This study is based on the transactional model of stress (18). This theoretical model explains the relationship between stressful events and psychological outcomes. The structure of the transactional model imposes the influence that personal and situational factors have on the appraisal of a condition (i.e., primary and secondary appraisal) (18). The primary appraisal of a threat of loss or harm depends on a person’s antecedent conditions related to the threatening stimulus and the individual’s psychological characteristics. Lazarus and Folkman (18) label the outcomes of the primary appraisal as harm/loss, threat, and challenge. Hence, migration, being a refugee, and the delivery of a newborn may threaten the ability of a mother to enjoy or fulfill her expected roles and therefore influence her mental status (19). In addition, situational variables can affect cognitive appraisal; a more severe perception of the event would lead to a more intense appraisal and, in turn, a more stressful perception. In secondary appraisal, the person evaluates whether she/he has the competencies and resources, such as social support, to deal with stressors and restore equilibrium with the environment (18). In the transactional model, adaptational outcomes are considered as the product of the situational and personal variables. For instance, social support and resilience improve the interactions between personal and situational variables (i.e., postmigration stress) and have negative relationships with the adaptational outcome (i.e., postpartum depression).

This study used a cross-sectional design. Convenience sampling was used to recruit Syrian postpartum refugee women in Istanbul, where 532,000 Syrian refugees live (20), through referrals from the nongovernmental organization (NGO), Refugee and Asylum Seekers Assistance and Solidarity Association (RASASA). Refugee women who were aged 18 years and older and had a live birth within the past year were eligible to participate. Exclusion criteria included self-reported diagnosis of a major mental disorder (such as schizophrenia or psychoses) and current use of antipsychotic or antidepressant medications at the time of the study. Data were collected between August 2022 and February 2023.

The sample size was determined before participants were recruited. A minimum of 200 mothers was required for the path analysis in structural equation modeling, with a medium effect size of 0.30, a power of 0.80, and a significance level of 0.05 (21).

This study was conducted per the ethical principles of the Declaration of Helsinki. Ethical approval was obtained from Koç University Committee on Human Research (Approval #: 2022.047. IRB3.019, Approval Date: 1/27/2022). Participants were informed about the study’s purpose and voluntary participation, including their right to decline or withdraw at any time without any influence on their care or services. Anonymity was assured, and participants were informed that their data would be used solely for research purposes. Eligible women who agreed to participate in the study provided recorded verbal consent. In addition, all participants were given information about available psychosocial services, including those offered by NGOs.

Following the referral of refugee women from the NGO to the primary researcher, a trained Arabic-speaking research assistant called each woman to explain the purpose of the study and review the eligibility criteria. Those eligible and who agreed to participate were asked to provide verbal consent to be contacted for a subsequent telephone interview and to schedule a convenient time to complete the survey over the phone. All participants were sent appointment reminders via text message before the interview. We approached 208 eligible mothers, of whom 8 refused to participate (response rate = 96.2%). Therefore, the final sample size comprised 200 Syrian refugee women.

At the beginning of the telephone interview, the research assistant asked the mother if she was in a private place and if she felt comfortable starting the interview without interruption. During the interview, the mothers’ responses were entered into Qualtrics XM (Qualtrics, Seattle, WA), a secure, password-protected, web-based software. The questionnaires took approximately 20–30 min to complete. After completing the questionnaires and after each interview, mothers were provided an electronic voucher ($10).

Descriptive statistics (i.e., means, standard deviation, and frequencies) were used to describe the sample sociodemographic and clinical characteristics and the key study variables (e.g., postmigration stress, intimate partner violence, and PPD). The normality of the data was checked using the Kolmogorov–Smirnov test. The dependent variable, EPDS, was scored as a categorical variable in the study: depression symptoms (EPDS ≥ 10) vs. no depression symptoms (EPDS < 9), as it did not meet the normality assumption. We first conducted bivariate analyses to examine the association between the outcome variable (PPD) and the participants’ characteristics. Chi-square or Mann–Whitney U tests were used to examine differences in postpartum depression symptoms by participants’ characteristics. Spearman rank correlations were used to examine the associations among continuous variables. Next, logistic regression was conducted to examine the psychological predictors of PPD, controlling for significant covariates identified in the bivariate analyses. The results of the logistic regression then provided preliminary evidence of relationships among the study variables, which informed the selection of variables included in the exploratory path model analysis (i.e., postmigration stress, social support, resilience, and PPD). While this study initially aimed to examine the indirect pathways between IPV and PPD, IPV was excluded from the path analysis due to a low report rate and a non-significant association with PPD in logistic regression.

The indirect effect was examined using the weighted least squares mean and variance adjusted (WLSMV) estimator, which is appropriate for categorical dependent variables. We used a top-down approach by including covariates based on bivariate analysis in the path model, removing the nonsignificant covariates from the path model one at a time while assessing the improvement in model fit (37). The bootstrapping technique with bias-corrected confidence interval estimates was conducted to examine significant indirect pathways, i.e., social support and resilience. The hypothesized model was evaluated for goodness of fit using the following measures: Chi-square goodness of statistic (p > 0.05), Comparative Fit Index (CFI ≥ 0.90), Tucker-Lewis Index (TLI ≥ 0.90), Root Mean Square Approximation (RMSEA ≤ 0.08), and standardized root mean square residual (SRMR < 0.08). The standardized coefficients were reported. The analyses were conducted using IBM SPSS version 28 and Mplus version 8.0 (38). Alpha was set a priori at 0.05 for all statistical tests.

Participants’ sociodemographic and clinical characteristics are summarized in Table 1. More than half of the participants were aged 27 years or older (51.0%), unemployed (99.0%), had less than a high school education (81.5%), and reported insufficient income (88.0%). Most were married (99.5%), multiparous (94.5%), and gave birth vaginally (66.0%), with 34.5% having experienced high-risk pregnancies. Only 3% of mothers reported intimate partner violence, while 25% had unmet mental health treatment needs. PPD symptoms were significantly differentiated by parity, high-risk pregnancy, history of depression, and unmet mental health treatment needs for PPD.

PPD was associated with all main variables (Table 2). Post-migration stress was associated with resilience (r = −0.229, p < 0.001) and social support (r = −0.193, p < 0.001). However, social support was not associated with resilience (r = 0.137, p = 0.052).

Regression analysis revealed that, after controlling for other variables in the model, the odds of PPD increased significantly by 33% for each one-unit increase in the anxiety score (Adjusted Odds Ratio [AOR]: 1.33; 95% CI: 1.17–1.53) (Table 3). Conversely, for each one-unit increase in the resilience score, the odds of PPD decreased significantly by 8% (AOR: 0.92; 95% CI: 0.867–0.979).

Figure 1 displays the path diagram for the standardized probit coefficients. The model adjusting for parity fit the data well, X² (2, N = 200) = 2.798, p = 0.247; CFI = 0.99; TLI = 0.94; RMSEA = 0.045[90% CI: 0.000–0.155]; SRMR = 0.056. In the model, there was a direct effect from postmigration stress to resilience (β = −0.266, p < 0.01) and social support (β = −0.199, p < 0.01). In addition, there was a significant direct effect from resilience (β = −0.589, p < 0.001) to PPD. Path analysis indicated that social support (β = 0.046, 95% CI: 0.012–0.108) and resilience (β = 0.163, 95% CI: 0.075–0.25) indirectly affect the relationship between postmigration stress and PPD (Table 4).

This study has some limitations. First, we employed convenience sampling to recruit Syrian refugee women via referrals from a single NGO, RASASA, in Istanbul. Although this method was selected to provide direct access to refugee women, recruitment was limited to those connected to RASASA and who voluntarily agreed to participate. Therefore, the sample may not represent the broader population of Syrian refugee women. Second, there may be other variables influencing PPD that were not examined or included in the model. Third, a cross-sectional research design was used in the current study, which may preclude causal or temporal inferences Thus, longitudinal research is needed to further test the path model explored in the current study. Fourth, while collecting data via telephone-based interviews is a reliable and accessible method (45), underreporting is still a concern due to the social desirability effect, as refugee women might feel hesitant to disclose issues related to their mental health, such as posttraumatic stress and intimate partner violence. Hence, further studies are required to focus on these critical issues, considering the limited information available on Syrian refugee women in Türkiye. Finally, whereas the path model had strong fit indices, there is potential overfitting due to the exploratory nature of the analysis and the sample characteristics. Therefore, to validate the indirect pathways, confirmatory designs and representative samples are recommended in future studies.

This study revealed that anxiety and resilience are important predictors of PPD in Syrian refugee women. Our findings also suggest indirect associations between postmigration stress and PPD through resilience and social support. Hence, clinical and policy initiatives are recommended to address mental health in refugee women through screening programs integrated in perinatal health care in both hospital and community-based settings. In addition, interventions that promote social support and resilience in Syrian refugee women may be beneficial in improving their PPD symptoms. Specifically, understanding the needs of refugee women and the intersection of their culture, gender roles, and migration policies is crucial to planning and implementing prevention interventions that enhance coping and resilience capacity. This is crucial as building social support and connections among refugee women should also be advocated (19). Finally, while our study is cross-sectional and exploratory, future community-based participatory research that involves collaboration between researchers and community stakeholders (46) serving refugee women is needed to better understand refugee women’s mental health and address their needs through action steps.