Background

Depression is the most common mental health disorder and the second leading cause of disability in the USA [1]. Prescription medications for depression, typically antidepressants, are a primary treatment modality in the USA and improve depression symptoms in most patients [2, 3]. Latinx individuals comprise an estimated 18.3% of the US population and represent the largest US minority group [4]. Among US Latinx adults, antidepressant use is associated with improved depression assessment scores and depression symptom remission [5]. Several factors related to depression medication use have been attributed to Latinx ethnicity and culture at the individual patient level. These factors include age, sex, stigma, familism, gender roles, health beliefs (such as perceptions of depression symptom severity and concerns about antidepressant side effects and discontinuation symptoms), personal health literacy, religious beliefs, and medical mistrust [5,6,7,8,9,10,11,12,13,14,15,16].

The COVID-19 pandemic exacerbated existing health disparities between Latinx and non-Latinx White (NLW) individuals and highlighted the need to shift the focus of accountability and responsibility of health care from the individual to public health and healthcare systems [17, 18]. Social determinants of health (SDH) are the conditions that people live, work, play, and learn that have a greater influence on health outcomes than genetic profile or medical care [17]. SDH inequities are rooted in racism and ethnic discrimination within economic, legal, educational, environmental, community, and healthcare systems and lead to health disparities and poor health outcomes among underrepresented populations. Ethnic determinants of health (EDH) are SDH with a unique effect for an ethnic group relative to the dominant group [18]. EDH-related health disparities are grounded in ethnic discrimination and confounded by the heterogeneity of the US Latinx population. Various quantitative and qualitative studies have identified systems-level factors associated with depression medication use among Latinx adults including economic stability, education level, health insurance, access to language and culturally concordant mental health care, and organizational health literacy [5,6,7,8,9,10,11,12, 19]. However, individual- and systems-level factors have not been assessed concurrently within a national sample of Latinx adults with major depressive episode (MDE) or prescription medication use for depressive symptoms, or with consideration for place of origin for more than 2 populations (Mexico and all other Latinx). Identifying the relationships of individual- and systems-level EDH factors between NLW and Latinx adults and among Latinx adults from different places of origin could provide a deeper understanding of the importance of individual- and systems-level EDH factors on depression medication use and support hypothesis building for future research.

For this study, EDH factors are based on the ethnic and disparities framework (Table 3 appendix) [18] and are categorized within the levels of analysis of environment-societal structures which includes geographic (metropolitan/non-metropolitan) location, education, and economic stability; and social-cognitive-behavioral which comprises stress and stress processes and cultural factors.

To assess EDH variables across a large, heterogeneous Latinx adult sample for hypothesis building, we compared national survey data of self-reported prescription medication use for depressive symptoms and proxy EDH variables among NLW, Latinx, and Latinx populations from Puerto Rico, Mexico, and other Latinx countries using an ethnic and disparities framework [18]. We hypothesize that differences between NLW and Latinx populations with MDE and with depression medication use will be noted among individual- and systems-related EDH factors; systems-related EDH variables such as access to prescribers and health insurance will no longer be different among populations with depression medication use.

Methods

We conducted a retrospective analysis of 4 years of National Survey on Drug Use and Health (NSDUH) datasets, from 2015 to 2018, using the online Restricted-use Data Analysis System (RDAS) for hypothesis building [20, 21]. The NSDUH is an in-person, nationwide, cross-sectional, annual survey sponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA). Survey participants answer questions about drug use, mental and emotional health issues, and mental health treatment. NSDUH data are weighted [22] to represent the US population. Annually, the NSDUH has responses from approximately 68,000 individuals ≥ 12 years old from all 50 states and the District of Columbia. The survey does not collect data from individuals who reside in Puerto Rico, a US territory with a majority Latinx, bilingual population, but does represent individuals within the USA who report Puerto Rico as their place of origin. NLW is defined by the NSDUH codebook as self-identified “Not Hispanic/Latino and White Only” and Hispanic/Latino as self-identified “Hispanic, Latino, or Spanish origin or descent” without regard to race. We choose to use the term Latinx to represent Hispanic/Latino individuals to reflect a gender inclusive term used in academic settings [23].

The SAMHSA Archive provides online access to the NSDUH RDAS files compiled by several years and accounts for changes in survey methodology over time [20,21,22]. RDAS allows for limited analysis of demographic and EDH factors that were collected and subsequently coded by NSDUH. EDH variables are proxies for EDH factors that were selected based on the ethnic and disparities framework (Table 3 appendix) [18]. Using the RDAS website, we compared EDH variables among NLW, Latinx, and Latinx adults with MDE from Puerto Rico, Mexico, and other Latinx countries (Table 1); then using the same EDH variables, this process was repeated for respondents who reported prescription medication use for depressive symptoms (Table 2). The RDAS will not return results that fail to comply with SAMHSA respondent disclosure limitations [21]. Relevant to analyses limitations, the RDAS would not return results for the individual populations of Cuba, Central or South American countries, or the Dominican Republic [20,21,22]. Therefore, the responses from these populations are found within the other Latinx countries group.

Table 1 Characteristics of respondents with major depressive episode in past year by ethnicity and place of origin
Full size table
Table 2 Characteristics of respondents with prescription medication use in past year for depressive symptoms by ethnicity and place of origin
Full size table

The NSDUH adopts the definition for major depressive episode (MDE) from the Diagnostic and Statistical Manual of Mental Disorders 5th edition (DSM-5) [24]. MDE was defined as “a period of at least 2 weeks of depressed mood or loss of interest or pleasure in daily activities and specified symptoms, such as problems with sleep, eating, energy, concentration, or self-worth” in their lifetime or past year. NSDUH respondents were asked questions about prescription medication use for a mental or emotional condition or for depressive symptoms in the past year which is represented by the recoded variable of “prescription medication use for depressive symptoms.” Respondents with unknown data for past year MDE measures or unknown prescription medication use for depressive symptoms were excluded by NSDUH [22].

For our analyses, we included all relevant and assessable (via the RDAS), individual- and systems-related EDH variables from NSDUH survey questions that correspond with the ethnic and disparities framework. [18] For the systems-related environmental-societal structure level of analysis, the construct domains and representative EDH variables used were geographic location of residence (metropolitan size vs non-metropolitan) and education (level [less than high school, high school, some college or associates degree, and college graduate]; limited English proficiency [speaks English very well and well to not very well at all]; and survey language preference [Spanish or English]). For the construct domain of economic stability, the EDH variables used were employment status (employed full time, employed part time, unemployed, not in work force), income status (federal poverty level, health insurance any type), and occupational stress (depressive symptoms impairment of ability to work [none to mild and moderate to severe]) [18]. Additionally, in this level of analysis, we added the construct domain of healthcare system; [26] the representative EDH variables were saw/talked to primary care or family medicine doctor, saw/talked to psychologist, saw/talked to social worker, and/or prescription medication use for depressive symptoms; all which were dichotomous yes or no.

For the individual-level social-cognitive-behavioral level of analysis, the representative EDH variables used were stress and stress processes in the past year (serious thoughts and plans of suicide [yes or no]; serious psychological distress [yes or no]; and MDE with severe role impairment [yes or no]). [18] Individual cultural factors were stigma (no mental health treatment due to concerns about confidentiality); beliefs about health and healing (no mental health treatment because not needed; and no mental health treatment because not helpful) which were all yes or no options; and religious beliefs influence on decisions (strongly disagree to disagree and in agree to strongly agree). The systems-level environment-physical characteristics and individual-level community and interpersonal levels of analysis did not have any construct domains with possible EDH variables from any NSDUH survey questions.

The RDAS generates 2-way cross tabulations that depict the number of times each of the possible variable combinations occurred in the sample data [21]. Results are reported as percentages and 95% confidence intervals (CIs) which reflect the survey’s stratified, clustered design. The independent variables of ethnicity were NLW, Latinx, and place of origin (Puerto Rico, Mexico, other Latinx countries). For each dependent EDH variable, responses from NLW to Latinx adults with MDE in the past year were compared as well as responses among adults with MDE from Puerto Rico, Mexico, other Latinx countries (Table 1). These analyses were then rerun for respondents who reported prescription medication use for depressive symptoms in the past year (Table 2).

Bivariate statistical calculations were performed via the RDAS using the survey package in R [27]. The RDAS calculates the 95% CI using the Taylor series linearization method, assuming a with-replacement design [28]. Statistical significance (P < 0.05) was determined using the Wald chi-square test of association. When no P value was reported by the RDAS output and when determining differences between Latinx place of origin populations, statistical significance was determined if the respective 95% CIs did not overlap [29]. This study received a status of not regulated from the University of Michigan Institutional Review Board (IRBMED HUM00147519) as it is an analysis of publicly available datasets and conducted using a publicly available tool. NSDUH RDAS dataset content is strictly controlled to prevent identification of individuals or their personal health information as indicated by the RDAS not returning results that fail to comply with SAMHSA’s disclosure limitations [25]. This analysis was completed in January 2023.

Results

The total unweighted sample size for 2015 to 2018 NSDUH respondents older than 18 years is 202,500. [15,16,17] This corresponds to a weighted US adult population (greater or equal to 18 years of age) of 245,838,000 people. Of these,15.9% are Latinx adults normalized to the national proportion of US Latinx individuals during the time of each survey [20,21,22]; reported place of origin for Latinx adults is Puerto Rico (1.8%; not living in Puerto Rico), Mexico (9.6%), and other Latinx countries (4.6%). Analysis of other individual Latinx countries was restricted by the RDAS not returning results [20, 21]. The total weighted sample size for respondents with MDE in the past year is 16,858,000 (Table 1). Latinx tend to be younger than NLW adults with MDE; more Latinx than NLW adults with MDE live in large metropolitan and less in non-metropolitan geographic locations.

General Status of Mental and Physical Health of Latinx and NLW Adults

Proportionately, the Latinx population reported a significantly lower prevalence of MDE yet similar rates of impairment with MDE compared with the NLW population, as exemplified by the lower rate of past year MDE and similar rates of overall health and depression severity reported (Table 1).

Disparities in the Care and Support Reported by the Latinx Population

While the prevalence of MDE may be lower among Latinx adults, there is a significant gap in the proportion of Latinx compared with NLW population with MDE that receive healthcare as shown by the significantly lower proportion of Latinx adults who saw/talked to a primary care or family medicine doctor (− 12.4% difference; P < 0.0001), saw/talked to a psychologist (− 4.3% difference; P < 0.0001), and reported prescription medication use for depressive symptoms (− 16.6% difference; P < 0.05) compared with NLW adults. Results suggest that several systems-level EDH factors could have contributed to these disparities among those with MDE. Economic stability was significantly different between populations with more Latinx than NLW reporting they were unemployed, living below the federal poverty threshold, and lacked health insurance. The highest education level achieved was also significantly different between populations; 9.8% more Latinx than NLW adults with MDE had less than a high school education. Less Latinx than NLW adults reported English language proficiency; 24% of Latinx adults with MDE spoke English well to not well at all. Additionally, less Latinx than NLW adults (− 22.5% difference; P < 0.05) were born in the USA. Responses to questions regarding individual-level EDH factors like concerns about confidentiality (a proxy for stigma), and beliefs about healing and health and reasons to not seek mental health treatment (proxies for health beliefs) were not significantly different between populations. Additionally, no difference was observed between Latinx and NLW adults with MDE and the influence of religious beliefs on their decisions.

Latinx and NLW Adults Who Took a Prescription for Depressive Symptoms in the Past Year

As shown in Table 2, a significantly lower proportion of Latinx adults took a prescription medication for depressive symptoms in the past year compared with NLW adults (28.4 to 40.6%, respectively; P < 0.0001). Overall, good to excellent health status was lower for both populations; however, it was significantly lower for Latinx compared with NLW adults with depression medication use. No statistically significant differences between populations regarding sex were reported. However, approximately 2.4 more females reported medication use for depression than males. Latinx adults proportionately take medication at earlier phases of adulthood compared to NLW adults. As NLW adults age, the use of medication increases; however, for Latinx adults, those between 26 and 49 years have the highest use of medications for depressive symptoms. Among those who reported depression medication use, a larger proportion of the Latinx than NLW population resides in large metropolitan geographic location (62.5 to 48.0%, respectively). Several systems-level EDH variables continued to be different such as poverty level, education status, limited English proficiency, and income. Individual cultural factors related to health beliefs, religious beliefs influencing decisions, and stigma were not different between the two populations. Similarly, seeing/talking to a primary care provider or seeing/talking to a psychologist was no longer significantly different between populations, indicating improved access to medical care for those reporting medication for depression in past year.

Latinx Adults from Puerto Rico, Mexico, and Other Latinx Countries

As shown in Table 1, respondents from Puerto Rico represent the largest proportion of Latinx adults with MDE, reported the lowest rate of good to excellent health, the highest rate of poverty, and highest rate of serious psychological distress. Yet adults with MDE from Puerto Rico have the highest rates of English proficiency, prescription medication use for depressive symptoms, health insurance, and access to primary care or family medicine doctors compared with adults from Mexico and other Latinx countries. In comparison, adults from Mexico represent the smallest Latinx population with MDE and lowest rates of serious psychological stress, prescription medication use for depression symptoms, and English proficiency. However, they report the highest rates of suicide, poverty, and being uninsured compared with the overall Latinx population. Interestingly, no differences were reported among Latinx populations for individual cultural factors such as stigma or health beliefs and seeking mental health treatment or religious beliefs influencing health decisions.

Latinx Adults from Puerto Rico, Mexico, and Other Latinx Countries Who Took a Prescription for Depressive Symptoms in the Past Year

After limiting the findings to respondents who reported prescription medication use for depressive symptoms, Latinx adults from Puerto Rico, Mexico, and other Latinx countries had no differences for geographic location, education level, seeing a primary care provider or psychologist, and health insurance (Table 2). Additionally, stress, depression severity, good to excellent health, beliefs about healing, and stigma were not different among Latinx populations with depression medication use. Significant differences among Latinx populations were observed with economic stability and education level as shown by the different rates of poverty, limited English proficiency, Spanish language survey, and individuals < 10 years living in USA.

Discussion

We conducted an analysis of individual- and systems-related factors using an EDH lens [18]. The rate of prescription medication use for depressive symptoms was 16.6% (P < 0.05) less for Latinx than NLW adults with MDE according to data from a national sample showed. Differences in systems-level EDH variables between Latinx and NLW populations with MDE were also found between populations with depression medication use. Of note, seeing a primary care doctor or seeing a psychologist was significantly different among Latinx and NLW populations with MDE but not for those with depression medication use. Individual-related EDH variables of overall health and serious psychological distress were significantly worse for Latinx than NLW populations who reported depression medication use. However, stigma, healing beliefs, and religious beliefs were not different between populations with MDE or depression medication use. Differences between certain Latinx populations with depression medication use were found with limited English proficiency and poverty.

Findings from this preliminary analysis align with previously reported differences between NLW and Latinx populations with depression at the systems-level such as access to a health care provider, poverty level, health insurance, English proficiency, and education level as barriers to medication use for depression. [3, 5, 12, 18] As previously reported, individual-level cultural factors such as stigma, health beliefs, and religious beliefs may be relevant for Latinx adults and medication use for depression. [5,6,7,8,9,10,11, 13,14,15,16, 29, 30] However, in our analysis, no differences in these factors were observed between Latinx and NLW adults with MDE or reporting medication use for depression. Poverty and English proficiency remained significantly different among Latinx populations reporting medication use.

The ethnic and disparities framework [18] that was applied in this study was modified to include a healthcare systems level of analysis as described by a health disparities framework [26]. While not specific to the Latinx population, it was needed to account for healthcare provider interactions and medication use.

Limitations

These analyses have several limitations including those associated with retrospective and bivariate analyses of national survey datasets. The NSDUH data are collected by self-report which can be biased; actual rates of diagnoses, objective assessments, and clinical evaluations may be different than what were reported. Because of some RDAS restrictions and NSDUH methodology, we were unable to assess unweighted sample sizes, make comparisons between included and excluded populations, or evaluate additional Latinx countries of origin or persons currently living in Puerto Rico. Unrestricted access to the NSDUH data could allow for more complex analysis of additional variables to create a model that could better determine differences among ethnic determinants. While the results are weighted to reflect a national sample, generalizability could be limited by lack of representation of Latinx adults from all places of origin. Several important EDH factors were not assessed because of RDAS restrictions and because the NSDUH does not ask relevant questions for environmental-physical characteristics or community and interpersonal levels of analysis [18]. The respective construct domains contain several EDH factors that could influence depression treatment and could be modifiable targets for future interventions such as discrimination, quality of housing, family support, other types of stress and stress management, as well as values, norms, and traditions.

When the RDAS would not provide a P value, we adopted no CI overlap between populations to represent statistical significance (P < 0.05). This more conservative method rejects the null hypothesis less often than standard statistical methods when the null hypothesis is true [28]. Thus, relying on the overlap of confidence intervals to determine significance may be too conservative an estimate and result in some significant findings not being reported. Limitations of the use of a bivariate analysis or cross tabulation and the chi-square test statistic in this study are that they do not reveal the cause, strength, or direction of the relationship(s) nor do they consider the effects of other variables on the relationship(s) which may or may continue to be statistically significant under multivariable adjustments.

Despite these limitations, these study findings provide some insight into relationships between Latinx and NLW populations with MDE and medication use for depressive symptoms. Results suggest that seeing a primary care provider or psychologist or social worker, serious psychological distress, and overall health status are important factors related to depression medication use among Latinx adults. Given the limitations of these analyses, multivariate analyses of EDH factors and other Latinx populations are warranted to further describe these relationships.

Conclusion

An analysis of NSDUH datasets from 2015 to 2018 using an ethnic and disparities framework supports previous reports regarding systems- and individual-related cultural factors and medication use for depression among Latinx adults. EDH variables that appear to be related to depression medication use among Latinx adults are seeing a primary care provider or behavioral health specialist, depression severity, and overall health status, whereas some individual factors such as stigma, religious beliefs, and health beliefs may not be as relevant. Based on these preliminary findings, expanding investigations of the described EDH factors; adding EDH factors related to built environment, social and community context, and family; and assessing other Latinx countries individually are warranted.