Data sources
A secondary data analysis was conducted using the 2016 Ethiopia Demographic and Health Survey (EDHS) [15]. The EDHS was conducted under the auspices of the Federal Ministry of Health (FMoH) and the Ethiopian Public Health Institute (EPHI), implemented by the Central Statistical Agency (CSA) and supported by ICF International, USAID, UNICEF, the World Bank, United Nation Population Fund (UNFPA) and UN Women [15].
The EDHS collected information from a nationally representative sample including both men and women to provide up-to-date estimates of demographic and health indicators. Data were collected using a probability-based two-stage stratified random sampling design. Data collection took place over a 5.5-month period from January to June 2016. A detailed information on the methodology is available in the EDHS report [15].
Study population and sample size
The sampling frame for the current study consisted of 10640 women aged 15-49 years who undergone anemia testing. After excluding women with missing data on high risk fertility behavior (n=1742) or any of the other covariates assessed in the current study (n=3099), 5799 women were retained for the final analysis.
Study variables
Outcome of Interest: Anemia status
Anemia status of women aged 15-49 years was the outcome of interest in the current study. Anemia cut-off points used in the EDHS were those recommended by WHO. For non-pregnant women, any anemia was defined as Hb<12 g/dl, and for pregnant women as <11 g/dl [16]. Blood samples were taken from a finger prick and analyzed using a HemoCue analyzer. Hemoglobin levels were adjusted for cigarette smoking and altitude. Detailed procedures are explained elsewhere [15]. Anemia was used as a dichotomous variable (no, yes) instead of using a continuous hemoglobin concentration.
Exposure variable: Maternal high-risk fertility behavior
The exposure variable of the current study was maternal high-risk fertility behavior. The definition of high-risk fertility behavior adopted by the 2016 EDHS was applied [15]. High-risk fertility behavior was constructed using four different variables; (i) mother aged < 18 years at the time of delivery; (ii) mother aged > 34 years at the time of delivery; (iii) mother of a child born after a short birth interval (<24 months); and (iv) mother of high parity (> 3 children). The presence of one of the four conditions was termed a single high-risk fertility category. Combinations of two or more conditions were referred to as multiple high-risk category.
Three exposure variables were used for the current analysis: (i) any high-risk fertility behavior vs. none; (ii) exposure to different categories of high-risk fertility behaviors v. none (categorized as exposure to single high-risk category vs. none and multiple high-risk categories vs. none); and (iii) the specific types of high-risk fertility behaviors v. none.
Covariates:
Several variables that are theoretically and empirically linked to anemia [6, 7] were included in the current analysis such as: maternal age (15–24, 25-34, 35-49), education (no education, primary, secondary or higher), occupation (no, yes), residence (urban, rural), number of household members (1-4, 5-6, ≥7), wealth index (poor, middle, rich), religion (non-Muslim, Muslim), body mass index (underweight, normal, overweight/ obese), current use of contraception (none, hormonal, barrier, physiological methods), pregnancy intention (intended, unintended), type of delivery (normal, cesarean), iron taken during pregnancy (no, yes) and maternal decision-making autonomy.
The household wealth index was used as an indicator to estimate the poverty-wealth status of households in the EDHS. It was calculated based on ownership of selected assets, such as televisions and bicycles, materials used for housing construction, and types of water source and sanitation facilities. The household population was then divided into five groups: first quintile (poorest), second quintile (poorer), third quintile (middle class), fourth quintile (richer), and fifth quintile (richest) [15]. In the current study, women were classified as poor if they belonged to the first or second wealth quintile, middle if they belonged to the third quantile, and rich if they belonged to fourth and fifth quintile.
Body mass index was defined as weight in kg divided by the square of height in m. Body mass index categories were underweight (<18.5), normal (18.5–24.9) and overweight/obese (≥25). Contraceptive methods that include pills, injections and implants were categorized as hormonal methods. IUD, condoms and female sterilization were classified as barrier methods. IUD was considered as a barrier method due to lack of data on the type of IUD (i.e. copper and hormonal IUDs). Physiological methods included periodic abstinence, withdrawal, lactational amenorrhea (LAM) and standard days method (SDM).
Women’s decision-making autonomy was measured by women’s participation in household decision making. Women were considered to participate in household decisions if they make decisions alone or jointly with their husband in all three of the following areas: (i) the woman’s own health care, (ii) major household purchases, and (iii) visits to the woman’s family or relatives [15].
Frequency distribution and descriptive statistics (mean ± standard deviation) were used to describe the characteristics of the study participants. Bivariate analysis using Rao-Scott Chi-Square test was performed to analyze the study participants according to their anemia status and high-risk fertility behavior. Three fully adjusted logistic regression models were created to analyze anemia status and each model contained a different fertility risk predictor (any vs. no risk; separate effects of single high-risk, multiple high-risk vs. no risk; specific types of high-risk vs. no risk).
Pairwise interactions between high-risk fertility behavior and each covariate were performed to examine differences in the relationship between high-risk fertility behavior and anemia status across different categories of covariates. There was a significant interaction between residence and high-risk fertility behavior on anemia status. Multivariable adjusted logistic regression models were performed to examine the independent association of high-risk fertility behavior on anemia status stratified by residence. All the covariates were entered simultaneously into all of the regression models.
Since the sampling design of EDHS was complex, sample weight (weighted frequency and weighted percentage) was used for all analysis to give accurate estimates for the population parameters. Statistical testing was performed using SAS University Edition for Windows (SAS Institute Inc., Cary, NC, USA.). Two-tailed p value < 0.05 was considered statistically significant.
Ethical considerations
The EDHS received government permission, used informed consent and assured participants of confidentiality. The data was downloaded after the purpose of the study was justified and approved by the DHS program. The current study was considered exempt from a full review because it was based on an anonymous public use of a secondary dataset with unidentifiable information on the survey participants.