Design and setting
We analyzed data from Women ASPIRE, a cross-sectional study of gendered physical and mental health concerns of Syrian refugee women living in non-camp settings in Jordan (N = 507). The key objective of this study was to understand which health concerns are most prevalent for Syrian refugee women who were engaged in health care systems, as well as modifiable risk factors to reduce negative health outcomes in order to help inform best practices in clinical, and policy settings to improve the lives of Syrian refugee women in Jordan. Additional details on the Women ASPIRE study have been published elsewhere [19, 23]. We limited our analyses to women who had a history of pregnancy and whose last pregnancy (which would have taken place in the year prior to data collection, 2017) was single, took place in Jordan, and ended in term live birth or miscarriage (N = 307). The study was set in four nongovernmental health clinics that primarily served women and children, one in each of four governorates in Jordan (al-Mafraq, Amman, az-Zarqā’, and Irbid), which hosted the majority of all registered Syrian refugees in Jordan . Research assistants with a degree in medicine, psychology, public health, or sociology approached women in the clinic waiting rooms, introduced the study, obtained informed consent, and enrolled eligible women into the study. Research staff then conducted a one-hour, face-to-face, interviewer-administered quantitative survey using Qualtrics software on electronic tablets. Upon completion of the survey, each participant received a food package (valued at 7 USD). The research assistants completed training in human subjects research, referral pathways, and survey administration.
We recruited eligible women by time- and venue-based systematic sampling between April and November 2018. We included women who were Syrian refugees (confirmed by UNHCR identification card), were ≥ 18 years old, lived in non-camp settings in Jordan, were visiting the clinic for their own health, and agreed to participate. We excluded women with any degree of cognitive dysfunction as determined by the Mini–Mental State Examination . Of all women invited to participate, 84% met the eligibility criteria and agreed to participate (N = 507).
For the present analysis, we aimed to identify factors associated with increased risk of miscarriage among our sample. We selected a series of demographic, physical and mental illnesses, and gendered inequities of health that have been previously found to be associated with miscarriage in comparable samples.
We measured age (in years) on a continuous scale, and defined advanced maternal age as age ≥ 35 years at conception . We determined literacy based on the standard definition (ability to read and write). We asked participants whether they were married and, if applicable, the age at first marriage. We then dichotomized the age at first marriage as ≥ 18 years or < 18 years to construct the child marriage variable.
We used the reduced Coping Strategies Index (CSI) to measure household food insecurity (Cronbach’s ɑ = 0.67). This measure, which has been validated for use with displaced Arab populations, and was selected to characterize the economic stability of the household (which has been shown to be associated with poor reproductive health outcomes, globally), rather than using a measure such as household income, given that most refugee families received similar financial assistance from various aid organizations. The reduced CSI is a standard set of five coping strategies that indicate household food insecurity . We assessed whether participants had resorted to the five coping strategies within the past 30 days using an individual “Yes” or “No” scale for each strategy. Then, we summed the number of “Yes” responses (0–5) and considered two or more “Yes” responses to indicate food insecurity.
We captured a simplified residential history using two items from the questionnaire: the governorate of permanent residence in Syria (Aleppo, al-Ḥaskah, al-Lādhiqīyah, al-Qunayṭrah, ar-Raqah, as-Suwaydā’, Damascus, Dar ‘ā, Dayr al-Zūr, Ḥamāh, Ḥimṣ, Idlib, Reef Demashq, or Ṭarṭūs) and the governorate of current residence in Jordan (‘Ajlūn, Amman, al-’Aqabah, al-Balqā’, al-Karak, al-Mafraq, aṭ-Ṭafīlah, az-Zarqā’, Irbid, Jarash, M’ān, Mādabā). We collapsed the categories of both variables based on geographical clustering.
Domestic physical violence
We measured the lifetime exposure to domestic physical violence, perpetrated by the current or most recent husband (Cronbach’s ɑ = 0.75) or a family member (Cronbach’s ɑ = 0.68), using a modified version of the Conflict Tactics Scale, which has been validated in a sample of Palestinians [27, 28]. We asked participants to report, using a “Yes” or “No” scale, whether they had experienced eight forms of physical abuse. Women who responded “Yes” to any one of the eight items were considered to have been exposed to physical abuse and those who responded “No” to all eight items were considered not to have been exposed.
Physical health characteristics
We asked participants whether they had ever been diagnosed with thyroid disease or diabetes mellitus with binary response options of either “Yes” or “No.” We also asked participants who had been pregnant at least once before about the outcome, order, and prenatal care for up to five pregnancies, in order of recency. To measure the number of previous miscarriages, we enumerated the number of pregnancies based on data we collected on the most recent five pregnancies experienced. We collected data on the five most recent pregnancies because we expected that most women in our sample would have come to Jordan from Syria within the past 5 years. Out of up to the four pregnancies preceding the last pregnancy—that resulted in miscarriage, and retained the variable as a continuous variable.
Mental health characteristics
We measured mental health using three validated scales—namely: the abridged Center for Epidemiologic Studies Depression (CES–D) four-item scale, Generalized Anxiety Disorder seven-item (GAD–7) scale, and Posttraumatic Stress Disorder Checklist for DSM–5 (PCL–5). CES–D measures symptoms of depression based on seven-day recall, GAD–7 measures symptoms of generalized anxiety disorder based on two-week recall, and PCL–5 measures symptoms of posttraumatic stress disorder based on 30-day recall [29,30,31]. We used the standard cut scores (≥ 4, ≥ 10, and ≥ 23, respectively) to interpret the screening results on a dichotomous scale (“negative” or “positive”). Cronbach’s ɑ for CES–D, GAD–7, and PCL–5 were 0.78, 0.85, and 0.94, respectively.
The primary dependent variable of this analysis was whether the most recent pregnancy ended in term live birth or miscarriage. We created a dichotomous variable based on the question “What was the outcome for this pregnancy?” We defined term live birth as the delivery of a neonate with signs of life after 37 weeks of gestation and miscarriage as spontaneous pregnancy loss before 20 weeks of gestation.
We computed summary statistics to describe the background characteristics and access to health services of the full sample (N = 307 women). We then performed a subgroup analysis of women of advanced maternal age to describe the rates of unplanned pregnancies and contraception use. Next, we estimated the relationships between independent variables and the most recent pregnancy outcome (term live birth or miscarriage) for the full sample using Pearson’s χ2 test or the Mann–Whitney U test. Finally, we fit a binary logistic regression model (with Firth’s correction for small-sample bias) to obtain adjusted estimates . We included in the model the full set of independent variables, namely “literacy”, “household food insecurity”, “region of permanent residence in Syria”, “region of current residence in Jordan”, “lifetime diagnosis of diabetes mellitus”, “lifetime diagnosis of thyroid disease”, “awareness of a reproductive health service in the area of residence”, “advanced maternal age at conception”, “prenatal care”, “number of previous miscarriages”, “child marriage”, “lifetime physical abuse by the current or most recent husband”, “lifetime physical abuse by a family member”, “CES–D result”, “GAD–7 result”, and “PCL–5 result”. We excluded eight incomplete cases from the multivariable analysis and we used the Benjamini–Hochberg method to correct for multiple testing. We used R (version 3.6.2) to perform all data analyses.
The Columbia University Morningside Institutional Review Board and the Jordanian Ministry of Health approved the study protocol. All participants provided written informed consent.