Migrants' mental health recovery in Italian reception facilities

The study investigates how initial reception in host-country facilities affects asylum-seekers’ mental health and whether specific factors predict PTSD. Motivated by evidence that migrants—especially asylum-seekers and refugees—experience elevated rates of mental disorders (notably PTSD), and that both pre- and post-migration stressors influence outcomes, the authors seek to fill gaps in systematic mental health screening during reception. Research questions: RQ1—What is the impact of first contact with the reception system on asylum-seekers' mental health? RQ2—Can we identify risk (e.g., traumatic experiences) and shielding (e.g., family ties, education) factors for PTSD? Hypotheses: H1—A 14-day stay in reception facilities improves psychological well-being; H2-1—Asylum-seekers with PTSD have different lifetime experiences than those without; H2-2—Past traumatic experiences and demographics can predict PTSD. The study emphasizes the relevance of early reception conditions for mitigating trauma and guiding policy and clinical practice.

Prior work documents high prevalence of mental disorders among migrants, particularly asylum-seekers and refugees, including schizophrenia, anxiety, depression, and especially PTSD. Pre-migration exposure to war, persecution, and torture, and post-migration adversities (legal insecurity, poor housing, limited integration) contribute to risk. Despite this, few studies longitudinally track mental health during stays in reception centers and systematic psychological screening is rare. The literature suggests that reception quality may shape outcomes and that identifying socio-demographic and experiential predictors of PTSD could inform prevention and care. Machine learning has been increasingly used for mental health detection but has been less applied to migrant PTSD predictors. This study builds on these gaps by quantifying short-term changes during initial reception and by modeling predictors from trauma histories and demographics.

Setting and participants: The study was conducted in reception facilities for asylum-seekers in Rome, Italy (municipally coordinated; medical assessments by the Migrant Health Unit, Local Health Authority Rome 2), from May to September 2021. Asylum-seekers underwent a mandatory 14-day COVID-19 quarantine. Of 150 admitted, 100 participants (6% female; mean age 30 ± 7.2 years) consented and completed all questionnaires administered by psychiatrists/psychologists with trained cultural mediators. Ethics: Informed consent obtained via same-language mediators; participation voluntary; approved by Comitato etico Lazio 2, ASL Roma 2 (No. 166.20; protocol 0029199). Data and code publicly available. Cultural mediation: Full-time, certified mediators matched by cultural/linguistic background provided culturally contextualized support throughout the 14 days. Instruments: - Q0 Demographics (sex, age, marital status, education, employment, supports, ties, time in Italy). - Q1 and Q6 WHO-5 Well-Being Index at entry and day 14. - Q2 PC-PTSD-5 screen. - Q3 Harvard Trauma Questionnaire (HTQ): traumatic events (Section C), DSM-related and culture-related PTSD symptoms (Section D). - Q4 TALS/SCI-TALS: 116 binary items across 9 domains on trauma/loss spectrum, subthreshold symptoms, temperamental traits. - Q5 LiMEs (Italian version): 59-item checklist of traumatic experiences and living difficulties before, during, and after migration. Statistical analysis: For RQ1 (H1), compared WHO-5 positivity at entry (Q1) vs day 14 (Q6) using McNemar exact test (p=0.05); power >0.999 with N=100. For RQ2 (H2-1), compared TALS domain scores between HTQ-positive (n1=23) vs HTQ-negative (n2=77) using left-tailed Mann–Whitney U-tests (p=0.05). Machine learning for RQ2 (H2-2): Data preprocessing included cleaning, standardization, and aggregation/generalization of categorical responses. Feature selection via univariate chi-square tests (p≤0.05); MRMR and ReliefF tested but not adopted due to poorer performance/sensitivity. Classification: Response variables were binarized PTSD symptoms from PC-PTSD-5 (Q2) and HTQ Section D (Q3). Predictors included HTQ Section C traumatic events and selected demographic items from Q0 (questions 2–5, 7–10, 12, 15–30). Models tested: SVMs, decision trees, ensembles (e.g., RUSBoost), and k-NN. Addressed class imbalance by inverse frequency class weights. Validation via 10-fold cross-validation repeated 20 times. Performance metrics: sensitivity (primary), specificity, fall-out (FPR), accuracy, and AUC. Model selection required accuracy >70%, AUC >0.70, and fall-out <0.25, prioritizing sensitivity to minimize missed PTSD cases.

• RQ1 (impact of reception stay): WHO-5 screening positivity dropped from 51% at arrival (Q1) to 21% at day 14 (Q6); McNemar exact test p<0.001. Item-level WHO-5 responses showed broad improvements (e.g., “cheerful and in good spirit” >half-time from 70% to 85%; “calm and relaxed” 76% to 87%; “fresh and rested” 72% to 83%; “daily life filled with interesting things” 64% to 74%; “calm and vigorous” 83% to 82% essentially unchanged). Among 8 participants positive to Q1 and PTSD tests (Q2/Q3), 5 improved to WHO-5 negative at day 14. - PTSD incidence: HTQ indicated 23% PTSD positivity; PC-PTSD-5 indicated 22% positivity with 83% agreement between the two instruments. - Traumatic exposures (HTQ Section C) were common: material deprivation 70%, war-like conditions 69%, witnessed violence 61%, disappearance/death/injury of loved ones 51%, bodily injury 50%, forced confinement/coercion 50%, torture 52%, forced to harm others 12%, brain injury 8%. - H2-1 (lifetime experiences differ by PTSD status): Across all nine TALS domains, HTQ-PTSD–positive participants had higher median scores than negatives with significant differences: Domain I median 0.5 vs 0.3 (U=356.5, p<0.001); II 0.59 vs 0.22 (U=376.5, p<0.001); III 0.48 vs 0.29 (U=462, p<0.001); IV 0.72 vs 0.28 (U=307, p<0.001); V 0.78 vs 0.22 (U=298, p<0.001); VI 0.58 vs 0.25 (U=367.5, p<0.001); VII 0.25 vs 0 (U=480, p<0.001); VIII 1 vs 0 (U=211, p<0.001); IX 0.17 vs 0 (U=555, p=0.002). - H2-2 (predictors of PTSD symptoms): Demographic variables were not selected as salient predictors and did not improve models. Traumatic events predicted specific HTQ symptoms: • Recurrent intrusive thoughts/memories—Quadratic SVM using war-like conditions, torture, forced to harm others, disappearance/death/injury of loved ones, witnessed violence (Accuracy 0.85; Sensitivity 0.86; Fall-out 0.16; AUC 0.84). • Feeling jumpy/easily startled—RUSBoosted Trees using all traumatic events (Accuracy 0.80; Sensitivity 0.82; Fall-out 0.19; AUC 0.84). • Sudden emotional/physical reactions when reminded—Medium Tree with modified cost function using all traumatic events (Accuracy 0.76; Sensitivity 0.72; Fall-out 0.21; AUC 0.75). Other PTSD symptoms could not be robustly predicted under selection criteria, potentially due to sample size or unmeasured factors.

The study demonstrates that a supportive, structured initial reception (over 14 days) can significantly improve asylum-seekers’ subjective well-being, addressing RQ1 and supporting the role of post-migration environment in recovery. Concurrently, elevated PTSD prevalence reflects the substantial pre-migration trauma burden. H2-1 is supported: PTSD-positive individuals exhibit higher lifetime trauma/loss spectrum features across all TALS domains, underscoring clinical relevance of subthreshold and spectrum presentations and the need to distinguish them from depressive symptoms. For H2-2, machine learning indicates past traumatic events—not demographics—predict particular PTSD symptoms (intrusive thoughts, hyperarousal, cue reactivity), while demographic factors like social ties or education did not emerge as shielding predictors in this cohort. These findings align with theory that pre-migration trauma drives risk, while early, trauma-informed post-migration care can mitigate distress and support adaptation. The results argue for timely mental health assessments, culturally informed care (including mediators), and reception models that avoid large, isolating facilities. Tailored interventions should particularly address sequelae of bodily injury, torture, and witnessed violence. Failure to address post-migration stressors risks undermining recovery.

This study provides quantitative evidence that immediate, supportive reception conditions improve asylum-seekers’ psychological well-being over a short period and that pre-migration traumatic exposures are key predictors of PTSD symptoms, while demographics did not act as shielding predictors in this sample. Policy and practice should prioritize trauma-informed, person-centered early reception, comprehensive mental health screening, and targeted support for those exposed to severe violence and deprivation. Future research should: (1) replicate findings in diverse migrant populations and settings; (2) extend follow-up to assess durability of improvements; (3) examine timing and access to reception systems; (4) evaluate the specific impact of cultural mediators and co-designed assessment tools; and (5) explore additional predictors of less predictable PTSD symptoms.

• Sample size of 100 limits power to detect predictors for several PTSD symptoms; models for many symptoms did not meet performance thresholds. - Short observation window (14 days) captures immediate effects but not medium- to long-term trajectories. - Response variables were binarized to meet model criteria, potentially reducing nuance. - The cohort was predominantly male (6% female) and located in a single metropolitan reception context (Rome), limiting generalizability. - Many participants had already lived in Italy ≥1 year (94%), so findings may not reflect newly arrived asylum-seekers. - Potential unmeasured confounding: other determinants of symptoms may not have been captured by administered questionnaires. - Use of Western diagnostic tools, although mitigated by cultural mediators, may still imperfectly capture culturally specific expressions of distress.