Utilization of key preventive measures for pregnancy complications and malaria among women in Jimma Zone, Ethiopia

Background: In Ethiopia, malaria infections and other complications during pregnancy contribute to the high burden of maternal morbidity and mortality. Preventive measures are available, however little is known about the factors influencing the uptake of maternal health services and interventions by pregnant women in Ethiopia. Methods: We analyzed data from a community-based cross-sectional survey conducted in 2016 in three rural districts of Jimma Zone, Ethiopia, with 3784 women who had a pregnancy outcome in the year preceding the survey. We used multivariable logistic regression models accounting for clustering to identify the determinants of antenatal care (ANC) attendance and insecticide-treated net (ITN) ownership and use, and the prevalence and predictors of malaria infection among pregnant women. Results: Eighty-four percent of interviewed women reported receiving at least one ANC visit during their last pregnancy, while 47% reported attending four or more ANC visits. Common reasons for not attending ANC included women’s lack of awareness of its importance (48%), distance to health facility (23%) and unavailability of transportation (14%). Important determinants of ANC attendance included higher education level and wealth status, woman’s ability to make healthcare decisions, and pregnancy intendedness. An estimated 48% of women reported owning an ITN during their last pregnancy. Of these, 55% reported to have always slept under it during their last pregnancy. Analysis revealed that the odds of owning and using ITNs were respectively 2.07 (95% CI: 1.62-2.63) and 1.73 (95% CI: 1.32-2.27) times higher among women who attended at least one ANC visit. The self-reported prevalence of malaria infection during pregnancy was low (1.4%) across the three districts. We found that young, uneducated, and unemployed women presented higher odds of malaria infection during their last pregnancy. Conclusion: ANC and ITN uptake during pregnancy in Jimma Zone fall below the respective targets of 95 and 90% set in the Ethiopian Health Sector Transformation Plan for 2020, suggesting that more intensive programmatic efforts still need to be directed towards improving access to these health services. Reaching ANC non-users and ITN ownership and use as part of ANC services could be emphasized to address these gaps. For the purpose of this study, we used data from a community-based cross-sectional survey that was conducted from October 2016 to January 2017 as part of the baseline evaluation of a larger cluster-randomized controlled trial to address barriers to safe motherhood options in Jimma Zone, Ethiopia (ClinicalTrials.gov identifier: {“type”:”clinical-trial”,”attrs”:{“text”:”NCT03299491″,”term_id”:”NCT03299491″}}NCT03299491), as described by Ouedraogo et al. (2019) [12]. Only women who had a pregnancy outcome (i.e. live birth, stillbirth, assisted abortion, miscarriage) in the year preceding the survey were eligible to participate. It was determined that 24 primary health care units (PHCUs) or clusters, and a total of 3840 women would be required for the trial’s baseline evaluation. The sample size calculation for the trial was based on detecting a 17% difference in the primary outcome (skilled birth attendance) between the intervention and control arms. A two-stage sampling strategy was used. We first randomly selected 24 primary health care units (PHCUs) or clusters from the 26 available in the three study districts (Gomma, Kersa, and Seka Chekorsa). From each PHCU catchment area, we then randomly selected 160 eligible women to achieve the target sample size of 3840. Eligible women were randomly selected for face-to-face interviews using a random number generator after conducting a listing exercise with HEWs, village leaders and members of the health development army. We sought informed consent for each woman after explaining the purpose of the survey and the risks and benefits associated with participation. Of the eligible women, 1.5% women (N = 56) refused to participate. No replacement was made for women who refused to participate. Trained interviewers used computer tablets to administer the questionnaire in the local language at the woman’s household. If no woman was available for the interview on the first attempt, the household was revisited on a later date. If no respondent was available after two attempts, the household was replaced by a randomly selected alternate. Survey questions ascertained socio-demographic characteristics of the participants and their maternal health services usage before, during and after pregnancy. The survey was conducted in three Woredas or districts of Jimma Zone: Gomma, Kersa, and Seka Chekorsa. Jimma Zone is situated in Oromiya region, approximately 7 h west by road from Ethiopia’s capital Addis Ababa. Oromiya is one of the regions in Ethiopia where the utilization of maternal and child health services remains suboptimal. The three districts in the study area had a population of approximately 260,000 in 2016. Maternal health services are provided at 26 health centers and 110 rural health posts located in these districts. Nationally, the 2016 EDHS reported that 48% of women who had a live birth in the 5 years before the survey did not attend ANC services [3]. Skilled birth attendance among the 2016 EDHS participants was also low, with only 20% of live births in the 5 years preceding the survey taking place in a health facility [3]. According to the 2015 Ethiopia National Malaria Indicator Survey, 58% of households had at least one ITN and 42% of pregnant women slept under an ITN in Oromiya region [13]. The primary malaria species of epidemiological importance in Ethiopia are Plasmodium falciparum and P. vivax, accounting for approximately 70 and 30% of the malaria cases, respectively [7]. The risk of malaria infection depends on the altitude, with the greatest risk of infection occurring below 2000 m [7]. In Gomma district, the altitude ranges from 1380 to 1680 m, compared to 1740–2660 m in Kersa district and 1580–2560 m in Seka Chekorsa [14]. The three chosen areas have similar climatic conditions, with a main rainy season occurring from June to August [14]. A short rainy season is also observable in February and March [14]. The incidence of malaria infection is typically greatest during both the rainy seasons while lower infection is observed during the rest of the year [7]. To ascertain ANC attendance in the three districts, women were asked whether they attended ANC during their last pregnancy and how many times they visited a health facility for ANC. Participants who reported not attending ANC were asked why they did not attend. To determine ITN ownership and utilization among the recruited sample, women were asked whether their household owned any ITN during their last pregnancy and how frequently they slept under an ITN during their last pregnancy (never, sometimes, often, always). We re-categorized the utilization of ITN to capture women who always used an ITN during last pregnancy and those who did not always use an ITN (never, sometimes and often). Finally, women were asked to report whether they were diagnosed with malaria during their last pregnancy. We considered the following socio-demographic variables: age (15–18 years, 19–24 years, 25–34 years, and 35–49 years), marital status (not married, married), ethnic group (Oromo, Amhara, Others), employment status (not employed, self-employed, employed), level of education (no education, primary, secondary, higher), decision-making about health care (husband or family member, self, jointly with husband), exposure to different media sources (not at all, at least once a week, more than once a week), frequency of contact with HEWs (not at all, less than once a month, once or more times a month), household size (≤4, 5–8, ≥9 household members), children in the household (≤3, 4–6, ≥7), reproductive history (i.e. total number of live births, miscarriages, stillbirths, and neonatal death), and last pregnancy intendedness. For employment status, the ‘not employed’ category included individuals who identified as housewives, students and unemployed s, while those identifying as farmers were categorized as self-employed. Following the steps defined by the DHS group, we used a principal component analysis (PCA) to construct a household wealth index combining households’ ownership of various durable assets (e.g. electricity, radio, television, refrigerator, mobile phone)), housing construction (type of materials used for floor, roof and exterior wall), type of toilet facilities, sources of water supply, and type of fuel used for cooking [15]. PCA is a statistical method that reduces a specified number of variables into a smaller number of dimensions or principal components [16]. Briefly, PCA allowed us to extract a principal component from our selected variables as a measure of socio-economic status and derive factor scores for the considered variables. Using the factor scores as weights for each variable, an overall score for each household was subsequently obtained, which can be interpreted as the household-specific wealth score. Based on their wealth score, the households were then divided into five equal quintiles (poorest, poorer, middle, wealthier, wealthiest). All the data were synchronized to a cloud server using Open Data Kit and we exported CSV files for data management and analysis. We performed all analyses using SAS statistical software version 9.4. We assessed the distribution of each variable through frequency tables. We used chi-square and Fisher’s exact tests, as appropriate, to explore associations between the variables and determine differences in proportions. Differences in means were assessed using t-tests. We considered a two-sided p-value of 0.05 as the level of statistical significance. Using unadjusted and adjusted logistic regression models, we generated odds ratio (ORs) and corresponding 95% confidence intervals to identify significant predictors of first ANC visit attendance. We considered all the aforementioned socio-demographic characteristics (i.e. age, marital status, employment status, level of education, decision-making about health care, exposure to television and radio, frequency of contact with HEWs, reproductive history, and last pregnancy intendedness) as potential predictors of ANC attendance in univariable analyses, and retained all significant variables (p > 0.05) in the multivariable models. We conducted a sub-group analysis among women who attended any ANC visit, to identify whether the participants who attended at least four ANC visits differed from women who attended three or fewer ANC visits with regards to their socio-demographic characteristics. We used multivariable logistic regression analyses to assess the relationship between ANC attendance (no attendance, at least one visit) and ITN ownership and utilization. We considered age, marital and occupation status, education status, household wealth, and whether the dwelling was sprayed with insecticide in the last year as potential confounders and effect modifiers. Confounding was assessed by looking at the strength of the association between each variable, and ANC attendance and ITN ownership and utilization. Effect modification was investigated by introducing interaction terms between main predictors and potential effect modifiers into the regression models. We similarly assessed the main predictors of malaria infection during pregnancy using logistic regression models, considering socio-demographic factors, ITN utilization and IRS as potential predictors. For the analyses of ITN ownership and use, and malaria in pregnancy, we performed subgroup analyses considering only women living within the catchment area of PHCUs at an altitude below 2000 m, where the risk of malaria is known to be higher. Our analysis considered the complex cross-sectional survey sampling design, wherein women were clustered within PHCUs, as failure to do so can lead to incorrect inferences [17]. We therefore incorporated clustering of the data in all the analyses through logistic regressions with random intercept models, using the PROC GLIMMIX function in SAS 9.4. We also ran model diagnostics, including assessment of the distribution of the residuals and random effects.