Ethiopia context

Ethiopia, the second populous country in Africa with an estimated population of about 105 million in 2107 [26], has recently experienced progressive declines in fertility. The total fertility rate (TFR) has declined to 4.6 children per women in 2016 from 5.5 in 2000 [27, 28]. Historically Ethiopia had one of the lowest contraceptive prevalence rate with little or no consideration to issues of contraceptive method mix. Family planning services began early in 1966 with the establishment of the Family Guidance Association of Ethiopia (FGAE), an International Planned Parenthood Federation affiliate [29]. In the 1980s, the Ministry of Health (MOH) added family planning to its maternal and child health program but family planning and population issues remained unpopular until the promulgation of a national population policy in 1993. The objectives of the population policy included increasing the prevalence of contraceptive use (CPR) to 44% by 2015 [30],

Ethiopia experienced remarkable progress in improving access to family planning since 2000 although contraceptive method mix has been dominated by one method, injectable. The use of modern contraception increased from 6.3% in 2000 to 35% in 2016 [27, 28]. Analysts have attributed the success of Ethiopia’s family planning programs to four important and interrelated factors: Political will, donor support, non-governmental and public-private partnerships and the government’s establishment of a network of Health Extension Workers [29, 31]. Olson and Piller (2013) provided a detailed review of the contribution of each of these four factors to the growth of family planning in Ethiopia, of which we echo the role of the Health Extension Program (HEP).

The HEP was launched in 2003 as part of the second five year Health Sector Development Program (HSDP) to increase access to promotive and preventive health services including family planning methods to the largely rural population of Ethiopia. Every village (Kebele) has a health post staffed with at least two female health extension workers (HEWs) who spend most of their working days in the community conducting house to house visits to teach and counsel women and family members on health promotion and prevention of diseases [32]. They help to generate demand for family planning, address misconceptions about family planning, counsel clients on the family planning methods, provide methods available at the health posts and refer clients to health centers in cases of extended needs. They provided mainly short term family planning methods (pills, condoms and injectables) but since 2009, with additional training, HEWs were able to insert implanon at the community health post under close supervision of health workers in the nearest health centers.

Building on the Population Policy and National Health Policy, the Ministry of Health developed a Reproductive Health Strategy (2006–2015) in 2006 which was updated in 2016 (2016–2020). While these previous policies and strategies gave little attention to the promotion of long acting reversible and permanent contraception, the revised reproductive health strategy of 2016 clearly articulates the government’s commitment to expanding access to LARP methods with as target to account for half of method mix by 2020 [33]. The Ministry also developed and costed the implementation plan for achieving the targets set in the strategy[34]. Among the efforts being made to attain this target includes an ongoing task shifting and task sharing with the implanon scale up program, IUD revitalization program in which HEWs and mid-level providers play a major role through demand creation, referral and insertion of implants and IUD [35].

Despite the rapid increase in contraceptive use, several challenges remain. There is a highly skewed method mix dominated by injectables, poor informed choice and higher contraceptive discontinuation rate [28]. Data on method information index shows that only 30% of users reported receiving adequate information for informed choice and nearly 35% of users discontinue a method within the first year of acceptance Above all, wider geographic and socio-economic disparities in contraceptive use exist [28]. Understanding the contextual factors that shape the uptake of LARP methods is thus important for programmatic interventions that aim at a rapid scale up of LARP methods, improve method choice and reduce contraceptive discontinuation rates.

Data source

This study uses data from Ethiopia Demographic and Health Surveys (DHS) conducted by the Ethiopian Central Statistical Agency (CSA) and ICF International. DHS is a nationally representative household survey of cross-sectional design. DHS surveys collect detailed information on individual and household characteristics, fertility, sexual behavior, contraception, and related reproductive behaviors from women of reproductive age (15–49 years). The latest 2016 DHS was used to examine the contextual factors that influence choice of long acting reversible and permanent contraception in Ethiopia.

In all the surveys, a two-stage stratified sampling is employed. At stage one, regions are stratified into urban and rural areas. Enumeration areas or clusters from each region proportional to the size of the national population were selected. In sampled households, all women aged between 15 and 49 years who consent to participate in the survey were interviewed. In the 2016 Ethiopian DHS, a total of 16,650 households and 15,683 women participated–with a reasonably high response rate of 98% and 95% respectively [28]. This study draws a sub-sample of married women of age 15–49 from the national sample and included a total of 6994 married women (weighted = 7352) of age 15–49 who were not pregnant and not infecund at the time of the survey.

The DHS included several questions that differentiate the exposure status of women; whether pregnant, postpartum amenorrheic, menopausal, infecund or fecund. Women are considered fecund at the time of the survey if they have been married or in union, are not menopausal and not postpartum amenorrheic and not pregnant but have had birth in the five years preceding the survey [28]. Sexually active unmarried young women (15–24) constituted a very small proportion (less than 8% of young women 15–24 years) due to low reports of sexual activity and so were not included. Moreover, women who reported they were infecund and menopausal at the time of the survey were also excluded.

Variables

The outcome variable in study is current use of long acting reversible and permanent contraceptive methods–a dichotomous variable coded ‘1’ if women or their partners were using implants, IUD and female or male sterilization and ‘0’ otherwise. Those coded as ‘otherwise’ include nonusers of contraception and users of short acting family planning methods (including injectables, pills, condom and natural methods). Several variables that have been associated with contraceptive use in literature were included as independent variables. We categorized them into three sets: individual and household levels, health facility and community-level variables.

Individual and household level variables in this analysis represent woman and household specific characteristics. These include age of women, education, religion, urban or rural residence, household wealth index, employment status, parity and fertility preferences. We recoded women’s age in to four categories (15–19, 20–29, 30–39 & 40–49 years). Education is recoded into three categories as; ‘no education’, ‘primary’ and ‘secondary and above’. The household wealth index is a composite variable based on data on household possessions and amenities collected in DHS surveys. This index categorizes women into five categories (poorest, poorer, middle, richer and richest). In the DHS, women are considered as employed if they had done any work other than their regular housework in the 12 months preceding the survey whether they paid or not. We then regroup women’s employment status into three categories: unemployed, paid employment, and unpaid employment. Parity is also recoded into four categories as the distribution of women by their number of living children (0/1, 2–3, 4–5 and 6⁺). Women’s fertility preference is recoded into four categories: desire for a child soon, desire more children later, desire no more children and undecided.

Access to family planning information was measured using the percentage of women in the cluster who heard about family planning messages through radio, television, newspapers or magazines or through mobile phones in the few months preceding the survey. Moreover, the DHS also asks women whether they were contacted by community health workers or family planning providers in the year before the survey. Access to family planning services was difficult to measure as the DHS did not include any direct question on access to services. In this paper, we used proxy variables of barriers to medical care—the DHS question on whether distance to facility is a big problem or not as a proxy for ‘distance to health facility’.

Other contextual community level variables considered include community level women’s empowerment, fertility norms, poverty. We created these cluster level variables by aggregating individual level data to the cluster level. These include cluster level poverty, which was created from the mean values of wealth index categories of the individual households aggregated at the cluster level, cluster level education created by aggregating values of the educational level of women in the cluster and a new category of regions, development region, based on their similarities in terms of economic activities, settings and socio-economic characteristics. Community level women empowerment variable was created by aggregating a composite variable constructed from variables on household and women’s health care seeking decisions. The composite variable indexes the proportion of women who participated in the decisions jointly with their husbands or alone. The variable community level fertility norms was constructed by aggregating the number of children desired at the cluster level. Similarly, the variable community level marriage norm was constructed by aggregating median age at marriage at the cluster level. The variable ‘development region’ is created by categorizing regions based on their dominant economic activity and based on the Ethiopian government’s classification of region into: ‘urban’ ‘agrarian’ and ‘emerging’ regions [35]. The first category (urban) includes the city administrations of capital Addis Ababa and Dire Dawa which are dominantly urban. The second category included dominantly rural regions with better distribution of health facilities and includes Amhara, Oromia, Tigray, and Southern Nations, Nationalities and People’s region and Harari. The remaining four regions (Afar, Somali, Benishangul Gumuz and Gambella) are not at the same level with the other regions in terms of provision of and access to healthcare and are thus classified as ‘emerging’ regions.

Data analysis

Data were analyzed using STATA software version 14. First, we conducted exploratory analyses of each of the variables as well as descriptive analysis looking at trends and patterns in the use of modern contraception. Bivariate regression analysis was done to examine associations between contraceptive use and the selected individual, household and community level variables. Variables were included into the multivariate mixed effects logistic regression if they had a significant association at the bivariate level (P-value <0.05). Multilevel mixed effects logistic regression analysis was used to identify the net effects of each explanatory variable after adjusting for potential confounders.

Given the hierarchical nature of the DHS data, we used a multilevel logistic regression analysis to capture covariance at the individual/household and cluster level. Multilevel modelling is a robust method for analyzing hierarchically clustered data such as the DHS [36]. It provides a mechanism for measuring the influence of community factors and unobserved community effects on health outcomes. Moreover, multilevel modelling corrects the estimated standard errors to allow for clustering of observations within units. We reported fixed effects odds ratios (OR) and 95% confidence interval (CI) after controlling for the effects of individual and household characteristics. Statistical significance was determined using two tailed Wald test with the significance level of alpha of 5%. We used Hosmer and Lemshaw goodness of fit statistics to check the fitness of the full model (model IV) and the result showed that the model fit was good. The random effects were expressed in terms of community level variance (δ2). Intra-class correlation coefficient (ICC) was used to examine clustering and the extent to which contextual factors explain the unexplained variance of the empty model.

Multilevel regression analysis

Table 3 shows results of the multi-level logistic regression analysis of the association of LARP methods with several individual, household and community level contextual factors (Models I-IV). Bivariate regression analysis was conducted to test the association between the outcome and explanatory variables, and variables were included in to the multivariate multi-level regression analysis based on their association at the bivariate level (S1 Table). Models I-III present results of the different individual, household and community level variables while model IV shows results of all variables included in the multilevel regression model. The results show that there is a considerable level of variation in the uptake of LARP methods at the individual, household and community levels. As reported by intra-cluster coefficient (ICC) nearly 34% of the variation in the uptake of LARP methods is attributed to whether a woman resides in one community or another (Table 3, model I). The variation in use of LARP methods was reduced by the addition of contextual factors. Nonetheless, the community and health facility variables included in the analysis do not fully explain the community-level variation in contraceptive use as several health facility and quality of care related variables were not measured through the DHS.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multilevel analysis of the association between individual, household and cluster level variables with use of long acting and permanent contraception, Ethiopia 2016.

https://doi.org/10.1371/journal.pone.0209602.t003

The regression analysis identified several individual, household and community level factors that influence the uptake of LARP methods. Among individual and household factors, age of women, wealth status, women’s employment and fertility preference were significantly associated with use of LARP methods. The odds of using LARP methods was significantly lower among adolescent of age 15–19 (OR, 0.53; 95% CI, 0.32–0.85) and older women of age 40–49 (OR, 0.63; 95% CI, 0.44–0.90) compared to women in their 20’s. Women’s education and place of residence did not show significant association in all the models. The odds of using LARP contraception increased as household wealth status increased. For instance, the odds of using LARP methods was nearly twice (OR, 1.91; 95% CI, 1.20–3.06) higher among women in the richest wealth category compared to those in the poorest quintile. Likewise, the odds of using LARP methods was 41% higher (OR, 1.41; 95% CI, 1.14–1.75) among women employed with cash income compared to those who were not working after adjusting for the effects of individual, household and community level variables. Among individual factors, women’s fertility preference was a stronger predictor of the choice of LARP methods. The odds of choosing LARP methods was more than twice higher among women who want to space, who want to limit or women who were not sure of the timing of their next child compared to those who desire to have a child soon (Table 3).

The study focuses on the role of cluster level variables in explaining the between cluster variability in the use of LARP methods. Model III shows that among the cluster level variables: exposure to family planning information, knowledge of LARP methods, region and women’s empowerment were significantly associated with use of LARP methods (Model III). After adjusting for the effects of individual and household factors, few contextual community level factors appeared significantly associated with use of LARP methods (Model IV): exposure to family planning information on media, knowledge of LARP methods, region and women’s empowerment were significantly associated with choice of LARP methods. Exposure to family planning information through media had a significant positive effect, with higher odds (OR, 1.24; 95% CI, 1.00–1.54) of using LARP methods among women with exposure to family planning information on the media compared to those who did not. Perceived distance to health facility had a marginal positive effect (OR, 1.21;95%CI,0.99–1.48) on the use of LARP methods with women who reported distance being not a big problem having higher odds of choosing LARP methods compared to those who reported distance is a big problem.

Women empowerment at the community level has shown significant association with the choice of LARP methods after adjusting for the effects of individual, household and other community level variables. Residence in a community with higher participation of women in household decisions increased the odds of using LARP methods by 37% (OR, 1.37; 95% CI, 1.03–1.83) compared to residence in communities where women’s participation in household decisions is low. Similarly living in clusters or communities with higher knowledge of LARP methods significantly increased women’s odds of using LARP methods. But, living in communities with higher mean desired number of children and lower median age at marriage did not show significant association with the uptake of LARP methods after adjusting for the effects of individual and household covariates. Corresponding to the regional variation in the uptake of contraception in Ethiopia, the regression analysis also showed that the odds of using LARP methods is significantly lower (OR, 0.45; 95% CI, 0.31–0.64) in the regions that are named emerging and characterized by pastoral nomadic way of life compared to the regions that are dominantly rural and agrarian areas. The uptake of LARP methods did not vary between the urban and mixed farming rural regions however.