Machine Learning Applications in Studying Mental Health Among Immigrants and Racial and Ethnic Minorities: An Exploratory Review

doi:10.21203/rs.3.rs-3575894/v1

Background

The use of machine learning (ML) in mental health (MH) research is increasing, especially as new, more complex data types become available to analyze. By examining the published literature, this review aims to explore potential gaps in the current use of ML to study MH in vulnerable populations of immigrants, refugees, migrants, and racial and ethnic minorities.

Methods

From October 2022 to June 2023, Google Scholar, EMBASE and PubMed were queried. ML-related terms, MH-related terms, and population-of-focus search terms were strung together with Boolean operators. Backward reference searching was also conducted. Included peer-reviewed studies reported using a method or application of ML in an MH context and focused on the populations of interest. We did not have date cutoffs. Publications were excluded if they were narrative or did not exclusively focus on a minority population from the respective country. Data including study context, the focus of mental healthcare, sample, data type, type of ML algorithm used, and algorithm performance was extracted from each.

Results

Ultimately, 13 peer-reviewed publications were included. All the articles were published within the last 6 years, and over half of them studied populations within the US. Most reviewed studies used supervised learning to explain or predict MH outcomes. Some publications used up to 16 models to determine the best predictive power. Almost half of the included publications did not discuss their cross-validation method.

Conclusions

The included studies provide proof-of-concept for the potential use of ML algorithms to address MH concerns in these special populations, few as they may be. Our review finds that the clinical application of these models for classifying and predicting MH disorders is still under development.

Machine learning

mental health

minorities

disparities

Common Mental Disorders (CMDs), including major depressive disorder, mood disorder, anxiety disorder, and alcohol use disorder, affect approximately one in five people worldwide (1, 2). More specifically, the global prevalence of post-traumatic stress symptoms is 24.1%, anxiety is 26.9%, sleep problems is 27.6%, depression is 28.0%, stress is 36.5%, and psychological distress is 50.0% (3).

Although racial and ethnic minorities have chronic exposures to systematic disadvantages and discrimination, paradoxically, black and Latinx populations have similar or better rates of mental health (MH) than non-Hispanic whites for most major MH disorders, including major depression (4–6). Furthermore, they have a lower lifetime prevalence of mood, anxiety, and substance use disorders than non-Hispanic whites (7, 8). Black and Latinx individuals, however, are at higher risk of persistence and disability from mental illness (8–10). Asian Americans have the best MH status compared to whites and other racial and ethnic minorities, but this is poorly studied (11). Ultimately, CMDs may disproportionately affect ethnic and racial minorities overrepresented in homeless, incarcerated, and medically underserved populations (4).

The intersection of mental illness stigma, interpersonal and structural discrimination, and low socioeconomic status perpetuates the disparities in MH outcomes (2, 12). Moreover, because minority populations and individuals with low socioeconomic status are more likely to have limited health literacy, have geographic inaccessibility to MH services, and lack medical insurance, MH disparities are further widened (13). Research shows that racial and ethnic minorities have a lower quality of MH services, are less satisfied with professional MH services, and have higher dropout rates than whites (7, 14–17).

Interestingly, immigrants tend to report having better MH than their counterparts in both their countries of origin and host countries, at least initially. This phenomenon is known as the "immigration paradox." Immigrants arrive healthier, but after facing the challenges and stresses of assimilation, their initial advantage deteriorates, and the healthy immigration effect (HIE) disappears (18, 19). Over time, the rates of mental illness worsen to match or exceed those of the general population, especially in psychotic disorders (18–20). A study from a Canadian national database shows that recent immigrants (landed within ten years) reported having better MH than native Canadians, but after the adjustment period, they became more likely to report worse MH levels than their Canadian‑born counterparts (18).

Refugees differ from immigrants in fleeing war, persecution, and/or political turbulence, whereas immigrants leave their native country to settle permanently in a foreign country. When broken down by immigrant admission categories, recently arrived refugees reported lower rates of self-reported MH than other immigrant groups and native-born Canadians (18). Asylum seekers and refugees are at substantially higher risk of psychiatric disorders and have up to 10 times the rate of post-traumatic stress disorder (20) due to exposure to expulsion factors such as war, poverty, and persecution.

Migrants experience unique MH stressors, resulting in different needs than native populations. The process of pre-migration, migration, and post-migration settlement requires changes in personal and social relationships and adopting new socioeconomic and cultural environments (19, 21, 22). These major transitions, with exposures to stress, helplessness, and poor resources, significantly impact mental well-being. A sense of powerlessness during the pre-migration and migration phases can initiate or worsen CMDs (23). However, during the post-migration resettlement phase, migrants may initially feel relief and success in reaching their destination, facilitating the HIE. Over time, the strain of rebuilding social networks while facing social alienation and reestablishing economic stability in the face of structural barriers and inequalities has negative consequences on MH status (19, 22, 24). Immigrants face racism, discriminatory and exclusionary policies, status loss, and sometimes violence, which further affect their mental well-being (19, 25, 26).

With more than 258 million people living outside their country of birth and millions more racial and ethnic minorities living amongst their majority counterparts, there is a need to understand and strengthen the MH resiliency of these populations. Clinicians and researchers have increasingly collected "big data" to aid this mission. This includes structured and unstructured data from electronic health records (EHR), smartphones, wearables, social media, and other large, complex sources. However, more robust data analysis methods are necessary to gain insights from this data. Machine learning (ML), the intersection of computer science, artificial intelligence, bioinformatics, and biostatics, offers methods to handle such data types (e.g., images, text, and tabular data) which cannot be handled by traditional analytical methods, like t-tests (27).

The use of ML in the health sciences is increasing. Several review articles discuss the current ML applications in MH and the development of better ML models in MH research but do not specifically include discussions of race, ethnicity, or immigration status (28, 29). Maslej et al. (30) conducted a rapid review and utilized a Critical Race Theory perspective to examine how race and racialization are defined in the application of ML in MH. However, their study focused on only Major Depressive Disorder and did not include other CMDs.

This review asks: what is the potential of ML to study MH in vulnerable populations of immigrants, refugees, migrants, and racial and ethnic minorities? Moreover, how feasible is it, and what are the limits of ML in how it is currently being used in this context? By exploring the published literature through this perspective, we aimed to examine potential biases and gaps in the literature.

Two reviewers (KP and AA) independently conducted exploratory searches in Google Scholar, EMBASE, PsycINFO and PubMed from October 2022 to June 2023. All queries had three components: an ML-related term (e.g. "machine learning," "artificial intelligence"), an MH-related term (e.g. "mental health," depression, "post-traumatic stress disorder") and a population of a focus search term (e.g. immigrant, refugee, minorit*, Latino). These terms were combined with the Boolean operators to create final search strings. Queries were conducted on titles, keywords, and abstracts. Backward reference searching was also conducted, reviewing references from the articles that matched our search criteria for more articles that could fit our inclusion criteria. See Appendix for full query syntax.

Inclusion criteria included: (i) the article reported using a method or application of ML in an MH context (ii) the primary population studied was immigrants, refugees, migrants, and/or racial and ethnic minorities (iii) the article was published in a peer-reviewed publication (iv) the article was available in English. We did not limit articles to just those published in America or have any date cutoffs. Articles were excluded if they were narrative in nature (e.g., commenting on future applications of ML in MH or did not add original contributions to ML in MH) or if they did not exclusively focus on a minority population from the respective country (e.g., a study of ethnically Chinese migrants in China would be excluded). Conflicts over inclusion were discussed, and a consensus was sought before the inclusion or exclusion of the publication in question.

Data were extracted from each article, including study context, the focus of mental healthcare, sample, data type, type of ML algorithm used, and algorithm performance. A narrative synthesis approach was applied.

To summarize the results of this review, we present them in three sections. The first section includes the results of the selection process. The second section details the characteristics of the selected studies, such as their area of focus, publication year and location, and data source. The third section highlights the machine learning models used in the studies for predicting and studying mental health outcomes.

3.1 Selected studies

Our search strategies resulted in a total of 22,082 listed articles from Google Scholar, EMBASE, PsycINFO and PubMed. Figure 1 shows the flow of our search strategy and results. All records from PsycINFO and PubMed were reviewed (60), an additional 280 records were reviewed from Google Scholar and the most relevant 100 were reviewed from EMBASE. Based on titles and abstracts, 79 were selected and further reviewed. Most of these records were excluded because they did not focus on the population of interest. Instead, they focused on majority populations and racially homogenous populations and/or did not include discussions about immigrant/migrant status. We also reviewed five abstracts from citation searching. Ultimately 13 publications were included in this review.

3.2 Publication Characteristics

Table 1 presents some high-level characteristics of the reviewed publications. All but two of the analyzed articles were published in the last three years, with two earliest from 2017 (31, 32). More than half of the papers were from the US or incorporated populations based in the US, four were from Europe, and the rest were from Asia. Among the 13 articles, five focused on refugee populations (32-36), three focused on Hispanic populations in the US (31, 37, 38), two focused on black individuals (39, 40), one on Native Americans (41), and the last two articles focused on Korean immigrants in the US (42) and immigrant populations in Europe (43). The areas of mental health focus included stress (35), ADHD (39, 40), trauma (32, 33), depression (36, 38, 40), PTSD (34), psychological distress (42), schizophrenia (43), suicidal ideation (37, 41), and substance abuse (31).

Surveys (32, 34, 36, 42), drawings (33), secondary data sets (including EHR data, surveillance data, and national sample sets) (31, 33, 37, 41, 43), internet-based posts (35, 38), and genomic sequencing data (39, 40) were analyzed in the included publications (see Table 2). Various populations were considered, and sample sizes varied widely due to the type of data collected and analyzed. For example, Augsburger and Elbert (32) enrolled 56 resettled refugees in a study to prospectively analyze their risk-taking.33 Goldstein, Bailey (37) used a retrospective dataset with 22,968 unique Hispanic patients, and Acion et al. (31) included 99,013 Hispanic individuals in their secondary data analysis. Children were also included in the reviewed studies; one examined the depression and PTSD levels of 631 refugee children residing in Turkey (34). Another study analyzed drawings from 2480 Syrian refugee children to find the predictors of exposure to violence and mental well-being (33). Other sample sets analyzed 0.15 million unique tweets from Twitter (35) and 441,000 unique conversations from internet message boards and social media sites (38). Genomic sequencing data was collected from 4,179 black individuals (40) and 524 black individuals (39).

Most reviewed studies used supervised learning intending to explain or predict certain MH outcomes. For example, to classify substance use disorder treatment success in Hispanic patients, Acion et al. compared 16 different ML models to an ensemble method they called "Super Learning" (31). Similarly, Huber et al. compared various ML algorithms, including decision trees, support vector machines, naïve Bayes, logistic regression, and K-nearest neighbor, to determine the model with the best predictive power for classifying Schizophrenia spectrum disorders in migrants (43). Two studies explored the impact of trauma exposure on MH using ML (32, 33). Two studies utilized social media data to understand MH at a population-health level through ML algorithms (35, 38). All study aims are found in Table 2.

3.3 Machine Learning Model Performance and Characteristics

Table 3 outlines a summary of ML characteristics and model performance. This review found that all 13 included publications fell into three categories: classification (31, 35, 37, 39-43), regression (32-34, 36), and unsupervised topic modeling (38).

The publications used a range of ML models, from one (32-34, 36, 39, 40, 42) to 16 (31). In studies where multiple ML models were used, the aim was often to compare the models to determine the best predictive power. For example, Acion et al. compared 16 models and evaluated them using the area under the receiver operating characteristic curve (AUC) to classify substance use disorder treatment success in Hispanic patients (31). Huber et al. compared five different ML algorithms, including decision trees, support vector machines, naïve Bayes, logistic regression, and K-nearest neighbor, to determine the model with the best predictive power for classifying Schizophrenia spectrum disorders in migrants (43). Two of the studies used linear regression (34, 36). All of the studies developed custom models to meet their study aims. The most common programs used in these studies were R (31, 32), SPSS (34, 42), and Python (35, 39, 40).

Predictors that were included in the modeling were sociodemographic characteristics (31, 34, 37, 41-43), and some also included MH variables and experiences (31, 32, 34, 37, 41-43) collected from EHRs or surveys. One study first determined which of the included 653 input variables (including sociodemographic data, childhood/adolescence experiences, psychiatric history, past criminal history, social and sexual functioning, hospitalization details, prison data, and psychopathological symptoms) were the best predictor variables and trained a final ML algorithm using only those (43).

Two studies did not report the best algorithm performance (37, 38). For the other studies, accuracy and AUC were commonly reported. For example, Acion et al. classified substance use disorder treatment success in Hispanic patients and found that the AUC of studied models ranged from 0.793 to 0.820, with the ensemble method achieving an AUC of 0.820, which was not significantly better than the traditional logistic regression model's AUC of 0.805 (31). Huber et al. identified a tree algorithm that differentiated native Europeans and non-European migrants with schizophrenia with an accuracy of 74.5% and a predictive power of AUC = 0.75 (43). In Liu et al., the trained ML model had an accuracy of 78% in predicting ADHD in African American patients (39). In a similar study to classify ADHD, depression, anxiety, autism, intellectual disabilities, speech/language disorder, developmental delays, and oppositional defiant disorder in African Americans, the model had an accuracy of 65% in distinguishing patients with at least one MH diagnosis from controls (40). A second prediction model aimed at predicting the diagnosis of two or more MH disorders had a low accuracy level, with an exact match rate of 7.2-9.3% (40). Khatua and Nejdl (35) analyzed tweets acquired from Twitter feeds from self-identified refugees and categorized them into themes of the immigrant struggle with an accuracy of 61.61% and 75.89%.

The included studies also used p-values to assess their ML algorithms. Goldstein and Bailey utilized multivariable logistical regression to examine the relationship between experienced discrimination and suicidal ideation in Hispanic patients (37). They found that 19.0% of Hispanic patients who experienced discrimination also experienced suicidal ideation, compared to 11.5% of patients that did not experience discrimination (p=0.001). Moreover, Hispanic patients had 1.72 greater odds of having suicidal thoughts if they experienced discrimination compared to those that did not (p=0.003). A study by Erol and Seçinti used regression analysis to study the relationship between PTSD and depression and various predictors in adolescent refugee minors (34). They found that moderate and severe changes in family income level and stress in food access predicted depression scores and PTSD symptoms (p < 0.01). Drydakis (36) used random effects models to estimate the relationship between the number of mobile applications that facilitate immigrants' societal integration and immigrants' integration, health, and mental health.37 The results showed a negative association between the number of standard m-Integration applications and adverse MH status (p < 0.01). Accuracy was also measured using importance and normalized importance (42), Root-mean-square error (RMSE) (32), and Least Absolute Shrinkage and Selection Operator (LASSO) coefficients (33).

3.4 Cross Validation

Six studies used internal cross-validation methods (31-33, 40, 41, 43). Only one study used an external data set to validate their ML algorithm (39). That external validation of the algorithm reduced the accuracy of their algorithm from 78% to 70-75% (39). Almost half of the included publications did not use or discuss their cross-validation method (34, 36-38, 42).

In recent years, there has been significant interest in the potential of ML to transform the field of MH research (29). Studies examining ML models have generally concluded that they outperform traditional statistical models, creating exciting possibilities for applying ML to studying MH in various populations. Recent advances in computational power and software availability have enabled researchers to reach new audiences and demonstrate the clinical value of ML. In particular, some studies have aimed to inform clinicians about the methods and applications of ML in the context of psychotherapy (44). However, while many of the reviewed papers provide proof-of-concept for the potential use of ML algorithms to address MH concerns, our review finds that the clinical application of these models for classifying and predicting MH disorders is still under development.

Despite ML's great interest and potential to transform MH research, few researchers have focused on specific and marginalized populations. In reviewing hundreds of articles on MH and ML, we found only a handful of studies specifically targeting immigrants, migrants, refugees, and/or racial and ethnic minorities. Many researchers simply included race as a variable in their models rather than designing ML algorithms to analyze these specific groups of individuals (45, 46). Moreover, as noted by Maslej et al. (30), most studies that considered African American and white samples used self-reported race or ethnicity or did not describe how this information was collected and thus were excluded from our analysis.

The current lack of ML models tailored to specific populations presents opportunities and challenges. On the one hand, it can help prevent the perpetuation of health disparities that arise when models built on majority populations are used to misclassify minorities (47). Performance differences in ML exist for different populations, especially with genomic data. For instance, one study externally validated their algorithm (39) on white Americans rather than African Americans and found that their algorithm's accuracy decreased. On the other hand, this lack of tailored models highlights the opportunity for researchers and clinicians to bridge the gap between what is known about majority populations and what is yet to be uncovered in other populations. Training ML models on other groups could expedite this process without being too resource-intensive.

One of the most common challenges in utilizing ML techniques to build classifiers for MH is the use of small sample sizes, which may limit the representation of the entire population and impact the generalizability of the classifier's accuracy estimate. This can be a practical limitation due to resource constraints in real-world clinical or diagnostic settings. However, researchers need to understand that using ML alone cannot address this issue (48). Most ML methods rely on supervised learning models, which are successful due to the abundance of training data. However, this training data requires human annotation, which can be time-consuming and costly. In the case of MH, there are insufficient publicly annotated datasets, making the quality of the data a significant concern for developing reliable models (46). Another challenge of using ML for behavioral diagnosis is validating the classification algorithms against questionnaires or clinical diagnoses, which are known to have self-report biases and are not completely accurate. This highlights the lack of established best standards in the diagnosis process for mental disorders and other psychiatric conditions (49). Future directions include the development of more robust and generalizable algorithms that can improve prediction capabilities. ML can be leveraged to understand the prevalence of MH conditions at a population level by using open-source and freely available data, which can be more accurate and less labor-intensive than traditional surveys. Furthermore, ML can enable the study of MH in children and adolescents in innovative ways (33, 40). The application of these models can be expanded to different sources and sample sizes, potentially leading to a rapid increase in their use in clinical settings.

There is also potential for the future application of ML and natural language processing (NLP) approaches to infer psychological well-being and detect CMDs in marginalized individuals based on their social media posts on platforms like Facebook and Twitter. Researchers must implement diagnostic criteria and tools that are precise and suitable for various online populations. Collecting personal information, such as sociodemographic characteristics and behavioral aspects, must be done in accordance with ethical considerations. These inferences can create online platforms that provide health information, support, and tailored interventions. Currently, the computational techniques and evaluations employed for collecting, processing, and utilizing online written data remain scattered throughout academic literature (50). Moreover, this potential is limited by factors such as class imbalance, noisy labels, and text samples that are either too long or too short, which can lead to performance and stability issues. The diversity of writing styles and semantic heterogeneity in different data sources can also cause a lack of robustness in model performance. Standardizing these measures can allow for the development of scalable approaches for automated monitoring of public psychological health in the future (40).

This review had limitations, including the possibility of missing relevant studies due to specificity in search terms. Future studies should consider using broader search terms to address these limitations. Additionally, the ethical and social implications of using ML in MH, including the potential for perpetuating existing biases and social determinants of health, should be carefully considered. Discussing ethical concerns is important when utilizing textual data related to MH, given the significance of privacy and security of personal information, particularly health data.

In conclusion, ML can potentially transform how we understand CMDs, particularly among vulnerable populations. Immigrants and refugees face unique challenges related to migration and resettlement that can negatively impact their MH status, including poverty, discrimination, and exposure to trauma. Meanwhile, African Americans and Hispanics in the US are overrepresented in marginalized populations and have higher persistence and disability from mental illness. This review has found that, to date, few studies have used ML to predict and classify MH in these populations, despite the wide gap in health disparities that persist in accessing quality MH services and outcomes. The use of big data and ML algorithms in the health sciences is increasing and holds promise, but more study of ML applications in MH is warranted.

Ethics approval and consent to participate: Not applicable
Consent for publication: Not applicable
Availability of data and materials: The search strings and data sets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.
Funding: Not applicable
Authors' contributions: KP and AA contributed to the conception, acquisition and analysis of this work. KP and AA drafted the work. KP, AA and MAG contributed to the revision of this manuscript. All authors have approved this work and agreed to its submission.
Acknowledgements: Not applicable

Steel Z, Marnane C, Iranpour C, Chey T, Jackson JW, Patel V, et al. The global prevalence of common mental disorders: a systematic review and meta-analysis 1980–2013. International journal of epidemiology. 2014;43(2):476-93.
Eylem O, De Wit L, Van Straten A, Steubl L, Melissourgaki Z, Danışman GT, et al. Stigma for common mental disorders in racial minorities and majorities a systematic review and meta-analysis. BMC Public Health. 2020;20(1):1-20.
Nochaiwong S, Ruengorn C, Thavorn K, Hutton B, Awiphan R, Phosuya C, et al. Global prevalence of mental health issues among the general population during the coronavirus disease-2019 pandemic: a systematic review and meta-analysis. Scientific Reports. 2021;11(1):1-18.
Satcher D. Mental health: Culture, race, and ethnicity—A supplement to mental health: A report of the surgeon general: US Department of Health and Human Services; 2001.
Rosenfield S. Triple jeopardy? Mental health at the intersection of gender, race, and class. Social Science & Medicine. 2012;74(11):1791-801.
Calzada EJ, Gulbas LE, Hausmann-Stabile C, Kim SY, Berger Cardoso J. Mental health issues within Latinx populations: Evaluating the state of the field. New and emerging issues in latinx health. 2020:45-62.
Ruiz P, Primm A. Disparities in psychiatric care: Clinical and cross-cultural perspectives: Lippincott Williams & Wilkins; 2010.
Kessler RC, Demler O, Frank RG, Olfson M, Pincus HA, Walters EE, et al. Prevalence and treatment of mental disorders, 1990 to 2003. New England Journal of Medicine. 2005;352(24):2515-23.
Breslau J, Cefalu M, Wong EC, Burnam MA, Hunter GP, Florez KR, et al. Racial/ethnic differences in perception of need for mental health treatment in a US national sample. Social psychiatry and psychiatric epidemiology. 2017;52(8):929-37.
Lê Cook B, McGuire TG, Zuvekas SH. Measuring trends in racial/ethnic health care disparities. Medical Care Research and Review. 2009;66(1):23-48.
Gong F, Xu J. Ethnic Disparities in Mental Health among Asian Americans: Evidence from a National Sample. Journal of Health Disparities Research and Practice. 2021;14(3):6.
Misra S, Jackson VW, Chong J, Choe K, Tay C, Wong J, et al. Systematic review of cultural aspects of stigma and mental illness among racial and ethnic minority groups in the United States: implications for interventions. American journal of community psychology. 2021;68(3-4):486-512.
Sanchez K, Ybarra R, Chapa T, Martinez ON. Eliminating behavioral health disparities and improving outcomes for racial and ethnic minority populations. Psychiatric Services. 2016;67(1):13-5.
Primm AB, Vasquez MJ, Mays RA, Sammons-Posey D, McKnight-Eily LR, Presley-Cantrell LR, et al. The role of public health in addressing racial and ethnic disparities in mental health and mental illness. Preventing chronic disease. 2010;7(1).
Safran MA, Mays Jr RA, Huang LN, McCuan R, Pham PK, Fisher SK, et al. Mental health disparities. American journal of public health. 2009;99(11):1962-6.
Miranda J, McGuire TG, Williams DR, Wang P. Mental health in the context of health disparities. american Journal of Psychiatry. 2008;165(9):1102-8.
Maura J, Weisman de Mamani A. Mental health disparities, treatment engagement, and attrition among racial/ethnic minorities with severe mental illness: A review. Journal of clinical psychology in medical settings. 2017;24(3):187-210.
Ng E, Zhang H. The mental health of immigrants and refugees: Canadian evidence from a nationally linked database. Health Reports. 2020;31(8):3-12.
Bas-Sarmiento P, Saucedo-Moreno MJ, Fernández-Gutiérrez M, Poza-Méndez M. Mental health in immigrants versus native population: a systematic review of the literature. Archives of Psychiatric Nursing. 2017;31(1):111-21.
Kirmayer LJ, Narasiah L, Munoz M, Rashid M, Ryder AG, Guzder J, et al. Common mental health problems in immigrants and refugees: general approach in primary care. Cmaj. 2011;183(12):E959-E67.
Rogler LH. International migrations: A framework for directing research. American Psychologist. 1994;49(8):701.
Bhugra D. Migration, distress and cultural identity. British medical bulletin. 2004;69(1):129-41.
Porter M, Haslam N. Predisplacement and postdisplacement factors associated with mental health of refugees and internally displaced persons: a meta-analysis. Jama. 2005;294(5):602-12.
Bhugra D, Ayonrinde O. Depression in migrants and ethnic minorities. Advances in Psychiatric treatment. 2004;10(1):13-7.
Noh S, Kaspar V, Wickrama KA. Overt and subtle racial discrimination and mental health: Preliminary findings for Korean immigrants. American journal of public health. 2007;97(7):1269-74.
Nakash O, Nagar M, Shoshani A, Zubida H, Harper RA. The effect of acculturation and discrimination on mental health symptoms and risk behaviors among adolescent migrants in Israel. Cultural Diversity and Ethnic Minority Psychology. 2012;18(3):228.
Callegaro M, Yang Y. The role of surveys in the era of “big data”. The Palgrave handbook of survey research: Springer; 2018. p. 175-92.
Shatte AB, Hutchinson DM, Teague SJ. Machine learning in mental health: a scoping review of methods and applications. Psychological medicine. 2019;49(9):1426-48.
Thieme A, Belgrave D, Doherty G. Machine learning in mental health: A systematic review of the HCI literature to support the development of effective and implementable ML systems. ACM Transactions on Computer-Human Interaction (TOCHI). 2020;27(5):1-53.
Maslej MM, Shen N, Kassam I, Rodak T, Sikstrom L. Race and Racialization in Mental Health Research and Implications for Developing and Evaluating Machine Learning Models: A Rapid Review. MEDINFO 2021: One World, One Health–Global Partnership for Digital Innovation. 2022:1088-9.
Acion L, Kelmansky D, van der Laan M, Sahker E, Jones D, Arndt S. Use of a machine learning framework to predict substance use disorder treatment success. PloS one. 2017;12(4):e0175383.
Augsburger M, Elbert T. When do traumatic experiences alter risk-taking behavior? A machine learning analysis of reports from refugees. PLoS one. 2017;12(5):e0177617.
Baird A, Cheng Y, Xia Y. Use of machine learning to examine disparities in completion of substance use disorder treatment. PloS one. 2022;17(9):e0275054.
Erol E, Seçinti DD. Examination of PTSD and Depression Levels and Demographic Data of Syrian Refugee Children during the Pandemic. Psych. 2022;4(2):215-25.
Khatua A, Nejdl W, editors. Struggle to Settle down! Examining the Voices of Migrants and Refugees on Twitter Platform. Companion Publication of the 2021 Conference on Computer Supported Cooperative Work and Social Computing; 2021.
Drydakis N. Mobile applications aiming to facilitate immigrants’ societal integration and overall level of integration, health and mental health. Does artificial intelligence enhance outcomes? Computers in Human Behavior. 2021;117:106661.
Goldstein EV, Bailey EV, Wilson FA. Discrimination and Suicidality Among Hispanic Mental Health Patients, 2010–2020: A Natural Language Processing Approach. Psychiatric services. 2022;73(11):1313-4.
Castilla-Puentes R, Dagar A, Villanueva D, Jimenez-Parrado L, Valleta LG, Falcone T. Digital conversations about depression among Hispanics and non-Hispanics in the US: A big‐data, machine learning analysis. 2021.
Liu Y, Qu H-Q, Chang X, Nguyen K, Qu J, Tian L, et al. Deep learning prediction of attention-deficit hyperactivity disorder in African Americans by copy number variation. Experimental Biology and Medicine. 2021;246(21):2317-23.
Liu Y, Qu H-Q, Mentch FD, Qu J, Chang X, Nguyen K, et al. Application of deep learning algorithm on whole genome sequencing data uncovers structural variants associated with multiple mental disorders in African American patients. Molecular psychiatry. 2022;27(3):1469-78.
Haroz EE, Walsh CG, Goklish N, Cwik MF, O’Keefe V, Barlow A. Reaching those at highest risk for suicide: development of a model using machine learning methods for use with Native American communities. Suicide and Life‐Threatening Behavior. 2020;50(2):422-36.
Choi S, Hong JY, Kim YJ, Park H. Predicting psychological distress amid the COVID-19 pandemic by machine learning: discrimination and coping mechanisms of Korean immigrants in the US. International journal of environmental research and public health. 2020;17(17):6057.
Huber DA, Lau S, Sonnweber M, Günther MP, Kirchebner J. Exploring similarities and differences of non-European migrants among forensic patients with schizophrenia. International Journal of Environmental Research and Public Health. 2020;17(21):7922.
Hitczenko K, Cowan HR, Goldrick M, Mittal VA. Racial and ethnic biases in computational approaches to psychopathology. Oxford University Press US; 2022. p. 285-8.
Prout TA, Zilcha-Mano S, Aafjes-van Doorn K, Békés V, Christman-Cohen I, Whistler K, et al. Identifying predictors of psychological distress during COVID-19: a machine learning approach. Frontiers in Psychology. 2020;11:586202.
Zhang Y, Wang S, Hermann A, Joly R, Pathak J. Development and validation of a machine learning algorithm for predicting the risk of postpartum depression among pregnant women. Journal of affective disorders. 2021;279:1-8.
Chen IY, Pierson E, Rose S, Joshi S, Ferryman K, Ghassemi M. Ethical machine learning in healthcare. Annual review of biomedical data science. 2021;4:123-44.
Cho G, Yim J, Choi Y, Ko J, Lee S-H. Review of machine learning algorithms for diagnosing mental illness. Psychiatry investigation. 2019;16(4):262.
Richter T, Fishbain B, Fruchter E, Richter-Levin G, Okon-Singer H. Machine learning-based diagnosis support system for differentiating between clinical anxiety and depression disorders. Journal of Psychiatric Research. 2021;141:199-205.
Calvo RA, Milne DN, Hussain MS, Christensen H. Natural language processing in mental health applications using non-clinical texts. Natural Language Engineering. 2017;23(5):649-85.

Table 1. Publication analysis

Characteristic	N	%	Reference
Year of publication
2017	2	15.4%	(31, 32)
2020	3	23.1%	(42) (43) (41)
2021	5	38.5%	(35, 36, 38-40)
2022	3	23.1%	(33, 34, 37)
Region
Asia
Turkey	1	8.3%	(34)
Jordan	1	8.3%	(33)
Europe
United Kingdom	1	7.7%	(36)
Germany	2	15.4%	(32, 35)
Switzerland	1	7.7%	(43)
US	7	53.8%	(31, 37-42)
Population of focus
Refugees	5	38.5%	(32-36)
Hispanics	3	23.1%	(31, 37, 38)
Native Americans	1	7.7%	(41)
African Americans	2	15.4%	(39, 40)
Korean immigrants	1	7.7%	(42)
European immigrants	1	7.7%	(43)

Table 2. Publication characteristics of included studies.

First Author (year)	Study Aim	Area of MH focus	Sample size and characteristics	Data analyzed
Acion (2017) (31)	Predict substance abuse treatment success using 17 different machine learning models	Substance abuse	99,013 Hispanic individuals	TEDS-D 2006-2011
Augsburger (2017) (32)	Assessed risk-taking behavior in refugees after exposure to trauma using a gamified BART	Trauma	56 Refugees resettled in Germany	Surveys and data on BART
Baird (2022) (33)	Used drawings by refugee children to estimate predictors of exposure to violence and mental wellbeing	Trauma	2480 Syrian refugee children	USF 2016 dataset
Castilla-Puentes (2021) (38)	To understand how Hispanic populations converse about depression by conducting big data analysis of digital conversations through machine learning	Depression	441,000 unique conversations about depression; 43,000 (9.8%) conversations were by Hispanics	Conversations from open sources like topical MH websites, message boards, social networks, and blogs
Choi (2020) (42)	Examined the predictive ability of discrimination-related variables, coping mechanisms, and sociodemographic factors on the psychological distress level of Korean immigrants in the U.S. during the pandemic	Psychological distress	790 Korean immigrants, foreign and US-born	Surveys
Drydakis (2021) (36)	Understanding associations between the number of mobile applications in use aiming to facilitate immigrants’ societal integration and increased level of integration, good overall health, and mental health	Depression	287 immigrants in Greece	Surveys
Erol (2022) (34)	Examine the PTSD and depression levels of Syrian refugee children and adolescents, the difficulties they experienced in access to food and education, and the changes in their family income, and evaluate the effects of these factors on symptom severities of depression and PTSD	Depression & PTSD	631 Refugee children living in Turkey	Surveys
Goldstein (2022) (37)	To examine the relationship between experiencing discrimination and suicidal ideation in Hispanic populations	Suicidal ideation	22,968 Hispanic individuals	Holmusk and MindLinc EHR datasets, 52,703 patient-year observations from 2010 to 2020
Haroz (2020) (41)	Develop a model using ML methods to better identify those at highest risk for suicide in Native American communities	Suicidal ideation	2,390 Native American individuals	Surveillance program data
Huber (2020) (43)	Differentiated native Europeans and migrants as to their risk of having schizophrenia	Schizophrenia	370 patients with diagnosed schizophrenia spectrum disorder	Hospital data from 1982 to 2016
Khatua (2021) (35)	Using social media data to identify the voices of migrants and refugees and analyze their MH concerns	General MH	0.15 million tweets, 2% from self-identified refugees	0.15 million tweets
Liu (2021) (40)	Used ML algorithms to distinguish ADHD, depression, anxiety, autism, intellectual disabilities, speech/language disorder, delays in development, and oppositional defiant disorder in blacks using the data from their genome.	ADHD, depression, anxiety, autism, intellectual disabilities, speech/language disorder, delays in development, oppositional defiant disorder	4179 black individuals	Genomic sequencing data
Liu (2021) (39)	Used ML algorithms to distinguish ADHD in blacks using the data from their genome.	ADHD	524 black individuals	Genomic sequencing data

Abbreviations: Balloon analogue risk task (BART), Treatment Episode Data Set-Discharge (TEDS-D)

Table 3. Machine Learning model characteristics from selected articles

First Author (year)	Outcome Variable	Predictors (Input variables)	ML technique	Cross-validation method (internal, external)	Type	Program used	Best algorithm performance
Acion (2017) (31)	Substance abuse treatment success	28; 10 patient characteristics, 3 treatment factors, referral type, problematic substance characteristics and mental health problem	LR, RLR, Lasso-LR, EN, RF, DNN, EL	Two-fold cross-validation (I)	Classification	R; H2O R interface and package rROC	AUC: 0.793 - 0.820 Best mode: EL
Augsburger (2017) (32)	Risk-taking behavior as measured using a balloon analog risk task (BART)	Exposure to different types of childhood maltreatment, experiences of war and torture, lifetime traumatic events and symptoms of depression and PTSD, sociodemographic factors	Stochastic GBM	Tenfold cross-validation with three repetitions (I)	Regression	R; gbm & caret	RMSE: 18.70, R^2: .20,
Baird (2022) (33)	Psychological trauma as measured on the GHQ-12	18 digitally coded features in self-portraits and free drawings	One model method used: LASSO-R	K-fold cross-validation (I)	Regression	Not reported	R-squared: 0.108
Castilla-Puentes (2021) (38)	Tone, topics, and attitude of digital conversations	Digital conversations	NLP and texting mining	Not used	Unsupervised- Topic modeling	CulturIntel	Not reported
Choi (2020) (42)	Psychological distress is measured using the Kessler Psychological Distress Scale (K10)	Demographic characteristics, three types of discrimination characteristics, three types of coping mechanisms	ANN	Not used	Classification	SPSS	AUC: 0.806
Drydakis (2021) (36)	Increased level of integration, overall health, and mental health	Number of mobile applications in use that facilitate immigrants' societal integration	Linear Regression	Not used	Regression	Not reported	p < 0.005
Erol (2022) (34)	Symptom severity of depression and PTSD	Demographic data, PTSD and depression levels, access to food and education, and changes in family income	Linear regression	Not used	Regression	SPSS	R-squared = 0.123
Goldstein (2022) (37)	Suicidal ideation in the past year	Experience of discrimination, demographics	Deep-learning NLP algorithms and LR	Not used	Classification	Not reported	Not reported
Haroz (2020) (41)	Suicide attempt, measured at 6, 12, and 24 months after an initial suicide-related event	73; demographic characteristics, educational history, past mental health, substance use, living status, history of domestic violence, participation in tribal activities, knowing anyone who died by suicide in their lifetime and number of indexed events	RF, SVM, Lasso-R, RLR	Repeated cross-validation with 10 iterations (I)	Classification	Not reported	AUC: 0.87
Huber (2020) (43)	Migrant status	653 variables	LR, DTs, SVM, and naive Bayes	5-fold cross-validation (I)	Classification	Not reported	DT Accuracy: 74.5%; AUC: 0.75
Khatua (2021) (35)	Tweets that fall into 3 themes: generic views, initial struggles, and subsequent settlement	Tweets	Bi-LSTM, CNN, BERT	Training and testing	Classification	Python	F1-Score: 61.61 - 75.89%
Liu (2021) (40)	MH diagnosis from EHR	Copy number variation	Multi-layer perceptron	Two-fold random shuffle test validation (I)	Classification	Python; Scikit-learn package	Accuracy: 65.7%
Liu (2021) (39)	ADHD diagnosis	Copy number variation	Multi-layer perceptron	Two-fold random shuffle test validation (E)	Classification	Python; Scikit-learn package	Accuracy: 75.4%

Abbreviations: Logistic regression (LR), Ridge logistic regression (RLR), Least Absolute Shrinkage and Selection Operator, (Lasso-LR), random forests (RF), deep learning neural networks (DNNs), Ensemble learning (EL), Lasso-Regression (Lasso-R), gradient boosting machines (GBM), Natural language processing (NLP), Artificial Neural Network (ANN), decision trees (DTs), support vector machines (SVM), Bidirectional Long Short-Term Memory (Bi-LSTM) and Convolutional neural network (CNN), Bidirectional Encoder Representations from Transformers (BERT), area under the receiver operating characteristic Curve (AUC), Root-mean-square error (RMSE)

No competing interests reported.

AppendixA.docx

Machine Learning Applications in Studying Mental Health Among Immigrants and Racial and Ethnic Minorities: An Exploratory Review

Status:

Journal Publication

Version 1

Abstract

Figures

1. Introduction

2. Methods

3. Results

4. Discussion

5. Conclusions

Declarations

References

Tables

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1