The data source of the study is TDHS-SM-2018, which was conducted by the Hacettepe University Institute of Population Studies.

The TDHS-SM-2018 is designed to produce estimates of important demographic characteristics and health indicators for Syrians living in Turkey. The survey aims to produce the first ever household level mother and child health indicators of the Syrian refugee population in Turkey.

A weighted, multi-stage, stratified cluster sampling approach was used in the survey. The framework on the Syrian population in Turkey was obtained from the Ministry of Interior, Presidency of Migration Management. Because of the unavailability of a frame that included each Syrian household in Turkey, a sample was created based on the population size of each quarter, which is Turkey’s smallest administrative unit. Stratification in the survey considered only one variable for the camp / non-camp population. The first stage of sample selection comprised selecting quarters from each stratum as primary sampling units. Following the creation of block lists by field teams that identified Syrian homes in the chosen quarters, the second stage of sample selection — the selection of households — was conducted. In order to interview 2000 households (target sample size) in 100 clusters for the TDHS-SM-2018 sample, 20 households from each cluster were chosen. All women between the ages of 15 and 49 who usually resided in the chosen households and / or were present there the night before the interview were considered to be eligible for the women’s questionnaire.

The fieldwork was conducted between 23 November 2018 to 12 February 2019. Data collection for the TDHS-SM-2018 sample was carried out in teams. Each team consisted of six people who are bilingual in Turkish and Arabic: three female interviewers, a male measurer, a female listing personnel and a team supervisor. Before the interviews were conducted, the listing personnel scanned the households in the determined cluster and listed the households where Syrian refugees live. Then, one of the female interviewees interviewed women from the households which were selected from the listed households. Since measuring instruments are heavy, men are preferred as the measurer and the male measurer obtained the height and weight of women aged 15–49.

In the survey, the technique of computer-assisted personal interviewing in which a computer or a tablet is used were applied. CSPro software (U.S. Census Bureau) was employed for the data processing.

In the TDHS-SM-2018, 2216 women aged 15–49 living in 1826 households were successfully interviewed on issues related to child health for all children under five years, together with socio-economic and bio-demographic characteristics of the Syrian refugees living in Turkey. Hacettepe University Ethics Commission reviewed and approved the TDHS-SM-2018 and informed consent was obtained from all interviewed women in the original study.

In this further analysis, the last children with a gestational duration greater than 32 weeks from the mothers’ singleton pregnancy, born in the two years before the research date and living with the mother were enrolled (unweighted n = 744).

The dependent variable of our study was “breastfeeding in the last 24 hours during the study period”.

As independent variables, the sociodemographic characteristics of family members [the place of residence in Turkey (non-camp and camp), region (central, east, west and south), highest education level of mother and father (no education, primary, secondary, higher), mothers’ literacy in Arabic (cannot read at all, able to read only parts of the sentences, able to read the whole sentence), mothers’ reading and writing abilities in Turkish, household size (< 6, 6–9 and ≥ 10 members), the presence of other wives of husband, smoking status of the household members], the economic status [existence of a person in the household working in a paid job, and existence of a person in the household receiving any payment of benefits], the maternal characteristics [maternal age at the first birth, and the current pregnancy status at the interview, mother’s body mass index (BMI) values, life satisfaction, satisfaction with financial status and social activities], the antenatal, birth and postnatal characteristics of enrolled children [mother’s desire for child when pregnant (soon: wanted at the time of conception, later: after became pregnant; does not want), mothers’ age at birth of the current child, the birth order, the preceding birth interval, the number of antenatal care attendances, the gestational duration of the child, the country where they gave birth (Turkey and Syria), the place of delivery (home, public hospital and private hospital), the type of delivery (Cesarean and vaginal delivery), child’s birth weight, sex, and age, starting time of breastfeeding (early initiation of breastfeeding (EIBF) and after), prelacteal feeding (giving the baby any fluid or nutrients, including sugared water, honey, tea, milk other than breast milk, infant formula, plain water, except breast milk in the first three days after birth), the presence of bottle feeding, and the presence of postnatal care in the two months after the birth were taken from TNSA-SM data.

In addition, the duration since migration to Turkey coded as “23 months before the birth of the child or after the birth”, “2– 3 years before the birth of the child”, “4–5 years before the birth of the child” and “at least 6 years before the birth of the child”.

In the study, a ten-point scale of satisfaction with the financial situation of the household was used to evaluate subjective economic well-being. Similarly, mother’s satisfaction with life is measured again using a ten-point scale: 1–10. One point denoted the worst case, 10 the best case. This variable was recoded into two categories ≥ 5 for satisfied, < 5 for dissatisfied. The social activities of the mothers were evaluated according to their exercise habits, visiting friends at home, using internet and watching television programs for women. Exercise habits information was collected with the question, “Do you do any physical activity such as sports, walking?“.

In Demographic Health Surveys (DHS), as in most surveys, the sample is selected with unequal probability to increase the number of cases available (and hence decrease sample variability) for specific locations or subgroups for which data are required. Weights are used when statistics are tabulated in order to provide the appropriate representation, and corrections for varying response rates are made when weights are calculated because of sample design [12]. Unweighted and weighted numbers of cases and percentage distributions of the data were calculated. Breastfeeding percentages according to case characteristics were calculated using a weighted database.

Using complex sample logistic regression, the difference between breastfeeding percentages according to household characteristics, maternal activities and antenatal and postnatal characteristics of last-born child were examined, and univariate odds ratios (OR) and 95% confidence interval (CI) were calculated.

The difference between the percentages of breastfeeding in the weighted sample for variables including household characteristics, maternal activities and antenatal and postnatal characteristics of the last-born child was analyzed with the Chi-square test. Variables with a difference of p < 0.10 were determined and included for multivariable binary logistic regression analysis. Among these variables, it was observed that there was a relationship between private hospital deliveries, frequency of antenatal follow-up, and the financial status of the household, and “delivery place” was further analyzed. There was a relationship between prelacteal feeding and bottle usage, and prelacteal feeding was selected for further analysis. Therefore, multivariable binary complex sample logistic regression analysis evaluated region (South vs. West, Central vs. West, East vs. West), respondent currently working (yes vs. no), currently pregnant (yes vs. no or unsure), place of delivery (home vs. public sector, private sector vs. public sector), history of prelacteal feeding (yes vs. no), preceding birth interval (first child vs. ≥ 36 months, < 18 months vs. ≥ 36 months, 18–35 vs. ≥ 36 months), duration between migration and birth of the child (2–3 vs. < 2 years., 4–5 vs. < 2, ≥ 6 vs. < 2 years) and child’s age in months as independent variables on breastfeeding.

Value of median breastfeeding duration was determined by linear interpolation of percentage of first group (beginning with the birth group) below 0.5 and previous group percentage using the following formula: median = ??−1 + [(??−1 − 0.5) / (?? −1−??)] ∗ (??) where ?? is the proportion breastfeeding for the first group where the proportion is below 0.5, pi-1 is the proportion breastfeeding for the preceding group, mi-1 is the midpoint value for the preceding group, and wi (two months) is the time width of the group taken as the difference between the midpoint value of the current group and the preceding group [12].

Analyses were performed with Stata 13 and SPSS 22 package program.