Complex Samples Logistic Regression Analysis of Predictors of the Use of Modern Contraceptives Among Married or In-Union Women in Ethiopia.

Background: Modern contraceptive use by currently married Ethiopian women has steadily increased over the last 15 years. The main objective of this study was to identify the magnitude and predictors of current use of modern contraceptives among married or in-union women aged (15-49) in Ethiopia using complex sampling. Methods: The study used data from the 2016 Ethiopia Demographic and Health Survey (EDHS, 2016) dataset. In the study a total number of 7,346 married or in-union women age 15-49 in Ethiopia were included. The current use of modern contraceptives among reproductive-age married or in-union women in Ethiopia, measured as ‘Yes’ (currently using a modern method) and ‘No’(using a folkloric method, traditional method, and no method) is considered as a dependent variable. Complex samples binary logistic regression was performed to examine the importance of each explanatory variables using SPSS version 23, and statistical signicance was attached at p=0.05. Results: The study found that age 40–44 years (AOR = 2.064), wealth index Middle (AOR = 1.553); Rich (AOR =1.71), working status yes (AOR = 1.279), and sexual activity in the last 4 weeks active in the last 4 weeks (AOR=3.614) were highly signicantly associated with current use of modern contraceptive. Conclusion: The variables age, religion, residence, region, wealth status, husband desire for children, women's working status, sexual activity in the last 4 weeks, and total number of children ever born were statistically signicant predictors for the current use of modern contraceptives among married or in union women in Ethiopia.

born were highly signi cantly predicted the use of modern contraceptives among married or in union women in Ethiopia.

Background
Contraception de ned as a practice of methods of preventing or planning of conception. It is the most cost-effective investments a country can make in its future. Moreover, it offers a range of potential bene ts that encompass economic development, maternal and child health, education, and women's empowerment. Contraception allows spacing of pregnancies, delaying pregnancies in young girls who are at increased risk of health problems from early childbearing, and preventing pregnancies among older women who also face increased risks. Contraception enables women who wish to limit the size of their families to do so. By reducing rates of unintended pregnancies, contraception also reduces the need for unsafe abortion. Some contraceptive methods help prevent the transmission of HIV and other sexually transmitted infections [1]. Contraceptive methods can be modern, traditional and folkloric methods of contraception. Modern method of contraception includes, intra uterine device (IUD), oral contraceptives, Pill, Progestogen-only pills (POPs) or "the minipill", Injectables, condom, Male sterilization (vasectomy) Female sterilization (tubal ligation) Emergency contraception pills, Lactational amenorrhea method (LAM), Sympto-thermal Method, etc. Contraceptive use is widely acknowledged as an important intervention towards achieving Millennium Development Goals (MDGs) four and ve as it has proven to improve maternal health which calls for reduction of maternal and child mortality ratio [2][3][4].
Modern contraceptive methods enable couples to enjoy sex without fear of the risk of pregnancy at any desired time [5]. It also helps couples and individuals realize their basic right to decide freely and responsibly when and how many children to have. The growing use of contraceptive methods has resulted in not only improvements in health-related outcomes such as reduced maternal mortality and infant mortality, but also improvements in schooling and economic outcomes, especially for girls and women [6]. The bene ts of modern contraception have long been established. It is estimated that up to 35% of maternal deaths and 13% of child mortalities could be averted whilst 25% of under-ve mortalities could be prevented if birth intervals were at least three years and by the use of various contraceptive methods in planning their families [7].
Contraceptive prevalence and the unmet need for family planning are key indicators for measuring improvements in access to reproductive health as asserted in the 2030 Agenda for Sustainable Development under target 3.7. "By 2030, ensure universal access to sexual and reproductive health-care services, including for family planning, information and education, and the integration of reproductive health into national strategies and programs" [8]. Globally, 62 percent of married women ages 15 to 49 use a method of family planning and 56 percent use a modern method. At a country level, use of any method of family planning among married women can vary signi cantly, ranging from 4 percent in South Sudan to 88 percent in Norway [9]. Contraceptive use has increased in many parts of the world, especially in Asia and Latin America, but continues to be low in sub-Saharan Africa [10]. The use of modern family planning has expanded rapidly across Ethiopia, rising from about 8 per cent of married women in 2000 to 36 per cent in 2016, according to the country's demographic and health survey [11].
Previous studies have considered the determinants of contraceptive use in Ethiopia [12,13]. In most of the previous studies, complex sampling design is not considered and since this data is from multistage sampling, which involves the use of different strata, clustered sampling techniques and unequal selection probabilities, it is inappropriate to conduct analysis without taking the survey sampling designs into account. Also, tting such data without considering the survey sampling design may lead to biased estimates of parameters and incorrect variance estimate [14]. Thus, the current study aims to see prevalence of current use of modern contraceptives and examining the predictors of current use of modern contraceptives among reproductive-age women in Ethiopia based on the 2016 EDHS data using complex sample logistic regression.

Data Source
Data for this secondary data analysis was extracted from the 2016 Ethiopian Demographic and Health Survey (EDHS) dataset to identify Factors of the current use of modern contraceptive among married and in-union women in Ethiopia. The author accessed the data by requesting it from the Demographic Health Survey Program Team through online (http:// dhsprogram.com/). The survey was conducted from January 18, 2016, to June 27, 2016, based on a nationally representative sample that provides estimates at the national and regional levels and for urban and rural areas. The survey target groups were women age 15-49 and men age 15-59 in randomly selected households across Ethiopia [11].

Sample Design
The 2016 EDHS data is strati ed and selected in two stages. In the rst stage, 645 EAs, 202 urban areas and 443 rural areas were selected using probability proportional to the size of enumeration area. Lists of households were prepared from the selected enumeration areas and served as a sampling frame for the selection of representative households in the second stage. Lastly, 18,008 households were selected for the study out of which only 17, 067 households were occupied. Complete interview was obtained from 16,650 households making the response rate 98%. In those interviewed households, 16,583 women were identi ed and 15,683 women completed the interview making a response rate of 95%. Therefore, sample data for the current study came from an individual record of 15,683 reproductive-age women. Details of the study design for these survey can be obtained from CSA, ICF and Rockville [11].

Study sample
This study restricted the analysis to women who were currently in union. Consequently, women who were currently not in any ardent union, who never had sex, who were currently pregnant, and were infecund were excluded from the dataset. The study sample selected was 6,989. We accounted for the multistage sampling design during analysis, which resulted in weighting the dataset using a weight variable in the dataset for reproductive-age women. Thus, the weighted study sample used for analysis was 7,346 reproductive-aged married or in-union women in Ethiopia [ gure 1]. We used ETIR71SV.sav ,SPSS data le from 2016.

Study Variables
The dependent variable in this study was 'current use of modern contraceptives' among reproductive-age married or in-union women in Ethiopia. In this study, the responses to the contraceptive use question 'no method' or 'modern method' or 'traditional' or 'folkloric method' were dichotomized and measured as 'Yes' (currently using a modern method) and 'No' (using a folkloric method, traditional method, and no method).
To predict current use of modern contraceptive methods among married or in union women's, several independent variables were selected based on literatures and their relevance [12,24]. These variables included; age in 5-year group of respondents, region, respondent and husband/partner educational level, place of residence, religion, wealth index, respondent current working status, age difference, media exposure, husband desire for children, sexual activity and total children ever born. Age difference was calculated by subtracting the husband/partner age with respondent age and grouped as 1= same or older than man and 2= man older than women.

Data Analysis
The study variable was coded as binary variable (Yes= 1 and No= 0) and the data analysis was performed using SPSS version 23. We account for the multistage sample design in the DHS dataset during the descriptive analyses, test of association, and logistic regression analyses. We selected a subpopulation (women who were currently in-union/living with a man); we used the "Currently/formerly/never in union" variable to specify the subpopulation during the analysis. The DHS calculate the individual sample weights to six decimal places but it is presented in the dataset without points. Thus, the sample weight for each case was computed by dividing the available weight in the dataset by 1,000,000 before using it in the analyses. This was done according to the guideline of DHS on handling the individual weight variable [11] .
The data analysis part includes rst descriptive statistics then chi-square test and nally binary logistic regression analysis. The descriptive statistical technique was used at the univariate level to describe the characteristics of the sample using frequency and percentage distribution for categorical variables. Bivariate analysis is used Pearson's Chi-square (χ2) test to examine the association between the dependent variable and the independent variable. In logistic regression, factors that showed unadjusted odds ratio (OR) with a P-value <0.25 were subjected to multiple logistic regression analysis. Adjusted odds ratio and con dence interval (CI) were respectively, used to measure the association between current use of modern contraceptive and predictor variables and their statistical signi cance in the nal model.

Results
Descriptive Statistics: Socio Demographic Characteristics A total of 7,346 sample of married or in-union women aged 15-49 were considered in this study from 2016 EDHS data set. The prevalence of use of modern contraceptives was 3,095 (42.1%). Six thousand one hundred eighty ve (84.2%) of the respondents were from rural areas and the remaining 1,161 (15.8%) were from the urban areas. Majority of the respondents had no education (61.0%), were sexually active in the last 4 weeks (81%), were found in the age group of 25-29 years (25.3%), were not currently working(69.8%), were not exposed to media(72.1%), were from poor family (40%) and were Orthodox Christians (41.1%). Chi-square Test of Association Table 1 above shows the results of the magnitude of prevalence of current use of modern contraceptives and background characteristics of the respondents. Of the study samples, most of the women (63.4%) residing urban place used modern contraceptive compared to about 38.2% of rural women. That is among a woman who lives in rural areas majority of them (61.8%) were not currently used modern contraceptive methods. Among women who participated in this study, the highest prevalence of modern contraceptive was from Addis Ababa (67%) followed by Amhara (52.9%) and SNNPR (48.2%) regions whereas the lowest prevalence for the use of modern contraception were from Somali (1.9%) region. Married or in union women, who are currently working showed high prevalence (50.6%) of use of modern contraceptives, while prevalence of the use of modern contraceptives for women not working is 39.5%. Among highly educated women, there was 63.9% prevalence of current use of modern contraceptive compared to 37% among women with no formal education. The chi-square test shows that, predictors age of respondents, region, place of residence, religion, women's educational level, wealth index, husband's desire for more children, total children ever born, working status of women, husband education level, age difference, sexual activity, and media exposure about family planning were statistically signi cant at 5% level of signi cance (Table 1).

Binary Logistic Regressions
The results of the binary logistic regression are given in Table 2, which displays the crude and adjusted odds ratio, signi cance level (p-value) and con dence interval for the coe cients associated with each of the independent variable entered into the logistic regression model. In the complex samples multiple logistic regression model, factors age, Region, residence, religion, wealth index, husband desire for children, women working status, sexual activity, and total children ever born were signi cantly associated with current use of modern contraceptives (Table 2).
Women within the age group of 40-44 years were 2.064 times more likely to use modern contraceptive methods (AOR = 2.064, CI: (1.461, 2.916)) than the reference age group 15-19 years. The odds of using modern contraceptive of married or in-union women in Amhara are 1.587 times (AOR: 1.587 CI: (1.270, 1.982)) higher than that of the odds of using modern contraceptive of married women in Tigray region which is the reference region. On the other hand, Married women who lived in Somali and Afar were 94.9% and 63.6% less likely to use modern contraceptive methods as compared to married women in Tigray respectively. While the odds of married women living in Oromia, Benishangul-Gumuz, SNNPR, Gambela, Harari Addis Ababa and Dire Dawa were not statistically signi cant. Women who resided in the rural areas were 44.7% less likely to use modern contraceptive compared to those women living in urban areas. Married or in-union women who were followers of Muslim and Other religions were (41.4% and 42.9% respectively) less likely to use modern contraceptives than those who were followers of Orthodox Christian. Women who were in middle and rich wealth status were (1.553 times and 1.710 times respectively) more likely to use modern contraception than those in the poor wealth status which is the reference group. Women who had husband who wants more children to have were 12.6% (AOR = 0.874 CI: (0.765, 0.999)) less likely to use contraceptive than women who desired same number of children with their husband. Women who had 1-2 number of children were 1.587 times more likely to use modern contraception (AOR: 1.587, CI: (1.257, 2.004)) than those married women with no children ever born. Similarly, the odds of married women with ve and above children were 29.1% less likely to use modern contraception than those with no children ever born. In addition, women working status was also statistically signi cant. The odds of married or in-union women who were currently working are 27.9% higher than the odds of married women who were not currently working.
Women who were sexually active four weeks before the survey were 3.614 times more likely to use a modern contraceptive method than women who were sexually inactive in the last 4 weeks. Also, the study veri ed that respondents education level, husband/partner's educational level, age difference between husband and women, and mass media exposure were signi cantly associated with modern contraceptive use in the bivariate analysis but have no statistical signi cant association after controlling for potential confounders ( Table 2  with modern contraceptive use compared with a very young age (15-19 years); which is consistent with results from Ethiopia and Zambia [13,16,17]. However, we found that there was no signi cant association between women aged (20-24 years) and contraceptive use. This study shows that the odds of using modern contraceptive of married or in-union women is high in Amhara region. On the other hand, Married women who lived in Somali had smaller odds than other regions. This nding is similar to a study in Ethiopia [12]. Residence was another signi cant factor, which is associated with current use of modern contraceptive in the current study. The odds of women who resided in the rural areas were 44.7% less likely to use modern contraceptive compared to those women living in urban areas. This result is consistent with the results of Ethiopia, and Sierra Leone [2,[18][19][20]. In addition, the study also revealed that religion is a major predictor of contraceptive use among women. Women belonging to Muslim and other religion (Catholic, traditional and other) were less likely to use contraceptive methods than Orthodox Christian's. This result is consistent with ndings from Ethiopia and Zambia, which reported religion as a major predictor of modern contraceptive use [12,16,21]. In this study women's who were married or in union, wealth status shows statistically signi cant effect on the current use of modern contraceptive method, as the level of women's wealth status increases, the likelihood of using modern contraceptives increases. Married women who were in the middle and rich wealth status were more likely to use modern contraception than those in the poor wealth status, which is the reference group. This result is similar to that of a study in Ethiopia, Ghana, and Sierra Leone [12,13,18].
Women's working status was a signi cant predictor of modern contraceptive use. Workers were more likely to use modern contraceptive methods than the non-working women's were. This nding was consistent with other studies [12]. Women who were sexually active four weeks before the survey were more likely to have used a modern contraceptive methods [5,18].
Women who had 1-2 numbers of children were 58.7% more likely to use modern contraception than those married women with no children were ever born. Similarly, the odds of married women with ve and above children were 29.1% less likely to use modern contraception than those with no children ever born.
This nding is similar to previous reports from Bangladesh, Tanzania, and Ethiopia which reported that as the number of total children increases, use of modern contraceptives increase [20,22,23]. Formal education viewed as a fundamental to increase individual understanding, and to adopt method of contraceptive use. Nevertheless, it is surprising to nd that both respondent's education level and husband's educational level had no signi cant effect on the use of modern contraceptives. This contradict multiple studies done by other researchers [16,18,21,25]. The variable media exposure is not signi cantly associated with modern contraceptive use which is inconsistent to previous studies showing that exposure to mass media is an important factor to use any form of modern contraception [12,26].

Conclusions
The study identi ed how socio demographic characteristics of married or in-union women affect the current use of modern contraception. The chi-square test of association indicated that a number of factors, including age, region, place of residence, religion, wealth status, total number of children, women and husbands education level, media exposure, husband desire for children, age difference, women working status and sexual activity in the last 4 weeks were associated with current use of modern contraceptives. While in the nal binary logistic regression model with complex sampling design found that age, religion, residence, region, wealth status, husband desire for children, women's working status, sexual activity in the last 4 weeks, and total number of children ever born were highly signi cantly predicted the use of modern contraceptives among married or in union women in Ethiopia.

Limitations Of The Study
This study has some limitations. There are certain important determinants that can predict current use of modern contraceptives, which are not included in this study. Therefore, our ndings need to be interpreted with caution, since women may have used or not used modern contraception for reasons not included in this study. In addition, the study did not adjust for other important factors like cultural, social, relationship-related factors that could in uence women's use of modern contraception that are not explored in the EDHS data set. Additionally, our use of secondary data for analysis has limited us to choose from available variables in the dataset.

Declarations
Ethics approval and consent to participate This study used secondary data of freely available data online. Permission to access the 2016 EDHS data was obtained from the DHS program by agreeing with the conditions of DHS data use stated on the DHS consent letter. Ethical clearance to conduct the survey was approved by the Ethical Review Board of Ethiopia Central Statistical Agency (CSA) and written informed consent was taken from study participants.

Consent for Publications
Participants were made aware in the consent form that the results of the study would be published but that no data would be presented to allow the identi cation of individuals.
Availability of data and materials The SPSS datasets used during the current study are available from the corresponding author on reasonable request.

Competing Interests
The authors declare that they have no con ict of interests.

Funding
No organization funded this research.
Authors' Contributions SC develops, wrote the proposal, analyzed the data, and wrote the manuscript. DT approved the proposal with some revisions, participated in analysis and manuscript writing. Both authors read and approved the nal manuscript.

Figure 1
Shows how the subsample of the study was derived