified underlying factors, or pathways, that potentially explain differences in mental health among immigrant and U.S.-born groups. We define a pathway to be significant if model-based predictions of risk of disorders that adjust for all covariates are significantly different from predictions that adjust for all covariates except the pathway variable. This analysis is similar to Oaxaca-Blinder (O-B) decomposition analyses32,33 used in previous studies that assess the percentage contribution of a variable or set of variables to predicted racial/ethnic disparities in access.34–36 Our analysis method extends the O-B decomposition method to nonlinear models. The method allows us to test, one by one, the contribution of each pathway variable to time-in-U.S. group differences in risk of psychiatric disorder, while adjusting for the other pathway variables and covariates. The analysis was implemented identically to the RDE method described above except we allowed group differences in the pathway variable to enter into the prediction in steps 2 and 3. As before, we fit a model for any psychiatric disorder and created a counterfactual group with immigrant group membership and U.S.-born characteristics. Next, we ranked individuals in each group based on their pathway variable scores, replacing scores for U.S.-born individuals with scores of equivalently ranked individuals from the appropriate immigrant group. We then generated predicted probabilities for these counterfactual populations (groups of immigrants with their own pathway variable scores and U.S.-born values for all other variables). These model-based predictions were compared with the full adjustment (RDE) predictions to estimate the contribution of each pathway variable to the group difference. This method is related to methods previously used in the racial disparities literature that generate predictions by adjusting for some, but not all model covariates in non-linear models.25,37,38 A key assumption of this method is that, after the pathway variable is changed through rank and replace, the relationship between the covariates and the outcome remains the same as in the original study sample. The analysis identifies changes in differences in rates of psychiatric disorder between the U.S.-born group and different time-in-U.S. immigrant groups with and without adjustment for pathway variables. This is not meant to measure temporal change, but rather changes in predicted immigrant-U.S.-born group differences that suggest potential pathways connecting time in the U.S. and risk of psychiatric disorder.
All analyses used sampling weights to provide estimates that are representative of the U.S., non-institutionalized Latino adult population. Standard errors for Tables 1 and and22 were estimated using Stata 9 software27 which accounts for the survey sampling design, and significance tests were performed using design-adjusted Wald tests. For group and pathway comparisons, we derived standard errors from a bootstrap procedure39 and considered predicted differences significant if their 95% bootstrap intervals did not include zero.
In unadjusted analyses, Latino immigrants living in the U.S. for 0–10 years had a significantly lower rate of any last year psychiatric disorder and any depressive disorder than U.S.-born Latinos (Table 1). Latino immigrants living in the U.S. for 11–20 years and 21+ years showed no significant differences from the U.S.-born. Unadjusted rates of any psychiatric disorder increased in step with the number of years spent in the U.S. Compared to U.S.-born Latinos, Latino immigrant groups experienced less discrimination, and were less satisfied with economic opportunities. Latino immigrants living in the U.S. for 11–20 and 21+ years reported less family cultural conflict and immigrants living in the U.S. for 0–10 and 11–20 years had stronger ethnic identity than U.S.-born Latinos. Latino immigrants living in the U.S. for 0–10 years rated themselves as having lower social standing than U.S.-born Latinos. Latino immigrants living in the U.S. 0–10 and 11–20 years were younger than U.S.-born Latinos whereas those living in the U.S. for 21+ years were older. Latino immigrants had lower education, had parents with lower education, were less English proficient, and were less likely to be insured compared to U.S.-born Latinos.
No differences in rates of psychiatric disorders were found between immigrant groups after adjustment for all model covariates, with the exception that Latinos living in the U.S. for 21+ years were more likely to have a last year anxiety disorder than the U.S.-born (Table 2). Positive correlates of last year psychiatric disorder included higher perceived discrimination, greater family cultural conflict, and English language proficiency. Individuals with higher social standing in the U.S., higher level of neighborhood safety, being married, current employment, and having parents with 8–12 years of education as compared to less than 8 years of education were less likely to have last year psychiatric disorder. In general, these correlates were similar for last year depressive and anxiety disorder. In addition, females were more likely to have last year depressive and anxiety disorders than males.
When adjusting for gender and age (Table 3, Section 1 labeled “Gender and Age Adjustment”), we found that greater time in U.S. for immigrants increased risk, or the predicted probability, of last year psychiatric disorder. The counterfactual immigrant groups with all U.S.-born characteristics except for time in the U.S. showed increasing risk of disorder with increasing time in the U.S. (see Table 3, Section 2 labeled “Full Adjustment (RDE)”). Compared to U.S.-born Latinos, immigrants living in the U.S. for 0–10 years had significantly lower predicted risk of any last year psychiatric disorder and any last year depressive disorder.
Different levels of discrimination between immigrant and U.S.-born Latinos led to a significant change in rates of any psychiatric disorder, any depressive disorder, and any anxiety disorder among Latino immigrants living in the U.S. for 11–20 years and 21+ years. Reducing perceived discrimination from the level of the U.S.-born Latino to the level of Latino immigrants living in the U.S. for 11–20 years and 21+ years led, respectively, to a 4.5% and 2.6% decrease in the predicted probability of any last year psychiatric disorder, a 1.7% and 1.4% decrease in the predicted probability of any last year depressive disorder, and a 3.4% and 2.4% decrease in the predicted probability of any last year anxiety disorder (see comparison of Table 3, Section 3 with Table 3, Section 2). Reducing family cultural conflict from the level of the U.S.-born Latino to the level of the Latino immigrants living in the U.S. for 11–20 years and 21+ years resulted in a greater than one percent decrease in the predicted probability of any last year psychiatric disorder (see comparison of Table 3, Section 4 with Table 3, Section 2). Adjusting the values of the other four pathway variables did not lead to changes in predicted risk of psychiatric disorders (Table 3, Sections 5–8). Figure 1 summarizes results for the pathway variable analysis by showing the change in predicted probability of last year psychiatric disorder due to the adjustment of each of the six pathway variables. Discrimination and family conflict emerge as factors that may explain differences in past year mental health among U.S.-born and immigrant groups with varying times in the U.S.
There are several limitations of the current study. Identifying causal pathways between exposure to the U.S. and mental health is not possible with the available cross-sectional data, but only suggestive of potential relations. In addition, identifying causality would require that there was no emigration out of the foreign-born cohorts. Immigrants may desire to return to their home country after falling ill in order to be in the company of their families. This “salmon bias” hypothesis40 may help to explain lower rates of mortality and morbidity among immigrants in comparison to the U.S.-born. Future study should take advantage of longitudinal datasets and surveys based in Latin America to test for this type of bias. Given the inability to establish causality, the results can only be considered as suggestive of pathways by which time in the U.S. augments the risk of mental disorders.
Another limitation is that we assess the contribution of pathway variables to foreign-born and U.S.-born group differences one variable at a time, assessing the independent effect of each pathway variable. This analysis does not allow for the possibility that time in U.S. and nativity may influence mental health through multi-stage pathways. For example, it is plausible that longer time in the US increases ethnic identity, which increases perceived discrimination (further reinforcing ethnic identity), leading to higher rates of mental illness. Analyzing multi-stage pathways would require more complex statistical models and stronger model assumptions. Our analysis is further limited because the NLAAS does not provide adequate sample size to assess differences by both time in the US and ethnic subgroups (e.g., Mexican, Puerto Rican, Cuban, and other Latino subgroups) while adjusting for other covariates. Our prior research41 found significant differences between these sub-ethnic groups, suggesting that investigating the interaction of time in the U.S. and sub-ethnicity would be worthwhile given a larger dataset.
Notwithstanding, the results bolster previous studies supporting the “acculturation hypothesis” whereby greater exposure to the U.S. increases rates of any mental illness among Latino immigrants despite concomitant increases in SES and other generally positive predictors of mental health.3–5,42 Protective mechanisms from the country of origin (e.g. family harmony) may erode over time because of the additional demands of having to negotiate different cultural contexts in the US.43 In some contexts, social integration and linguistic isolation may limit the strain on an immigrant. In other contexts, the lack of ethnic enclaves and the rejection of immigrants, even those with social mobility, may force immigrants to constantly confront discrimination and a negative social mirror. The costs of family conflict and repeated exposure to discrimination44 linked to the experience of upward social mobility may increase emotional distress and leave immigrants without social supports. As a result, the emotional consequences and severed family ties associated with acculturation may actually outweigh the benefits of social mobility. Time in the US could also proxy for increased acculturative stress caused by the tension between the demands of mainstream society, available resources, and the desire to remain anchored in the values and norms of Latino society (eg. familism). Having a bicultural identity that facilitates navigating differences between the traditional values, beliefs, and socialization practices of the family’s original culture as well as those of the host culture may be critical to safeguard against mental disorders. Future studies should assess how different domains (i.e., biculturalism, gender roles, aspirations of social integration, expectations of fair treatment, availability of ethnic enclaves) mediate risk for mental illness after extended contact in the US. Our study results did not support the immigrant paradox, finding no mental health differences between U.S.-born and immigrant Latinos, with the exception that Latino immigrants living in the U.S. for 21+ years, were more likely to have a last year anxiety disorder than the U.S.-born.
Our findings could be explained by the mounting evidence45 that it is not nativity per se that accounts for the immigrant paradox effect but rather differences in contextual and interpersonal circumstances. In our analysis, we found that higher levels of discrimination and family conflict among U.S.-born Latinos (compared to immigrant Latinos) play a significant role in increasing risk of psychiatric disorders. This complements previous research16 finding that more acculturated Mexican immigrants reported higher rates of discrimination compared to less acculturated immigrants. As Latinos become more assimilated in U.S. society, they may become more perceptive to differential treatment and unfairness,46 or intermingle more with non-Latinos and interpret many inter-cultural interactions as discriminatory, all of this leading to augmented risk for psychiatric disorders. Because acculturation involves an alterable cognitive shift in one’s desire to belong within mainstream culture, the negative impact of discrimination might operate differently for different immigrants. For example, immigrants may receive negative social feedback suggesting exclusion from mainstream culture, despite having a social profile (e.g. education/occupation) apparently necessary to belong. Those with little desire to belong to mainstream US culture may ignore this negative feedback and therefore be less prone to mental illness than those with a strong desire to assimilate and be part of the “American dream”.
Erosion of family bonding by greater time in the U.S. has been identified47,48 as diminishing the protective close knit relations and solidarity among Latino members. Increased intergenerational conflict may arise as family members integrate their values and norms to better align with U.S. society. Our findings imply that interventions that combat discrimination and prevent family erosion during the integration process might be critical to ensure maintenance of mental health for Latino immigrants.
There are individuals in our sample that reported having both any depressive disorder and any anxiety disorder in the last year, suggesting there may be subgroups of the population (e.g., those with PTSD and major depressive episode) that have high rates of co-occurring disorders. In this paper, we analyzed anxiety and depression separately given differences in the disorders’ risk factors,49 correlates, and recommended treatments.50 However, we realize that these mental disorders are interrelated51,52 and recommend future studies investigating immigrant group differences in subpopulations with co-occurring disorders.
We find that last year risk of any psychiatric disorder, any depressive disorder, and any anxiety disorder increase in step with time in the U.S. for Latino immigrants. This might explain the desire for immigrant groups to maintain their cultural heritage53 as a way to buffer against potentially negative repercussions of assimilation on their mental health. Understanding this progression and identifying policy implications based on these results is made difficult by the fact that nativity differences in psychiatric disorder disappear when using regression models that adjust for all available covariates. In this paper, we present a method to examine these pathways, and find that increasing discrimination and family cultural conflict are possible pathways by which greater time in the U.S. associates with increased rates of psychiatric disorder.
The project was supported by the National Institute of Mental Health (grants U01 MH06220 and P50 MHO73469). The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Mental Health.
ContributorsB. Cook originated the study and led the conceptualization, design, and all aspects of writing the article. M. Alegría and J. Lin assisted with the conceptualization and design of the study, and assisted with drafting the article. M. Alegría, J. Lin, and J. Guo contributed to analysis and interpretation of the data. Statistical expertise was provided by J. Lin. All of the authors reviewed drafts and contributed to critical revisions of the article.
Human Participant Protection
The institutional review boards of the Cambridge Health Alliance, the University of Washington, and the University of Michigan approved all recruitment, consent, and interviewing procedures for the National Latino and Asian American Study.
Benjamin Lê Cook, Center for Multicultural Mental Health Research, Cambridge Health Alliance, Harvard Medical School, Somerville, MA.
Margarita Alegria, Center for Multicultural Mental Health Research, Cambridge Health Alliance, Harvard Medical School, Somerville, MA.
Julia Y. Lin, Center for Multicultural Mental Health Research, Cambridge Health Alliance, Harvard Medical School, Somerville, MA.
Jing Guo, Department of Health Studies at the University of Chicago, Chicago, IL.