Predictors of mental health outcomes of three refugee groups in an advocacy-based intervention: A precision medicine perspective

Abstract

OBECTIVE:

Precision medicine is an area with great potential for mental health, but has made limited gains prognostically in predicting effective treatments. For refugees exposed to violence, culture may be a crucial factor in predicting treatment outcomes.

METHOD:

For this study, 290 participants from three regions (Afghanistan, the Great Lakes region of Africa, and Iraq and Syria) participated in a randomized controlled trial of an advocacy-based intervention. Emotional distress symptoms were measured prior to intervention, mid-intervention (3 months), post-intervention (6 months), and follow-up (6 months after the end of intervention). Number of traumatic events, resource access, social support, and English proficiency were tested for potential predictive effects on intervention outcome.

RESULTS:

Multi-level generalized linear models revealed that Afghans’ (B = −0.259, SE = 0.108, p = .013), and Great Lakes Africans’ (B = −0.116, SE = 0.057, p = .042) emotional distress symptoms improved as a function of the intervention, while Iraqis and Syrians showed no intervention effects. For Afghans, English proficiency (B = −0.453, SE = 0.157, p < .01) and social support (B = −0.179, SE = 0.086, p = .037) were most strongly correlated to emotional distress, while for Africans, resource access (B = −0.483, SE = 0.082, p < .001) and social support (B = −0.100, SE = 0.048, p = .040) were the strongest predictors of emotional distress.

CONCLUSIONS:

Response to advocacy-based interventions and active ingredients may be influenced by culture; findings have implications for refugees and precision medicine.

In 1967, Gordon Paul noted, “The question towards which all outcome research should ultimately be directed is the following: What treatment, by whom, is most effective for this individual with that specific problem, and under which set of circumstances?” Today, while still struggling to answer clearly that seminal question, precision mental health research holds great promise for improving the effectiveness of mental health interventions (Delgadillo & Lutz, 2020). As a longstanding area of research and interest for clinical scientists, precision medicine aims to identify multivariable predictors of risk and treatment success in specific groups of people, which indicate good “fit” with an intervention. However, in spite of interest in these differential predictors of fit, precision medicine (sometimes called personalized medicine) research is almost exclusively focused on WEIRD (Western, European, Industrialized, Rich, and Democratic) sample groups. This means that there is a lack of knowledge about whether these treatment effects generalize to populations in the rest of the world, which is a weakness of considerable significance, given the greatest mental health burden is in low- and middle-income nations, especially among those from conflict-affected regions (Schumann et al., 2019). Precision researchers have been slow to expand to global mental health, which aims to reduce mental health disparities across countries and address impacts of social structural issues on mental health of minoritized groups including refugees (Bemme & Kirmayer 2020). Research indicates that culture, ethnicity, and social context play a major role in mental health disorder etiology (Rasmussen et al., 2015); thus, the lack of focus on cultural differences in mental health treatment is a major oversight in precision medicine aims.

There are more types of effective interventions for common mental health disorders (CMDs) than ever before, yet clear guidelines for determining which treatment is optimal for an individual continue to elude researchers (Delgadillo & Lutz, 2020). Average treatment effects among a host of psychological, pharmacological, and neurological empirically supported treatments tend to be equivalent (Tolin, 2010; Norton & Barrerea, 2012). While an entire population may be likely to recover with a given evidence-based treatment (EBT), it is not currently known if a specific individual is more likely to benefit from Cognitive Behavior Therapy, a pharmacological treatment, or some other intervention (Carl et al., 2020; Mayo-Wilson et al., 2014). The implications of this lack of clarity are critical, because individuals for whom the first treatment is ineffective are less likely to seek out another form of treatment (Wang et al., 2005). Targeting treatments precisely to individuals could improve stagnant intervention rates for CMDs.

While there is preliminary evidence that precision, or personalized, psychotherapy may show effective results (Nye et al, 2023), and while there have been dozens of correlates for positive and negative treatment outcomes identified in studies such as Project MATCH (Kuhlemeier et al., 2021), research is largely unable to identify prospectively variables which predict outcomes a priori, rather than predicting outcomes post hoc (Delgadillo & Lutz, 2019; DeRubeis, 2019). Part of the reason for this difficulty is that treatment outcomes are likely driven by a cumulative effect of multiple small interactions, rather than one variable, and identifying small, cumulative factors is a laborious and time-consuming process (Kuhlemeier et al., 2021).

An additional problem with precision medicine in mental health is the molecular, neurobiological focus of much of this research (Deif & Salama, 2021). Biomarkers of mental health vulnerabilities have been unsuccessful at identifying findings that might guide clinicians in any functional way (DeRubeis, 2019) and environmental predictors have likely been overlooked (Peterson, 2020). Therefore, a sole focus on biomarkers rather than including social and structural determinants of mental health has hampered precision medicine research. This is counter to the trajectory of research demonstrating the relevance of the social environment in predicting mental health outcomes, research that has gained speed and empirical support, in spite of the lesser interest and funding in this area (Glasgow et al., 2018). Some precision medicine researchers have theorized that factors such as culture and culture-related traits such as religiosity, coping styles, attachment styles, beliefs about mental health, values, and culturally-rooted experiences may play a major role in precision interventions (Muela Arapico & Méndez, 2020; Norcross & Wampold, 2019). Identifying culturally responsive treatments could propel precision medicine forward, and provide intervention guidance for larger populations of understudied groups.

Precision Mental Health in Global Contexts

In terms of global mental health and efforts to reduce gaps in mental health resources worldwide, a focus on social determinants of health is warranted, as intervention effectiveness is historically understudied in non-WEIRD, low- and middle-income countries (LMICs; Schumann et al., 2019). Refugees with diverse backgrounds could provide useful information for comparing risk, correlates, and predictors of mental health intervention success in diverse groups, yet they represent a soberingly understudied population in precision medicine.

There are now more than 100 million forcibly displaced persons globally (UNHCR, 2023) most of whom have been forced from their homelands due to violent conflicts (Bemak & Chung, 2015). Refugees and other persons displaced due to conflict and violence exposure are at elevated risk for developing CMDs, especially depression, anxiety, and PTSD (Bogic et al., 2012). Most civilians in war-exposed regions will experience at least one potentially traumatic event as a result of the related conflicts (Fasfous et al., 2013), such as witnessing or experiencing violent attacks, bombings, torture, sexual violence, separation from family, and forced migration. Exposure to torture and a higher number of traumatic events experienced were correlated with greater risk for PTSD (Ibrahim & Hassan, 2017; Steel et al., 2009) and mood disorders such as depression and anxiety (Bogic et al., 2012; Giacco et al., 2018). These disorders tend to be chronic at disproportionately higher rates when compared to non-displaced persons.

In addition to experiences during violent conflict, prolonged exposure to adversity (social factors), such as detention, insecure residency status, low access to material services, inability to find work and other post-migration stressors, can also cause compounding mental strain (Li et al., 2015; Miller & Rasmussen, 2010). A combination of premigration, displacement, and postmigration stressors combine to make displaced persons one of the most vulnerable demographics for mental health burdens (Bemak & Chung, 2017), and anywhere from 39% to 100% of refugees who remain in low-to-middle income countries (LMICs) manifest PTSD, compared to 1–4% of the general population in WEIRD countries (Atwoli et al, 2014; Karunakara, 2004).

In addition to trauma exposure, the current environment not only plays a role in development of stress disorders, but also substantially moderates the relationship between potentially traumatic event (PTE) exposure and PTSD symptomology (LeMaster et al., 2018; Miller & Rasmussen, 2010; Porter & Haslam, 2005). Migration and post-migration stressors are more highly associated with mood disorders than traumatic exposure (Bogic et al., 2012). Environment-based factors that negatively moderate the relationship between a potentially traumatic event and poor mental health outcomes, include receiving documentation for resettlement (Raghavan et al., 2013) and perceived social support (Gorst-Unsworth & Goldenberg, 1998), suggesting that changes in one’s environment could affect mental health outcomes. These environmental stressors based around resource deprivation are predictive of physical and mental health (Almeida et al., 2002), and may play a major role in the etiology of stress-induced disorders like PTSD, depression, and anxiety.

Given the role resource access plays in mental health, a precision medicine approach ought to consider the role of current stressors in treatment approaches. Due to the role of stress in disorder development, some researchers argue that optimized treatments for trauma-exposed persons with PTSD and mood symptoms may include ameliorating chronic stressors rather than focusing primarily on cognitive processing of past traumatic experiences (LeMaster et al., 2018; Miller & Rasmussen, 2010). However, the strength of the relationship between resource access and mental health may vary cross-culturally, and therefore inform treatment approaches (Blackwell et al., 2023).

There are compelling reasons, theoretically and empirically, to suspect that culture may influence treatment effectiveness. Empirically, while some studies find configural, or pattern, invariance for measurement mental disorders such as PTSD, metric and scalar invariance are not found cross-culturally (Rasmussen et al., 2022; Rasmussen et al., 2015). Symptoms of PTSD often vary in salience cross-culturally, rather than topographical homogeneity (Grasser et al., 2021), such as varying factor structures or salience of different symptoms. If mental health symptoms vary cross-culturally, we might also expect that certain components of treatment, which target those symptoms, may produce disparate treatment outcomes. For example, if intrusive memories of acute traumas bear the weight of symptomatic burden for individuals of a certain cultural background, while hypervigilance is more prominent in another group due to prolonged chronic stress, a treatment focused on alleviating current stressors may improve symptoms in the latter, but not the former, despite similar overall severity.

In addition to empirical justification for variable mental health etiologies across cultures (Grasser et al., 2021), there are strong theoretical reasons for expecting different treatment outcomes based on varying experiences of violent, conflict, and displacement cross-culturally. Pre-migration and post-migration stressors are substantially different across refugee groups, which likely affect the etiology of mental disorders (Li et al., 2016). Different geopolitical contexts result in varying levels of acute trauma and chronic stressors, as well as time spent in flight. While some conflicts result in refugees from that region reporting high numbers of traumatic events, other conflicts may lead to fewer traumatic exposures on average, but longer periods of flight or living in refugee camps, resulting in prolonged states of stress (Blackwell et al., 2023). Additionally, post-resettlement, refugees from different regions often report different needs and sources of distress impacted by their cultural, religious, and previous educational and financial backgrounds, and racialization (Choe et al., 2022; Grasser et al., 2021). Refugees from middle-income nations often have some savings or higher education history which can buffer financial stress upon resettlement, however it can also result in a stark contrast in comparing their previous social status to their current one, prompting distress. In contrast, refugees from lower income nations often are resettled with no savings or safety net and experience financial hardship almost immediately; however, their situation may be improved compared to previous experiences—with available healthcare and other resources—leading to perceived improvements in social status (Hess et al., 2019; Thela et al., 2017). As disparate life experiences may result in different etiologies of mental distress, it seems likely that cultural background may also impact the types of mental health interventions that are most effective.

Present Study

The overarching goal of the current study was to examine whether certain individual difference variables were associated with differential outcomes in response to a community-based, resource-focused intervention; specifically, a reduction in mood disorder symptomatology. While risk factors have been identified which predict vulnerability to mental health symptoms in refugee groups, research has not yet examined how these factors may affect response to treatment. While traditional cognitive-behavioral and exposure therapies are effective at reducing mental health symptomology in refugee populations (Singla et al., 2020), it is unclear which different types of individuals respond better (or worse) to a treatment focused on ameliorating resource deprivation and strengthening social ties. Based on baseline analysis which found stronger correlations between traumatic events and mental health for Iraqis and Syrians and strong moderating effects of chronic stress on the relationship between traumatic events and mental health for Afghans and Great Lakes (Blackwell et al., 2023), we expected that Afghans and Great Lakes Africans would evidence significant effects of the intervention on culturally-specific distress, while Iraqis and Syrians would not, due to their higher levels of traumatic exposure and higher baseline levels of psychological distress, (Goodkind et al., 2020). Secondly, we predicted that, as an explanation for this discrepancy and consistent with cross-sectional research, resource access would play the strongest role in reducing mental health symptoms for Afghans and Africans, while prior traumatic events would play a larger role for Iraqis and Syrians. In addition to resource access as a potential active ingredient of intervention, we also selected several other variables which have been shown to impact mental health outcomes (trauma exposure) and were expected to change as a result of the intervention (social support and perceived English Proficiency) to examine potential mechanisms of change. In addition to expecting that these factors would change as a result of the intervention, previous research has shown that all predictors—resource access (Goodkind et al., 2014; Kasujja et al., 2022), trauma exposure (Steel et al., 2009), social support (Ahmad et al., 2020; Sundvall et al., 2021; Watcher et al., 2018), and, for resettled refugees, English proficiency (LeMaster et al., 2018; Sharifian et al., 2020)—are predictive of mental health symptoms for refugees from these regions and, at baseline, for the refugees in this sample (Blackwell et al., 2022; Blackwell et al., 2022; Goodkind et al., 2020).

Methods

Refugee Well-Being Project

Based on research that refugees often do not seek mental health services due to stigma and lack of familiarity, and that mental health intervention can be ineffective without addressing social and economic needs, a cost-effective, culturally adaptive intervention for reducing distress through community-based advocacy and mutual learning between refugees and student advocates was developed among resettled refugees in a large Southwestern city. In this intervention, the Refugee Well-Being Project (RWP), refugee families are paired with university student advocates who engage in mutual learning and cultural exchange and who aim to help newcomer families meet their self-reported needs and equip them with tools to navigate their new environments independently. For students, RWP participation is a two-semester course, with one semester focused on training and advocacy education, and the second focused on supervision and supporting advocacy efforts once students are paired with families for a six-month period. There were two components to the intervention: (1) Learning Circles, and (2) Advocacy. In Learning Circles, families and students meet weekly for 2–3 hours to share their cultures and experiences, engage in individualized learning, and plan weekly goals. For the advocacy component, advocates spend 4–6 hours weekly engaging in mutual learning and mobilizing resources to meet family needs. This might include helping families navigate health care systems, apply for jobs, practice English, or set up appointments with children’s teachers. To ensure fidelity of intervention implementation, extensive process data was collected, including weekly progress reports and logbooks from student advocates, observations of Learning Circles, and interviews with student-refugee dyads to explore their experiences working together.

Research Design

A randomized controlled trial was implemented to test the effectiveness the RWP intervention on a number of levels, including alleviating mental health symptoms experienced by newcomers. Participation was open to all refugees 18 and older from Afghanistan, the Great Lakes Region of Africa, and Iraq and Syria who had arrived in the U.S. within the past three years and were living in the city where the study took place. Recruitment was coordinated through bilingual/bicultural research team members from the three regions. These team members contacted all households on a complete list of refugees resettled in the given time frame, which was compiled in coordination with the two resettlement organizations in the city, as well as community networks. Each eligible adult in the household could make an independent decision about study participation; a household was only considered to have refused participation if no adults from the household agreed to participate. Of 161 households contacted, 143 (88.8%) agreed to participate in the study; 17 households were not interested in participating, and one household was ineligible because the one adult in the household was unavailable at the time the intervention would be held. Participants were excluded if they reported severe cognitive functioning difficulties, imminent suicide risk, severe mental illness, or participation in existing trauma-focused treatment. Only two participants were excluded due to these criteria and were referred for other treatment. Quantitative data from an interview battery of 524 questions were collected in participants’ homes at four time points: pre-, mid-intervention (three months), and post-intervention (six months) and then a follow-up at six months following the end of the intervention. Participants in the waitlist control group were given brief psychoeducation and referrals to mental health resources and were offered Narrative Exposure Therapy (NET) if they scored high on a measure of PTSD symptoms. Participants in the intervention condition who exhibited elevated levels of PTSD symptoms were also offered NET. Waitlist control participants were offered the opportunity to participate in the treatment immediately following the third cohort of implementation.

Approval for this study was granted from the Institutional Review Board at the University of New Mexico. Please see (Goodkind et al., 2020) for further information on overall study design.

Participants and Procedure

Participants were 290 individuals in 143 households who participated in a longitudinal RCT over 12 months. One-hundred and nineteen individuals (N=119) were part of the intervention group. Participants were all refugees resettled in the United States who had been living in the United States for less than three years at the start of the study and were compensated for their time in increasing amounts at each time point for their participation in the interviews ($20, $30, $40, $50). Their nations of origin included Afghanistan (35.5%), Iraq and Syria (32.8%), and the Great Lakes region of Africa (31.7%) —including Burundi, Democratic Republic of Congo, Republic of Congo, and Rwanda. With respect to gender, 52.4% of participants were women, and no participants identified as non-binary. Questions regarding racial identity were not included in the study, as many resettled refugees from these regions do not identify with the races assigned based on official U.S. classification and find these limited questions confusing. For example, Iraqis, Afghans, and others with Middle Eastern or North African ethnicity are classified as White in the U.S., despite many individuals feeling racialized as non-White due to their physical appearance and/or religious dress. All Great Lakes Africans would be racialized as Black in the United States. Further demographic information is provided in Table 1.

Table 1.

Demographic Information for the Three Geocultural Groups

	Afghan N (%)	Great Lakes African N (%)	Iraqi and Syrian N (%)
Gender  Male	52 (50.5%)	43 (46.2%)	46 (48.4%)
Female	51 (49.5%)	50 (53.8%)	49 (51.6%)
Marital Status  Single	39 (37.9%)	35 (37.7%)	22 (23.2%)
Married	50 (48.5%)	54 (58.1%)	66 (69.5%)
Widowed	12 (11.7%)	2 (2.1%)	5 (5.3%)
Divorced	2 (1.9%)	2 (2.1%)	2 (2.1%)
	Mean (SD)	Mean (SD)	Mean (SD)
Household Size	4.6 (2.0)	6.4 (2.3)	4.1 (1.6)
Weeks in US	23.5 (23.6)	26.84 (21.2)	39.9 (34.6)
Age	31.9 (10.6)	34.4 (12.2)	37.7 (11.1)
N	102	88	95

Quantitative data were collected in participants’ native languages (e.g., Arabic, Dari, French, Kiswahili, and Pashto) via personal interviews in their homes with a bilingual and bicultural interviewer who was a native speaker in the select language. These languages included: Arabic for Iraqi and Syrian participants, in their respective dialects; French, Kirundi/Kinyarwanda, and Kiswahili for Great Lakes African participants; and Dari and Pashto for Afghan participants. Using the TRAPD (Translation, Review, Adjudication, Pretesting, and Documentation) process, measures were translated and back-translated from English into each of the utilized languages (Survey Research Center, 2016). All data were collected through interview because of high illiteracy rates in the sample. All measures used in the study have a history of successful implementation with refugees and culturally diverse groups.

Measures

Mental Health Outcome

Newcomer Symptom Checklists.

Given criticism that traditional measures and factors of common mental disorders lack cross-cultural validity and applicability, three measures of culturally-specific, transdiagnostic emotional distress were created and validated as part of this study (Choe et al., 2023). The measures’ items were created using an inductive qualitative approach with a culturally diverse team to elicit expressions, understandings and idioms which reflect transdiagnostic idioms of distress. Exploratory factor analysis found high internal reliability for the Afghan, Great Lakes African, and Iraqi scales (α = 0.92, .91, and.91, respectively). Participants responded how frequently they experienced various symptoms on a five-point Likert-type scale where 0 reflected “never”, and 4 reflected “almost always.” Items were then averaged and the final score ranged from 0 to 4. Means and standard deviations were .76 (SD = .69) for Afghans, 1.95 (SD = .59) for Great Lakes Africans, and 1.35 (SD = .61) for Iraqis and Syrians.

Baseline Predictor

Trauma Exposure Scale.

The Checklist for Trauma Exposure was a baseline assessment formulated specifically for this study and included items from multiple validated measures including Weine’s trauma exposure questions (Horowitz et al., 1995), the Harvard Trauma Questionnaire (Mollica et al., 1992), and the Posttraumatic Stressors Scale (Foa et al., 1993). Participants were asked if they had personally experienced a potentially traumatic event such as “Being witness to a physical assault, beatings, or torture,” “Imprisonment,” and “Suicide of a family member or loved one.” Unless items specifically indicated witnessing an event, participants were only asked if they had personally experienced an event. There were 27 items, and the measure was scored from 0–27 with a higher number indicating greater exposure to traumatic events.

Time-Varying Co-Variates

Access to Resources.

The Difficulty Obtaining Resources scale (Sullivan & Bybee, 1999) functioned as a time-varying measure of Resource Access, which was reverse-scored to aid in ease of interpreting results. Participants rated it currently to obtain resources they needed in the 12 specific life domains, on a scale of 1 to 4. Access to resources was computed as the mean difficulty over all the resources respondents reported accessing, and then reverse scored such that higher results would reflect positive outcomes like the other covariates. For resources they were not accessing—such as childcare for participants without children—the item was not included in the mean. Average Cronbach’s alpha’s for these scales was .68.

Social Support.

The Multi-Sector Social Support Inventory Scale (MSSSI; Layne et al., 2009) measures social support from three sources over the past month: family, ethnic community, and non-ethnic community. As the intervention focused on connecting refugees with students from their non-ethnic community, only the non-ethnic community subscale was entered into the analysis as a time-varying covariate. Each of the three parallel scales included the same nine items related to attachment and perceived support (e.g., I can count on members of my family if I need help). Response choices ranged from 0 (never) to 4 (almost always), and total scores can range from 0 to 4. Higher scores reflect greater perceived social support. The MSSSI has been used or validated in all 3 geocultural groups in this study (Sierau, 2019; Soller et al., 2018). Cronbach’s alpha for non-ethnic community social support was .89 (M = 1.92, SD = 0.71).

English Proficiency.

Perceived English Proficiency (PEP; Rumbaut, 1989) was a time-varying covariate which measured how well participants understand, speak, read, and write English, and final scores were the mean of the four items. PEP is correlated with objective measures of English proficiency (Rumbaut, 1989). Response choices ranged from 0 (not at all) to 3 (like a native) and total scores are averaged and range from 0 to 3 (M = 1.08, SD = 0.66). Higher scores reflect greater perceived English proficiency. Cronbach’s alpha was .92.

Data Analytic Approach

A two-level multilevel model was used in the analysis with time points nested within participants. While participants were nested within households in this study, a three-level model would overparameterize the model for the subgroup analysis, so household level was not included in the model. Power was determined using Optimal Design software for power estimation (Spybrook, et al., 2011) to detect small-to-medium differences in post-intervention intercept or slope of change over time (d = 0.25) at 2-tailed p < .05, using 3-level multilevel modeling (4 time points at level 1, individuals within households at level 2, and households at level 3). As this study sought to look at predictors of recovery and change within each geocultural group, only two levels were supported by the model to detect similar effects with sufficient power (>.80).

Analysis of these data involved two steps. First, using the glmmTMB package in R (Brooks et al., 2017), growth curve analysis was used to examine the strength of predictors on distress symptoms, as well as the interaction between time and intervention group. Second, the models were examined for each of the three geocultural subgroups to determine if different predictors played a role in outcomes among those groups.

Each model contained four random effects, allowing (a) intercept or level of outcome to vary among adults, and (b) time slopes to vary among adults. A negative binomial model was used in the analysis to reflect the shape and dispersion of the outcome variables, which were zero-inflated, and over-dispersed, violating the assumptions of a traditional count distribution.

The equation for the analysis was: Level 1: Time; ${\text{Outcome}}_{\text{tij}}={\text{ß}}_{0\text{ij}}+{\text{ß}}_{1\text{ij}}(\text{Time})+{\text{ß}}_{2\text{ij}}(\text{resource access})+{\text{B}}_{3\text{ij}}\text{(perceived English proficiency)}+{\text{ß}}_{4\text{ij}}{\text{(social support) B}}_{5\text{ij}}(\text{time}\ast \text{intervention}){+}_{\text{etij}}$. Level 2: Individual ${\text{ß}}_{0\text{ij}}={\text{ß}}_{00\text{j}}+{\text{B}}_{01\text{j}}{\text{(number of traumatic events)+r}}_{0\text{ij}}.{\text{ß}}_{1\text{ij}}={\text{ß}}_{10\text{j}}+{\text{r}}_{1\text{ij}}$. Outcome is cultural distress, time is the different participant responses over each time point, number of events is the traumatic event exposure at baseline, and time*intervention is the slope coefficient of change for the intervention group.

Including missed interviews and skipped items, 2.0% of the data matrices included data missing at random (Little’s MCAR v2 (df = 67378) = 37630.71, p = 1.00). Maximum likelihood estimation was used to estimate parameters after listwise deletion of missing data.

Data Availability

Due to the highly vulnerable nature of this population, neither the data nor the materials have been made available on a permanent third-party archive to avoid potential risk to participants; however, requests for the data, materials, or study analysis code may be made to Dr. Jessica Goodkind at the University of New Mexico.

Results

Table 2 presents a summary of findings for all three geocultural groups. Results from these analyses revealed variable findings with respect to intervention effects and significant predictors of cultural distress.

Table 2.

Treatment and Predictor Effects on Cultural Distress Among the Three Geocultural Groups

	Afghan	Great Lakes African	Iraqi and Syrian
	Estimate (SE)	Estimate (SE)	Estimate (SE)
Fixed Effects  (Intercept)	1.885(.453) ***	1.573 (.269) ***	1.76(.355) ***
Independent Variable  Traumatic Events	.092(.024) ***	.021(008) **	.043(.008) ***
Social Support	−.193(.085)*	−.100(.048)*	−.034(.038)
Resource Access	−.158(.151)	−.483(.082)***	−.263(.055)***
English Proficiency	−.453(.158)**	−.068(.109)	−.125(.061)*
Time*Intervention	−.259 (.110)*	−.116(.057)*	.026 (.032)
−2 Residual	204.8	−846.1	−1024.2
AIC	−383.6	1716.1	2074.2
BIC	−339.4	1757.1	2121.2
N	102	88	95

Note:

^*p < .05

^**p < .01

^***p <.001

Analyses of Demographic Group Differences

There were no significant group differences in marital status or gender. Great Lakes Africans had significantly larger household sizes (B = 2.02 SE = 0.25, p < .01), or around 2 more persons per household, on average, than the other two groups. Iraqis and Syrians were significantly older (B = 4.63 SE = 1.4, p < .01), or by an average of 4.6 years, and had been in the United States for significantly longer than the other two groups (B = 14.85 SE = 3.37, p < .001), or by about 15 weeks. Additionally, as discussed in the hypotheses, Iraqis and Syrians reported higher levels of traumatic exposure (B = 1.12, SE = 0.81, p < .05) and higher baseline levels of psychological distress (B = 0.44, SE = 0.06, p < .001).

Afghans

There was a significant interaction of the intervention and time on culturally specific distress among Afghan participants (B = −0.259, SE = 0.108, p = .013), indicating a decrease in emotional distress scores of .259 at each time point, and an effect size of d = 2.39. However, the significance and strength of the effect size did not substantially change with the addition of predictors, indicating that intervention effects cannot be explained by changes in resource access, social support, or English proficiency, nor can intervention effects be explained by previous trauma exposure. Unexpectedly, the relationship between resource access and emotional distress across time was non-significant (p = .29), however, traumatic exposure (B = 0.092, SE = 0.023, p < .01), social support (B = −0.179, SE = 0.086, p = .037), and perceived English proficiency (B = −0.453, SE = 0.157, p < .01) were related to outcomes, such that higher traumatic exposure was predictive of more severe distress symptoms, while higher social support and higher perceived English proficiency were predictive of lower symptoms. See Table 2 for a summary of findings.

Great Lakes Africans

There was a significant interaction of the intervention and time on culturally specific distress among Great Lakes African participants (B = −0.116, SE = 0.057, p = .042), indicating a mean decrease in emotional distress scores of .116 at each time point and an effect size of d = 2.03. However, as with the Afghan group, the significance and strength of the effect size did not substantially change with the addition of predictors, indicating the intervention effects cannot be explained by changes in predictors. As predicted, the relationship between resource access and emotional distress was the strongest predictor of mental health, with increased resource access predicting lower distress symptoms (B = −0.483, SE = 0.082, p < .001). Additionally, traumatic exposure (B = 0.021, SE = 0.007, p < .01) and social support (B = −0.100, SE = 0.048, p = .040) were related to mental health outcomes, with higher traumatic exposure predicting more severe distress symptoms and higher social support predicting lower distress symptoms. Perceived English proficiency, however, had no significant relationship to psychological distress (p = .53).

Iraqis and Syrians

As expected, there was no significant intervention effect on mental health outcomes over time among Iraqi and Syrian participants (p = .418). While trauma exposure was significantly related to mental health as expected, this effect was quite small (B = 0.043, SE = 0.008, p < .001), contrary to expectations that higher levels of trauma exposure and emotional distress would drive treatment response. Also unexpectedly, resource access had the largest effect on mental health outcomes for this group, such that greater access to resources reduced severity of mental health symptoms (B = −0.263, SE = 0.055, p < .001). While social support did not impact mental health outcomes, higher perceived English proficiency was related to better mental health (B = −0.125, SE = 0.062, p = .043).

Discussion

Findings from this study move researchers closer towards determining what effective interventions look like for non-WEIRD groups historically neglected by outcomes research, but leave open the question of why these treatments work for some groups, but not others. Our hypotheses around the trajectory of mental health symptoms over time for each geocultural group operated as expected, with resettled Afghans seeing substantial mental health improvements of an advocacy-based intervention over time, Great Lakes Africans experiencing significant, but small, improvements of the intervention over time, and Iraqis and Syrians experiencing no mental health improvements from the intervention. However, our hypotheses around what factors might explain these differences were not clarified in these analyses. Though we expected findings might offer clarity for what active ingredients of the intervention might account for these changes, they actually suggest that unmeasured and unhypothesized intervention components likely account for these disparate outcomes. While resource access was not a significant predictor of mental health for Afghans, and improvements in English language ability and improved social support were, these factors did not impact the effect of the intervention over time. For Africans, greater resource access and to a lesser extent greater social support from their non-ethnic community played a significant role in their mental health symptoms reduction, but again, in spite of significant intervention effects, they did not predict the intervention effect over time, suggesting that, while the intervention was effective at reducing mental health symptoms, it was effective for unknown reasons.

These findings are both surprising and informative, yet leave substantial questions that might be explored in future work. The intervention proposed and implemented was specifically designed with the theory that by improving access to resources, social support, and English proficiency, mental health would improve. Support for our hypotheses would add depth to the debate over the superiority of psychosocial versus trauma-focused interventions, by suggesting that the intersection of shared geocultural background and resettlement experience may make resource-focused aid of primary importance to some groups, while others may see benefit from psychological-focused interventions (Miller & Rasmussen, 2010; Haushofer et al., 2020; Singla et al., 2020). However, while all three of the hypothesized predictors were important to mental health, and while the intervention was effective at decreasing mental health symptoms, the intervention was not effective as a function of the hypothesized mechanisms of action. Future research might explore other potential mechanisms by directly comparing advocacy-based approaches to cognitive and behaviorally-based interventions. For instance, potential factors to explore in future work might include household size, age, length of time since resettlement, and other between-group differences that may impact treatment outcome. Great Lakes African’s larger household sizes, or Iraqis and Syrian’s higher age (4.6 years) or time since resettlement (15 weeks) may account for difference in effectiveness of intervention approaches.

While these findings fail to explain why some groups saw mental health benefits from the intervention, while others did not, findings that mental health outcomes for this intervention varied by geocultural origin still have important implications that might be used to inform precision mental health for refugee groups. First, our findings suggest that resettled Afghan and Great Lakes African refugees experiencing mental health distress symptoms may benefit from an advocacy-based intervention. However, while resettled Iraqi and Syrian refugees may experience other benefits from an advocacy-based intervention, and despite the strong association between resource access and mental health, this particular model may not provide the reduction in distress symptoms other groups reported across time.

Second, it is important to note that number of traumatic events had a significant effect on culturally-specific emotional distress, but, unexpectedly, for all three groups this effect was quite small. This was especially surprising for Iraqis and Syrians, who we predicted would show a robust association between trauma exposure and mental health symptoms due to their higher levels of reported traumatic exposure. However, like Africans, resource access was the largest predictor of mental health for Iraqis and Syrians. Thus, we conclude that the effect of the intervention did not overlap with the effect of resource access for Iraqis and Syrians to the expected degree, and our original hypothesis that poorer outcomes for Iraqis and Syrians might be explained by the higher salience of prior traumatic events was not supported.

There could be several reasons for this finding. One potential explanation is that Iraqis and Syrians tended to have higher education levels and more economic stability upon arrival to the United States than the other groups studied (Harjanto & Batalova, 2022). It could be that the resources needed were slower to emerge and that Iraqis and Syrians needed more sophisticated help (such as renewing professional licenses) than advocates could provide swiftly. Another explanation could be that the contrast between their life pre- and post-displacement was starker than the other two groups such that even with improved resource access, their economic situation was still far worse than they experienced in the past. Results suggest that Iraqis and Syrians may be in greater need of both advocacy and traditional mental health support systems than the other groups examined in this study.

Limitations and Future Directions

There are several limitations of this study worth noting. First, while this sample size was large for studies with refugee populations, it is still statistically small for identifying predictors of longitudinal outcomes. Past precision medicine research has hypothesized that the reason specific predictors of treatment outcome are difficult to identify a priori is that there are likely dozens of factors which independently have small impacts, but when considered together they result in meaningful individual differences (Kuhlemeier et al., 2021). There could be many cultural or contextual factors which we did not think to account for which might predict effectiveness of intervention outcomes. Displaced persons are under-resourced and under-researched, while also being especially vulnerable to violence and its mental health consequences. To better answer the question of what types of treatments are most effective for individuals from diverse cultures, it is crucial that funders invest in studies that can recruit larger sample sizes so various interventions and predictors can be compared rigorously. Larger sample sizes would also allow for a deeper examination of the effect of additional predictors and interaction effects of mental health outcomes. It is possible that factors such as age, gender, religious background, or marital status affect outcomes (Arfken et al., 2018; Perera et al., 2016). While the demographics of our sample were fairly consistent, there were between-group differences in some areas (age, family size, and time in US), so it is difficult to conclude with certainty if the differences we found are based on cultural differences or on demographic factors. Second, although intervention fidelity was assessed through weekly progress reports and logbooks from student advocates, observations of Learning Circles, and interviews with student-refugee dyads to explore their experiences working together, we did not calculate intervention fidelity ratings. Third, it is important to note that while persons from the same region often share many cultural similarities, there is also substantial heterogeneity within geocultural groups, influenced by factors such as class, religion, linguistic history, ethnicity, and region. While understanding how cultural background may impact treatment response is an important area of research, it should not be taken to mean that an intervention is prescriptively appropriate or inappropriate for every individual from that region, rather that, on average, one treatment may be more appropriate than another for a particular presenting problem. The potential for “broken leg” exceptions, or factors which render actuarial predictive methods invalid, should not be ignored even when general findings can help effectively guide clinical decisions (Meehl, 1957). Future research should seek to account for other potential demographic predictors of treatment response, such as time since resettlement, age, family size, and more. Fourth, in a similar vein, while these data are longitudinal, a longer follow-up period would provide data on the long-term effectiveness of this advocacy intervention. Funding for costlier, long-term follow-ups can detect delayed intervention effects, and bolster claims of long-term impact (Forman et al., 2012).

It is also worth mentioning that this study is limited by its singular focus on mental health outcomes. While important, there are other important factors, such as quality of life and health, which these analyses did not include—and for some of those outcomes Iraqis and Syrians did see improvements (Goodkind et al., 2020). A singular focus on mental health should not restrict providers and aid workers from attention to the benefits of material aid. However, what these findings suggest is that effective treatments for emotional distress may vary by culture. While Iraqis and Syrians may receive many other kinds of benefits from an advocacy-based intervention, different types of interventions in conjunction may be necessary to target mental health. However, for displaced individuals from Afghanistan and the Great Lakes region of Africa, findings not only suggested that an advocacy-based intervention is effective for treating culturally-specific symptoms of emotional distress, but also offered potential mechanisms of action that may inform a precision mental health approach: indicating that improved English proficiency and social support for Afghans and increased access to resources and social support for Great Lakes Africans may be the factors producing these successes. Identifying these active ingredients is critical for developing and disseminating effective interventions for some of the world’s most vulnerable groups (Blackwell et al., 2022). Should future research among refugees resettled in the United States replicate these findings, these differences may be able to provide precision medicine guidance for treatment types and targets among diverse violence- and conflict-exposed groups.

Public Health Significance Statement:

An advocacy-based intervention reduced symptoms of distress for Afghan and Great Lakes African refugees, but not Iraqi and Syrian refugees, highlighting varying mental health needs. Further, different predictors of mental health outcomes were important to each group, suggesting a heterogeneous path to symptom remission. These findings may be used to help expand precision medicine to diverse populations.

Masked Narrative Description.

The data reported in this manuscript were collected as part of a larger, longitudinal data collection. Findings from the data collection have been reported in separate manuscripts. MS1 (published) reports the main effect of the intervention studied on outcomes, but does not examine culturally specific measures of distress, or seek to identify specific mechanisms of change; MS2 (published) and MS3 (published) explores the trauma of family separation specifically; MS4 (published) looks at correlations of mental health symptoms cross-sectionally; MS5 (published) explores specifically economic stressors on the entire sample; and MS6 (published) explores sources of social support in this sample. MS 7 (submitted) will focus on exploratory factor analysis for the culturally specific measures used in this study. However, this will be the first and only manuscript to look at longitudinal changes in culturally specific measures of distress for our three groups, and attempt to identify the potential active ingredients of change in mental health over time.