Trends and factors associated with iron supplementation during pregnancy among women who had pregnancy in the preceding ve years of the 2016 EDHS: Multivariate decomposition analysis

Introduction: Study estimated in 2019 iron deficiency related anemia accounts 1.89% and 1.07% of total disability adjusted life years (DALYS) to reproductive age women globally and sub saran Africa respectively. Anemia at the time of pregnancy may cause small sized baby, preterm birth, and perinatal, neonatal and maternal mortality. Iron and folic acid supplementation during pregnancy will minimize the risk of iron deficiency and risk of anemia. Methods: for this time series analysis, dataset of 2005, 2011 and Ethiopian demographic health surveys was used. A total weighted sample of 7234 in 2005, 7786 in 2011 and 7543 in 2016 are encompassed in this study. The Logit based Multivariate decomposition analysis for non-linear response outcome was calibrated to detect the factors contributed to the change in iron supplementation between 2005 and 2016 study period. The Logit based multivariate decomposition analysis utilizes the output from the logistic regression model to assign the observed change in iron supplementation over time into two components. Stata version 16.0 was used to analysis the data. Results: proportion of iron supplementation during pregnancy increased by more than a third (32%) between the period of 2005 and 2016. From the output of multivariate decomposition regression, major of the change in iron supplementation was due to change in coefficients. Women’s age, working s status and region of living (all except Dire Dawa) were variables as sources of change in iron supplementation under the category of change in coefficients. Additionally, age, region, wealth status, employment status/working status, ANC follow up to the last pregnancy and media exposure were variables attributable to the change of iron supplementation during pregnancy under the category of endowment. Conclusion: rate of iron supplementation at the time of pregnancy increased dramatically over the last decade in Ethiopia. Change in the proportion of women aged ≥ 35 years, population of each region, wealth status (poorer, middle & richer), working women, ANC follow up and media exposed as well as change in behavior of women aged ≥ 35 years, employed women and region of living were sources of change for iron supplementation between 2005 and 2016 period.


Introduction Background
Anemia is persisting as a giant health obstacle in the world. According Study estimated in 2019 iron deficiency related anemia accounts 1.89% and 1.07% of total disability adjusted life years (DALYS) to reproductive age women globally and sub saran Africa respectively (1,2). In spite of dramatic progress in maternal health programs in the last decades, anemia remains standing health problems in middle and low income countries (3,4). As per World Health Organization (WHO) statement, more than 40% of pregnant women are suffering from anemia globally and more than fifty percent of anemia was attributable to iron deficiency (2). Anemia at the time of pregnancy may cause small sized baby, preterm birth, and perinatal, neonatal and maternal mortality (5,6). Iron and folic acid supplementation during pregnancy will minimize the risk of iron deficiency and risk of anemia(2). To overcome these problems, WHO recommended any pregnant women to take 30 mg to 60 mg of iron and 0.4 mg of folic acid supplementation daily through the mouth (2, 7).
In Ethiopia, the performance of iron supplementation at the time of pregnancy is not satisfactory and has not reached the World Health Organization recommendations and evidence about trend of iron supplementation and contributing factors to change of iron supplementation over time are limited. This study aimed to assess the trend and factors associated with change of iron supplementation during pregnancy in Ethiopia using 2005, 2011 and 2016 demographic and health survey data. Findings from this study may support policymakers and program planner through providing information for evidence based decision making and wise utilization of scarce resources. In addition, key insights may get about the effect of characteristics and population structure changes for the improvement of iron supplementation during pregnancy period over the last 11 years.

Methods and study setting
Ethiopia is one of the sub-saran countries and the second largest in populated in Africa. The fourth Ethiopian health sector development plan (HSDP IV) introduced the current three tiered health system.
Based on the 2007 Population and Housing Census (PHC) projection estimate a total population of 103,000,000 in 2021. According to the 2007 census a majority of the population (83.6%) was living in rural areas, the average household size was 4.7 persons and 47% of the total women were between 15 and 49 years of age (15). Primary healthcare (health posts, health centers, and primary hospitals) is delivered at the primary level, serving up to 100,000 population; secondary level services are offered by general hospitals; and tertiary level services are given by specialized hospitals (16). This study used the data of 2016 Ethiopian demographic health survey (EDHS), which was researched from January 18, 2016, to June 27, 2016.

Variables of the study
For this study, iron supplementation was an outcome variable of interest. Women who supplemented with iron while they are pregnant were coded as "1" and those not iron supplemented were coded as "0".
The potential independent variables after deep literature review considered for this study were; Maternal age, Maternal and partner Education, Marital status, Maternal occupation, Wealth status, perceived distance to a health facility, Desirability of last pregnancy, Birth order, media exposure and Autonomy to decide on health care, ANC visits, place of residency, religion, region.

Operational definition
Media exposure: Created by combining whether a respondent reads the newspaper, listen to the radio, and watch television and labeled as "exposed" and coded "1". If the subject was exposed to at least one of the three Medias labeled as "not exposed" and coded as" 0". Autonomy to decide on health care.: women who decide their health issue alone are labeled as " autonomous" and coded as "1", who decides jointly with their partner were labeled as "jointly" and coded as "2", while those responded their partner alone decides on their health issue were labeled as "un autonomous " and coded as " 0".
Working status: women who are categorized under "not working" were coded "0" and those who categorized as professional/technical/managerial, clerical, sales, agricultural, services, skilled manual, unskilled manual or others were considered as "working" and coded "1".

Sample Size and Sampling Procedure
A two staged stratified clustered sampling method was used from all 11 geographic areas (9 regions and 2 city administration). The first strata included a total of 540 EAs in 2005, 624 in 2011 and 645 EAs were selected proportionally to EA size of nine geographical regions and two administrative cities.
In the second strata each eleven administrative division was further sub grouped into urban and rural, bringing results of 21 sampling strata. At this moment, by means of equal probability strategy, 27 to 32 households were selected from each cluster. If women had more than one birth history within the last five years preceding the survey, the most recent births were considered for this study.

Data management and analyses
To adjust for the complex sampling scheme of the demographic health surveys, to restore the representativeness of the survey and to get reliable statistical estimates, data were weighted and "svyset" STATA command was applied in all descriptive and analytical part of analysis. Svyset command was defined as "svyset as svyset [pw=wgt], psu (v021) strata (v022)". All statistical testing was performed at a 95 percent level of significance.  The E component refers to change in iron supplementation over the 11 years imputable to differences in endowments or characteristics. The C component refers to change in iron supplementation over the 11 years imputable to differences in coefficients or effects. Percent contribution with a 95% confidence interval (CI) of coefficients and a p-value <0.05 were reported. Table 1  percentage of women who perceived the distance from health facilities is not a big problem was elevated from 27% in 2005 to 42% in 2016. However, point percentage of women who had exposed to media was decreased nearly by 3% percent within the last 11 years of study periods.    Table 3 below compares how much of the change of iron supplementation during pregnancy between 2005 and 2016 is imputable to changes in women's characteristics (endowments) and how much to the effects of these characteristics (coefficients). It is apparent that the increment of proportion of iron supplementation was attributable to the coefficient effects of incorporated independent variables.

Regression-based decomposition results
Relatively, the effect of coefficient is more pivotal (coefficient=0.0718, representing 78.9 percent share of the total change) than the part explained by the endowment effects (coefficient= 0.2533; percent shared=22.08 %). The compositional differences related to women's working status/employment status was the second most positive contributor to the change of iron supplementation during pregnancy (+2.62%), particularly for women who was working/employed compared to women who were unemployed.
Among the very significant contributor compositional factors, positive change of iron supplementation during pregnancy was due to Amhara region (+1.8%), SNNP region (+0.8%), Somalia region (+0.45) and age greater than 35 years. However, Oromia region contributed to the change of iron supplementation negatively (i.e. decrement by 3.72%). Even though statistically significant, compositional factors like wealth status (poorer, middle and rich), media exposure (exposed) and region of living (Addis Ababa, Harari, Gambela, Benishangul and Diredawa)were among the fewer contributors to change of iron supplementation over 11 years of periods.   Coefficient =0.42297***, percent= 130.9 *represents 5 percent, ** represents 10 percent and *** 1% percent level of significance

Changes imputable to effects of coefficients
Holding the effect of change in compositional factors constant, nearly four fifths changes in iron supplementation during pregnancy was attributable to the difference in effect of coefficient between the two study periods (table 3). In regard to change of iron supplementation attributable to change in coefficients between the two surveys, the most important explanatory variables that provide significant contributions are women's age, working status/employment status and region of living (all except Dire Dawa). The change of women's behavior aged 35 years negatively contributed (decrease by 6.75%) to change of iron supplementation. Another interesting contributor was that being employed women increased iron supplementation during pregnancy by 5 point percentage. To illustrate, being employed women increases the possibility of taking iron tablets during pregnancy.
Concerning region of residency, just over a third (31%) change of iron supplementation was imputable to change behavior of Oromia region population. Change of behavior of population of SNNP, Amhara and Somalia regions negatively contributed to the change of iron supplementation during pregnancy between 2005 and 2016 years of study i.e. decrease by 15%), 6.5% and 2.8% respectively. Negligible contribution were also observed due to change of behavior of population living in Afar, Benishangulgumuz, Gambela, Harari and Addis Ababa compared to counterpart Tigray region. Outputs from the multivariate decomposition regression revealed that the rise in iron supplementation use between 2005 and 2016 is imputable more to the changes in effects of characteristics (coefficients) than to the changes in sample population composition (endowments). Variables such as women's age, working status and region of living (all except Dire Dawa) were the contributing factors under category of change in coefficients of multivariate decomposition. The contribution of change in endowment was due to change (either decrease or increase) in proportion of media exposed women, ANC visits, working women, wealth status (poorer, middle and richer), region of living and women's age (being ≥

years).
Holding the effect of change in effect of endowment constant, being employed (working) displayed predominant contribution to the rise of iron supplementation. This finding is in line with other studies done in India (21). This can be explained through, employed women may have opportunity to access information related to health in general. Employed women are more likelihood of visits health facilities due to their ability to offer cost of transport and medicine. Other reason might be related to their education level, that educated women are more likely to be hired to job and educated women have better knowledge about the severity and consequence of the problems (14). But, negative relationship was observed between women aged ≥35 years old and iron tablet uptake. This outcome is contrary to other study done in Ethiopia (8), which surmised that as age increases, the likelihood of taking an iron tablet increases. A possible explanation for this might be that sample size and method of analysis difference. As women's age increases the proportion of iron uptake diminishes, this might be due to older women might think of themselves as experienced and confident about their pregnancy outcome and refused to take iron tablets. The more aged, the more likely to have and to manage a large family.
This finding, while preliminary, suggests that older pregnant women needs special support related to awareness on iron deficiency related pregnancy outcome. Change in behavior of population residing in Oromia, SNNP and Amhara region decreased iron supplementation rate in a striking way.
From the compositional factors, ANC follow up was the dominant contributor to the change of iron uptake, in which the increment of proportion of women who had at least one ANC visits showed significant impact on the elevation of iron tablet uptake. This finding is similar with other studies (8,9,11,22). The possible explanation for this moment is, as iron supplementation is one part of ANC services, women who had exposed to ANC services are more likely to get health information related to iron supplementation and discussion with health care providers about the importance of iron supplementation during pregnancy is inevitable. Additionally, increase in proportion of women who lived Amhara region increases the rate of iron supplementation utilization during pregnancy. However, diminished iron supplementation utilization. Decrease in composition of women who resides in middle and richer households also showed slight fall in prevalence of iron supplementation. Event though, the period of study is quite enough to appreciate significant change of iron supplementation, Even though, the authors tried to analyze change of iron supplementation, due to its secondary nature of the dataset many predictor variables might have missed in this study. (Table 4)

Conclusions
Over the last 11 years, iron supplementation rate rose significantly among pregnant women in Ethiopia.
The Logit based decomposition analysis noted that about 79% percent of rise of iron supplementation over the period of 2005 to 2016 was attributable to the change of effect of characteristics (coefficients).
Variables such as women's age, working status/employment status and region of living (all except Dire Dawa) were the major sources of Change of effect of coefficients. In regard to change of composition of characteristics, age, region, wealth status, employment status/working status, ANC follow up to the last pregnancy and media exposure were the contributory compositional factors. It is better to focus program implementers on rising iron supplementation of Tigray, Harari, Gambela and dire Dawa regions as well as older women.