Structural determinants of depressive symptoms among refugees and host communities in South Sudan

Evidence from explainable machine learning

Bibliographic Data

ID22066494
AuthorsHyojin Im (0000-0001-9577-6714, Virginia Commonwealth University, corresponding author), Ashwag Abdulrahim (0000-0003-0226-0036, Virginia Commonwealth University)
Year2026
Volume14
Pages1778986-1778986
Publication date2026-04-14
Open AccessYes
TypeARTICLE
VenueFrontiers in Public Health (JOURNAL)
Journal identifiersISSN: 2296-2565 • E-ISSN: 2296-2565
PublisherFrontiers Media SA (PUBLISHER • CH)
DOI10.3389/fpubh.2026.1778986
PMID42058092
OpenAlexW7154280437
LanguageEN
References cited36

Background Depressive symptoms in displacement settings are often framed as consequences of refugee status, which can obscure the shared structural conditions emphasized in the Social Determinants of Health (SDoH) framework that shape mental health for refugees and host communities and limit the identification of practical intervention targets. This study examines the relative contribution of health, socioeconomic, protection, and contextual factors to depressive symptom severity among adults living in displacement-affected settings in South Sudan. Methods We analyzed nationally representative data from 3,055 adults (2,066 refugees, mean displacement duration 11.15 years; 989 host community members) from the 2023 Forced Displacement Survey. Depressive symptom severity was measured using the PHQ-9. We compared Elastic Net regression, Random Forests, and Extreme Gradient Boosting (XGBoost) using 10 × 5 nested cross-validation. The best-performing model was interpreted using SHapley Additive exPlanations (SHAP) to estimate the marginal contribution of each predictor in PHQ units. Results Mean depressive symptom severity was low to moderate overall ( M = 4.43, SD = 5.00) and did not differ by population type ( p = 0.783). XGBoost showed the highest predictive performance [Root Mean Squared Error (RMSE) = 4.47, R 2 = 0.247], significantly outperforming Elastic Net regression ( p = 0.006). Model explanations identified self-rated health status as the dominant predictor (19.3% of total importance), followed by perceived community violence (11.1%), perceived poverty (9.6%). Age (9.3%), discrimination (9.2%), food insecurity (8.6%), and citizenship (8.2%, pooled model only) contributed at moderate levels, whereas social protection (3.4%) and remittances (0.7%) contributed minimally. Predictor profiles were broadly similar across refugee and host models, with differences primarily in magnitude rather than rank ordering. Conclusion Depressive symptoms in South Sudan appear to be structured primarily by health, material hardship, and protection-related gradients rather than refugee status per se . Findings support integrated, area-based public health responses that link mental health support with primary health care access, poverty-oriented assistance, and protection and safety interventions rather than programming organized primarily around legal status distinctions

Food insecurity · Mental health · Population · Poverty · Psychological intervention · Refugee · COVID-19 and Mental Health · Education and experiences of immigrants and refugees · Migration, Health and Trauma

Tools

Open DOIOpen Access