Study area and data source
The study was conducted in Ethiopia, which is located in the North Eastern part of Africa, also known as the horn of Africa, lies between 30 and 150 North latitude and 330 and 480 East longitudes. This study used the EDHS 2016 dataset which were conducted by Central Statistical Agency (CSA) in collaboration with the Federal Ministry of Health (FMoH) and the Ethiopian Public Health Institute (EPHI). Data were accessed from their URL: www.dhsprogram.com by contacting them through personal account after justifying the reason for requesting it. Then reviewing the account permission was given via the email. A cross-sectional study design using secondary data from 2016 EDHS was conducted. All reproductive age women who were married, fecund and or sexually active were included in the study and those women who were sexually active 30 days prior to the survey were excluded.
Sample size determination and sampling procedure
A total of 9,125 weighted reproductive age women who were married, fecund and/or sexually active were included (Figure1).
The 2016 EDHS sample was stratified and selected in two stages. In the first stage, stratification was conducted by region and then in each region stratified as urban and rural, yielding 21 sampling strata. A total of 645 EAs (Enumeration Areas) (202 in urban areas and 443 in rural areas) were selected with probability proportional to EA size in each sampling stratum. In the second stage, a fixed number of 28 households per cluster were selected with an equal probability systematic selection from the newly created household listing.
Variable measurement
The outcome variable for this study is dichotomized as unmet need (yes/no) which was generated from a constructed EDHS variable. It is the sum of unmet need for spacing and limiting and reproductive age women who were married, fecund and/or sexually active have unmet need if they don’t want any more children or want to delay their next birth for at least two years but not using contraception. Pregnant or amenorrheic women with unwanted or mistimed pregnancies or births were also considered to have unmet if they were not using contraception at the time they conceived [15, 17, 24].
Data processing and analysis
Data cleaning was conducted to check for consistency and missing value. Recoding, labeling and exploratory analysis was performed by using Stata/SE version 14.0. Descriptive statistics was used to present frequencies, with percentages in tables, graphs and using texts. Sample weight was used in order to compensate for the unequal probability of selection between the strata that were geographically defined, as well as for non-responses.
Multilevel analysis was conducted after checking that the data was eligible for multilevel analysis that means Intra-cluster Correlation Coefficient (ICC) greater than 10% (ICC=12.74%). Since DHS data are hierarchical, i.e. individuals (level 1) were nested with in communities (level 2), two-level mixed-effects logistic regression model was fitted to estimate both independent (fixed) effects of the explanatory variables and community-level random effects on unmet need for family planning. The log of the probability of unmet need for family planning was modeled using two-level multilevel model as follows:
Log [ = β0 + β1 Xij+ B2 Zij+µj+eij
Where, i and j are the level 1 (individual) and level 2 (community) units, respectively; X and Z refer to individual and community-level variables, respectively; πij is the probability of unmet need for family planning for the ith women in the jth community; the β’s indicates the fixed coefficients. Whereas, β0 is the intercept-the effect on the probability of unmet need for family planning in the absence of influence of predictors; and uj showed the random effect (effect of the community on unmet need for family planning for the jth community and eij showed random errors at the individual levels. By assuming each community had different intercept (β0) and fixed coefficient (β), the clustered data nature and the within and between community variations will be taken in to account.
During analysis first bivariable multilevel logistic regression was fitted and variables with p-value less than 0.2 were selected to build the 3 models (model1-3).Then the analysis was performed in four steps: Model 0 (empty model or null model/ without explanatory variable); Model 1 (only individual-level factors) Model 2 (only community factors); and Model 3 (both individual and community-level factors). The measures of association (fixed-effects) estimates the associations between likelihood of women to have unmet need for family planning and various explanatory variables were expressed as Adjusted Odds Ratio (AOR) with their 95 % Confidence level. A variable in which its p-value <0.05 was used to declare statistical significance. The measures of variation (random-effects) were reported using ICC, Median Odds Ratio (MOR) and proportional change in variance (PCV) to measure the variation between clusters.
The ICC shows the variation in unmet need for family planning for married reproductive women due to community characteristics. The higher the ICC, the more relevant was the community characteristics for understanding individual variation in unmet need for contraceptive for married reproductive women. The ICC was calculated as follows: [ICC= ], where is the estimated variance of clusters. MOR is defined as the median value of the odds ratio between the area at highest risk and the area at lowest risk when randomly picking out two areas and it was calculated using the formula [ MOR=exp( ) ]. In this study MOR shows the extent to which the individual probability of having unmet need for family planning for married reproductive women is determined by residential area. PCV measures the total variation attributed by individual level factors and area level factors in the multilevel model.
The presence of multicollinearity was checked among independent variables using standard error at cut off point of ±2 and there was no multicollinearity. Log likelihood test were used to estimate the goodness of fit of the adjusted final model in comparison to the preceding models (individual and community level model adjustments).