The Analysis of Socio-Economic and Demographic Factors Associated with Contraceptive Use Among Married Women of Reproductive Age in Rwanda

2.1. Source of Data

The present cross-sectional study used secondary data from the 2014/15 Rwanda Demographic and Health Survey (RDHS), the most recent national dataset on current contraceptive use. This survey was conducted from 9^th November 2014 to 8^th April 2015, where 12793 households were selected. The sampling used in this survey was a two-stage stratified method; in the first stage, 113 primary sampling or enumeration areas were selected from urban areas and 379 from rural areas. The sampling was done with probability proportional to the number of households residing in the village. In the second stage, systematic sampling was used for all households existing in the selected village and 26 households were selected in each village. This survey had three types of questionnaires: a questionnaire for households, a questionnaire for men and a questionnaire for women. Women’s questionnaire was used to collect information for all women aged 15-49 residing in the selected household on background characteristics, fertility preferences, contraception, reproduction, marriage and sexual activities and pregnancy among others and the data from this questionnaire were considered in the present study. The current study considered only currently married women or women living with partner. The eligible women response rate was 99.5% ; more details on sampling techniques and data collection can be found in [9].

2.2. Response Variable-Contraceptive Methods

There are different methods of contraceptive use, such as pill, IUD, injection, male condom, female condom, female sterilization, male sterilization, periodic abstinence, withdrawal, implants/Norplant, lactational amenorrhea (LAM) and standard days method. In the survey women were asked whether they used any of these methods. In the current study, a woman is considered as current contraceptive use if she uses at least one of the above mentioned methods and non-contraceptive use if otherwise and this was coded as “1” and “0” for contraceptive use and non-contraceptive use respectively.

2.3. Independent Variable

The dependent variables used in this study have been used elsewhere [7, 14-17]. These variables are classified as socio-economic and demographic factors: current age of the respondent, the age at first cohabitation (recorded as: <15, 15-17, 18-22, 22-49), woman’s level of education (no education, primary, secondary, tertiary), heard about family planning in last few months in newspapers/magazine (yes, no), heard about family planning in the last months on television (yes, no), visited by family planning worker during last 12 months (yes, no), husband’s or partner’s desire of children compared to the wife or partner (same, husband wants more, husband wants fewer, don’t know), wealth index of her household (poorest, poorer, middle, richer, richest), province of residence (Kigali, south, west, north, east), place of residence (urban, rural), currently working (yes, no), religious affiliation (Catholic, Protestant, Seventh Day Adventist, others), husband’s or partner’s education level (no education, primary, secondary, tertiary), decision-making for using contraceptive (mainly respondent, mainly husband or partner, joint decision, others), person who usually decides on health care (respondent alone, respondent and husband/partner, respondent and other person, husband/partner alone, someone else, others), number of living children (0, 1, 2, 3, 4 and more), size of the family (continuous) and age of the husband or partner (continuous).

2.4. Statistical Methods

Based on the sampling design used when collecting the data in Demographic Health Survey, the current study uses survey logistic regression model in order to account for the complexity of sampling design and account for heterogeneity between observations from same cluster.

2.5. Model Formulation

Let y_ijk denote the contraceptive use status of woman i from stratum j and cluster m, with i = 1, 2, 3, ... ,6841, j = 1, 2, 3, ... ,60 and m = 1, 2, 3, ... ,492. The dependent variable is defined as a dichotomous variable such that y_ijm = 1 if the women i is using any type of contraceptive methods and y_ijm if the women i is not currently using any type of contraceptive methods. It is assumed in the present study that the response variable y_ijm is Bernoulli distributed as y_ijm|μ_ijm~Bernoulli (μ_ijm), where μ_ijm is known as the mean and is given by E(y_ijm) = μ_ijm. It is related to the explanatory variables as:

g(μ_ijm) = X'_ijmβ + U'_ijmδ

where g(.) is the logit link function, β is a p-dimensional vector of regression of categorical explanatory variables, namely, woman’s education level, wealth quintile of the household, visited health facility last 12 months, respondent currently working, woman’s age at fist cohabitation, religion, woman’s current age, heard about family planning on the radio last few months, heard about family planning on the TV last few months, heard about family planning in the newspaper/magazine, province, husband desires children, husband/partner’s education level, age of the husband/partner) and δ is a q-dimensional vector of regression of continuous covariates (age of the husband/partner, size of the family).

2.6. Model Fitting and Testing

The analysis in this study was done utilizing Proc Survey logistic from SAS software version 9.4. The Taylor series method was used in this study as variance estimator and this is the default in SAS. The model fit statistics in survey logistic was assessed based on Akaike Information Criteria (AIC), the Schwarz Criterion (SC) and -2 Log-Likelihood (-2LogL) principles. The model test was done based on Likelihood ratio, score and the Wald test principles.

3.1. Descriptive Analysis

Table 1 presents the descriptive statistics of the participants. Out of 6841 married women or women living with their husband or partner 52.7% of them were currently using any type of contraceptive methods, where 5.9% and 46.8% were traditional and modern contraceptive methods respectively. The mean age of the respondents was 32.81, minimum was 16 and maximum was 49. The mean of the size of the household was 5.0; minimum was 1 and maximum was 15 members. The mean age of the husband or partner was 37.39, minimum was 18 and maximum was 95 years old. It is observed from the Table that 83.0% of the respondents were from rural areas whereas 17.0% were from urban areas. It is also observed that 16.5% of the respondents had no level of education, 70.5% had primary level, 10.3% had secondary education level and only 2.7% of them had tertiary education level. It can be seen from the Table that 55.0% of respondent heard about family planning on the radio, 7.1% on the TV and 6.1% in newspapers or magazines.71.8% of respondents visited health facilities in last 12 months before the period of survey, and 46.3% of them were told about family planning at health facilities. It is observed from Table 1 that 38.2% of the respondents were Catholic, 45.2% were Protestant, 12.8% and 3.2% were from religions other than Catholic, Protestant or Seventh Day Adventist.

The model fit is tested based on the Akaike Information Criterion (AIC), the Schwarz Criterion (SC) and the negative of twice the log likelihood (-2LogL). AIC and SC are used to compare the model of intercept only and the model of intercept and covariates. It is noted that the smaller the value of AIC and SC the better the model fit. It was observed from Table 2 that fitting the model with covariates and intercept had a small AIC and SC compared to model with intercept only. Consequently it was used in the analysis. The global null hypothesis was tested based on Likelihood ratio test and Score test and these two tests were found to be significant (Table 3) therefore the global null hypothesis stating that the slope parameters are zero is rejected.

Table 4 provides the unadjusted and adjusted odds ratios of survey logistic regression model, used to identify the determinants of contraceptive use among women of childbearing age. Wealth index of household, woman’s education level, woman’s age at first cohabitation and her age at the period of survey, her religious affiliation, size of the family, whether the woman heard about family planning on the radio, on TV or in newspapers/magazine, her province of residence, whether she was working or not at the period of survey, whether she was visited by family planning worker in the last 12 months prior to the survey, number of living children, husband/partner’s desire for more children and current age of the husband or partner were found to be significantly associated with the contraceptive use among married women of childbearing age in Rwanda. The results of fixed effect are summarized in Table 4 and the results of interaction effect are summarized in Table 5 and Fig. (1).

The odds of contraceptive use decreases with wealth quintile of the household. A woman from poorer and poorest household was 0.752(p-value=0.0068) and 0.743 respectively (p-value=0.0146) times less likely to use a contraceptive method compared to a woman from richest family. The present study did not find a significant difference between a woman from middle and richest household and association with contraceptive use among women of childbearing age.

It is noted from the results that the age at first cohabitation was found to be significantly associated to contraceptive use. A woman who had her first cohabitation at age less than 15 years old was 0.595 (OR=0.595, p-value=0.0235) times less likely to use a contraceptive method than a woman who had her first cohabitation from age group 22-49 years old. However, there was no significant difference in the current contraceptive use among woman who had her first cohabitation at the age 15-17, 18-21 and a woman who had her first cohabitation at 22-49 years old.

It was observed from the analysis that the working status of the woman was significantly associated with current contraceptive use among married woman or woman living with her partner. A woman who was not working at the period of survey was 0.769 (OR=0.769, p-value<.0001) less likely to use contraceptive method compared to a woman who was working at the period of the survey.

It is observed from the Table that the current age group of the woman had a significant effect on current contraceptive use among women. A woman who was in age group 25-34 and 15-24 at the period of survey was 1.584 and 1.222 respectively (OR=1.584, p-value=0.0007 and OR=1.222,p-value=0.0125) times more likely to use a contraceptive method than a woman who was in the age group 35-49 years old.

A woman who was not visited by a family planning worker in the last 12 months prior to the survey was 0.784 (OR=0.784, p-value=0.0001) times less likely to use contraceptive than a woman who was visited by a family planning worker in the last 12 months prior to the survey. The use of contraceptive varied with the woman’s province of residence. A woman from Eastern, Northern and Southern province was 1.331, 1.590 and 1.237 respectively (OR=1.331, p-value=0.0001, OR=1.590, p-value <.0001, OR=1.237, p-value=0.0288) times more likely to use contraceptive methods than a woman from Western province. But a woman from Kigali did not show any statistical difference compared to a woman from Western province.

The number of living children was significantly associated with the use of contraceptive method. The results revealed that a woman who had one child was 0.495 (OR=0.495, p-value <.0001) less likely to use contraceptive method compared to a woman who had four children or more. It was also found that a woman who had two children was 0.768 (OR=0.768, p-value=0.0058) less likely to use any type of contraceptive method compared to a woman who had four children or more.

A woman who did not hear about family planning in newspaper/magazines was 0.682 (OR=0.682, p-value=0.0030) times less likely to use contraceptive method compared to a woman who heard about family planning in newspapers/magazines. A woman whose husband desired more children or fewer children than her was 1.319 and 1.278 respectively (OR=1.319, p-value=0.0033, OR=1.278, p-value=0.0084) times more likely to use a contraceptive method compared to a woman whose husband/partner desired the same number of children as her.

A woman who did not hear about family on the radio last few months prior to the survey was 0.859 (OR=0.859, p-value=0.0077) times less likely to use contraceptive method compared to a woman who heard about family planning on the radio last few months prior to the survey.

It was also noted from the results that the age of the partner has an important significant effect on his wife or partner’s current contraceptive use. The results revealed that for a unit increase in age of the husband, the odds of contraceptive use by his wife or partner reduced by 1.5% (OR=0.985, P-value=0.0002).

3.2. Interaction Effects

The current study considered various two-way interaction effects but only women’s education level and her religious affiliation was significant and therefore used in the analysis. The results are summarized in Table 5 and Fig. (1). It is observed from Fig. (1) that contraceptive use increases with level of education among women from Catholic and women from religions other than Catholic, Protestant and Seventh Day Adventist. It is also observed from the Figure that there is a big gap between women with tertiary education level and women with no education level on contraceptive use among women from religions other than Catholic, Protestant and Seventh Day Adventist. It is observed from the Figure that contraceptive use is higher among Seventh Day Adventist women with secondary education, followed by women with primary education. It is also observed from the figure that contraceptive use was higher among Protestant women with no education and tertiary education followed by women with secondary education.