Study design and setting
This concurrent mixed-methods study utilizes data collected from married youth in Jordan between 2022 and 2023 as part of the ongoing Gender and Adolescence: Global Evidence (GAGE) longitudinal study (see Jones et al. [18] for more details). GAGE is a 10-year mixed-methods study following approximately 20,000 adolescent girls and boys in Jordan, Ethiopia, Bangladesh, Lebanon, Palestine, Rwanda, and Nepal. Informed by the emerging evidence base on adolescent well-being and development, GAGE draws on a capabilities conceptual framework that explores the interconnectedness of young people’s capabilities, the change strategies underpinning programming interventions, and the community and broader policy contexts in which they are living (see Fig. 1) [19]. In the case of SRH, the focus is on access to information, services and support that enable young people to maximize their health and well-being at a pivotal time in the life course and reap the ‘triple dividend’, with immediate impacts on adolescents today, longer-term impacts on adolescents as they transition to adulthood, and intergenerational impacts for those adolescents’ future children [20].
Fig. 1
GAGE conceptual framework
In Jordan, GAGE is following a cohort of more than 4,000 adolescents living in five governorates– Amman, Mafraq, Irbid, Zarqa, and Jerash– who were initially enrolled at baseline in 2018 to 2019. At baseline, the sampling frames were based on household vulnerability and, for Jordanians, aimed to be representative of socioeconomic vulnerable Jordanian youth, with an emphasis on understanding the refugee dynamic and marginalized groups. Therefore, the sample includes Syrian refugee youth living in either formal UNHCR camps (Zaatari and Azraq), ITSs, or host communities, and includes vulnerable Jordanian youth living in host communities in the same locations. The baseline sample included a random sample of Syrian refugee and vulnerable Jordanian adolescents aged 10–12 years and 15–17 years, as well as a purposeful sample of marginalized adolescents, including adolescents with a disability and adolescents married before the age of 18 years. The majority of the qualitative sample was drawn from the larger quantitative sample. For more details on the baseline sampling strategy, see Jones et al. [18]. At follow-up in 2022 and 2023, 2,940 youth were surveyed (a 71% follow-up rate).
Study population
For this analysis, the quantitative analysis focuses on the cross-sectional survey data collected at follow-up; baseline data is only used for the measurement of assets to minimize the amount of missing data. The study population for this analysis is married female Syrian refugee youth and vulnerable married female Jordanian youth, defined as those aged 15–24 years at the time of the 2022–2023 survey administration. To minimize repetition, we often refer to these simply as ‘Jordanians’ or ‘Syrian refugees’ in the text. There were 41 Jordanians and 272 Syrian refugees, most (56%) of whom live in host communities. All respondents were surveyed in person between July 2022 and May 2023 by a trained local team of enumerators who spoke fluent Arabic. The follow-up questionnaire [21] was developed based on previous rounds of data collection and tested in both English and Arabic, simultaneously. All questions were self-reported and collected via computer-assisted personal interviewing in Arabic.
The qualitative analysis used data from 117 in-depth interviews with Syrian youth, 12 focus group discussions with their caregivers, and 24 key informant interviews with community leaders and service providers. Among the in-depth interview participants, 30 were married and 23 of those were married before their eighteenth birthday. This sample is detailed and used in other recent work drawing on the GAGE sample [22]. These interviews were carried out between July and October 2022 by a trained team of qualitative researchers who spoke fluent Arabic. The qualitative interviews were recorded and then transcribed by native Arabic speakers into English.
Quantitative measures
The main independent variables of interest in this study are refugee status and residence. Eligible participants self-reported their nationality as Jordanian or Syrian. All Jordanians in this sample live in host communities. For Syrian refugees, three residence groups were created based on whether they reported living in host communities, formal UNHCR-run camps, or ITSs. Due to Syrian refugees migrating within Jordan between the baseline (2018–2019) and follow-up (2022–2023) survey rounds, residence is defined through a combination of baseline and follow-up survey locations, to form three distinct groups: always living in host communities (n = 152), ever living in formal camps (n = 86), and ever living in an ITS (n = 34). Although the sample size for Syrians living in ITSs is relatively small, this group is fundamentally different than those in other locations [23, 24], such that it is not contextually appropriate to combine.
The two primary outcomes of interest in this study are any recent contraceptive utilization, and type of contraceptive method recently used. Recent contraceptive utilization is operationalized via a binary indicator for those who self-report currently or recently using any form of contraception. We cannot determine the timeframe for ‘recent’ contraceptive use from the survey data; that is up to the respondents’ interpretation. We created a binary indicator for any recent modern contraceptive use, defined as currently or recently using oral contraceptive pills, intrauterine devices (IUDs), implants, injectables, condoms, emergency contraception (EC), or lactational amenorrhea (LAM). We also generated binary indicators for recent use of each type of contraceptive method (oral contraceptive pills, IUDs, withdrawal method, implants, injectables, condoms, EC or LAM). All contraceptive use outcomes were collected via self-report.
Existing literature suggests that age, educational attainment, wealth, governorate, residing in an urban setting, and number of living children are associated with contraceptive use among married women in Jordan [8,9,10, 16]. As such, the following factors are included in this analysis: age at marriage, difference in age with spouse, cohabitation with spouse, educational attainment, number of living children, and household wealth. Household wealth is measured via a decile index created from a set of 20 baseline household assets generated using principal component analysis [25].
Quantitative analysis methods
We first generated and explored descriptive statistics for demographic characteristics and recent contraceptive use among married vulnerable Jordanian and Syrian refugee female youth in Jordan. For continuous variables, means and standard deviations were reported. For binary variables, frequencies and proportions were reported. Statistical differences from formal comparisons between Jordanians and Syrian refugees, as well as Syrians in refugee camps, host communities and ITSs, are reported through t-tests for continuous variables and Chi-squared tests of independence for binary variables.
Next, we explored factors associated with any recent contraceptive use among married Syrian refugee female youth in host communities or refugee camps using a multivariable linear probability model with the following specifications:
$$\:{y}_{i}=\alpha\:+{\beta\:}_{1}{Hostres}_{i}+{X}_{i}^{{\prime\:}}\gamma\:+{u}_{i},$$
(1)
where \(\:{y}_{i}\) is the outcome of interest (any recent contraceptive use); \(\:{Hostres}_{i}\) is an indicator for residing in a host community; and \(\:{X}_{i}^{{\prime\:}}\) is a vector of predictive factors listed above. Syrian refugees living in ITSs and Jordanians were not included in this linear probability model due to sample size limitations. A linear probability model is used instead of logistic regression for ease of interpretation by readers (note that findings are qualitatively the same if logistic regression is used [26,27,28], see Supplemental Table 1 for logistic regression results).
All quantitative analysis was done in Stata version 18.
Qualitative analysis methods
The interviews were transcribed into English and then coded thematically in the qualitative analytical software MAXQDA, using a thematic codebook that was informed by detailed oral debriefings with the national research team. This debriefing process was carried out after interviews in each locality and then followed by a longer debriefing, during which the team discussed similarities and differences across research sites (both those that were explicitly stated by respondents as well as those that were more implicit and observed by the team). The coding was carried out in English by a team of three coders who had been working with the Jordanian longitudinal dataset previously. The coding process combined a mixed deductive and inductive approach, starting with a pre-defined set of codes based on the GAGE capabilities conceptual framework [29]. The coders had weekly debriefing meetings to discuss any emerging issues with relation to the application of the codes, inductively identified a number of additional thematic codes that were context-specific. The theoretical approach that was used in the coding process could best be categorized as reflexive thematic analysis [30]. A sub-sample of transcripts were double coded to check for consistency between coders. Unless otherwise stated, all quotes are from different participants.