Food insecurity associated with attendance to antenatal care among pregnant women: Findings from a community-based cross-sectional study in Southern Ethiopia

Purpose: Enrollment to antenatal care (ANC) is still not universal in Ethiopia. This study examines whether household food insecurity affects antenatal care attendance or not, as well as other factors associated with antenatal care. As optimal antenatal care is vital for the improvement of maternal and child health, the study will contribute to the efforts in improving maternal and child health. Patients and Methods: A community-based cross-sectional study was conducted among 707 pregnant women at or above 3 months of self-reported pregnancy in Southern Ethiopia. Multi-stage sampling was employed to obtain the study units. Data were collected using an interviewer-administered structured questionnaire. Logistic regression analysis was conducted to identify the independent factors associated with study outcome. Results: Out of a total of 707 study subjects, the majority (71%) of the study women visited a health facility for ANC service. The odds of ANC use was lower for women who were not in marital union (adjusted odds ratio (AOR)=0.39, 95% confidence interval (CI)=0.16–0.97), and those from food insecure households (AOR=0.50, 95% CI=0.32–0.79). ANC attendance was higher for women from high socio-economic status (AOR=2.62, 95% CI=1.29–5.29), with planned pregnancy (AOR=1.82, 95% CI=1.16–2.85) and a perceived risk from danger signs (AOR=4.32, 95% CI=1.60–11.67). Conclusion: While the overall ANC use was high, women experiencing food insecurity and those with unplanned pregnancy were having lower odds of ANC attendance among others. Interventions targeting at enhancing women’s attendance to ANC service might be realized through commitment from the agriculture, economic, as well as health sectors by increasing productivity and providing special attention to women in the pre-pregnancy and pregnancy period. Moreover, educating women so that they can recognize that every pregnancy is risky and promotion of family planning to reduce unplanned pregnancy could improve attendance to the ANC service. The study was conducted in Arba Minch Zuria woreda (woreda: an administrative unit corresponding to district in other parts of the world), Gamo Gofa Zone, South Ethiopia in March 2015. The woreda has 29 rural kebeles (smallest administrative units in the woredas in Ethiopia). Based on projection for the year 2014/2015 from the 2007 Ethiopian national census, the total population of the woreda was 205,204, with an estimated number of pregnant women equal to 7,101. At the time of the study, the woreda had six health centers and 37 health posts. Located at 505 km South of Addis Ababa, the capital city of Ethiopia, Arba Minch was the main town of both the Woreda and Gamo Gofa zone.37 The study employed a cross-sectional study design among pregnant women at or above 3 months of pregnancy in a community setting. Women reported their pregnancy status and months of gestation by themselves. The study participants were selected randomly from nine kebeles in the woreda. Those women who were not able to provide information either for serious acute illness or any other disability were excluded from the study for fear of not getting valid information. Likewise, those who did not live in the study area for at least 6 months were also excluded from participation in the study. This study is part of a previous study on 707 women38 from a thematic project on maternal and child healthcare utilization and food insecurity in Southern Ethiopia. The sample of women considered in the study was larger than sample sizes calculated using open epi software from ANC use estimates of two previous studies (86.3% and 28.5%) in the same region.17,22 The assumptions considered were 95% confidence, a pregnant population of 7,101, a design effect of 2, and a margin of error of 5%. There are 29 rural kebeles in Arba Minch Zuria Woreda.37 The estimated number of pregnant women for each kebele was obtained by multiplying the total population of the kebeles by 3.46% (conversion factor for estimated number of pregnant women) in Southern Nations Nationalities and Peoples Region (SNNPR). The study subjects were selected using a multi-stage sampling technique. In the first stage, nine kebeles were selected from the 29 kebeles randomly. Proportional allocation of the sample was done to each kebele based on the number of pregnant women available in the kebele. A list of pregnant women was prepared using the current data from the monthly updated family folders in the health posts of each kebele. Secondly, the required numbers of pregnant women were selected randomly from the list in each kebele by using a SPSS generated random number of codes pre-assigned to each pregnant woman in the list. (Family folder: A service delivery tool or registration book at household level that contains members’ health and health related information and household characteristics, irrespective of their relationship in the household and marital status). Pregnancy identification was based on self-reported data from women themselves. Identification of pregnancy is made initially by leaders of the smallest 1–5 network of households at grass root level in their locale and continually reported to leaders of 30 households in the women development army (HDA) and then to the health extension workers (HEWs) in the health posts. While reports are collected in every contact of the HEWs with the HDA leaders, it is updated and compiled on a monthly basis. After receiving a day long intensive training on the data collection process and instruments, seven public health nurses and two health officers collected the data and supervised the data collection, respectively. The questionnaire used for data collection was structured and administered by interviewers after being pre-tested in a kebele not selected for the study. ANC use was assessed using a questionnaire adapted from literature on ANC service attendance,16–25 and the Household Food Insecurity Access Scale (HFIAS) questionnaire by the Food and Nutrition Technical Assistance (FANTA III) was used to assess the household food insecurity status of women [39]. Health Development Army (HDA) leaders guided the data collectors and supervisors to the respective kebeles as well as households’ of the selected women. The list of pregnant women selected for the study with their respective kebeles was provided in advance to the data collectors. The questionnaire initially prepared in English was translated to the local language Gamotho by an expert in the language and back translation was done to check its consistency with the original meaning by another expert having a very good skill in both languages. Data was collected using a structured Gammotho version questionnaire. Before commencing data collection, a pre-test was conducted in Chano Dorga kebele (one of the kebeles in the woreda that was not selected for the study) on 5% of the samples or 36 respondents. Based on the pretest, difficult or confusing items were revised. Each woman was interviewed in a separate private place to avoid social desirability bias. A completeness and update check was made on 30% of randomly picked samples of folders in each kebele, and the data collection period was set in line with the recent update. Attended ANC service: when a woman had at least one visit to a health institution for checkup purpose during the current pregnancy. Knowledgeable on danger signs: women who spontaneously mentioned two of the three key pregnancy danger signs (vaginal bleeding, swollen hands/face, and blurred vision) or three of the four key labor and delivery danger signs (severe vaginal bleeding, prolonged labor (>12 hours), convulsions, and retained placenta), or two of the three key postpartum danger signs (foul-smelling vaginal discharge, severe vaginal bleeding, and high fever). Perceived risk from danger signs: Women who knew at least one of the key danger signs during pregnancy, labor, and delivery and the postpartum period and perceive that any of them could threaten their life (kill them or cause illness to them). A household was classified as mildly food insecure if members of the household sometimes or often worry about not having enough food, and/or are not able to eat preferred foods, and/or eat a more monotonous diet than desired and/or some foods considered undesirable, but only rarely in the last 4 weeks (30 days).39 A household was classified as moderately food insecure if members of the household have started to reduce the number of meals or the size of meals in order to cut back on quantity, and/or sometimes or often eat undesirable foods or a monotonous diet, rarely or sometimes in the last 4 weeks (30 days).39 A household was classified as severely food insecure if members of the household have experienced any of the three conditions (going a whole day and night without eating, going to bed hungry, or running out of food) even once in the last 4 weeks (30 days).39 A household was considered as food insecure when it has any of the food insecurity conditions mentioned above (mild, moderate, or severe food insecure), otherwise it was classified as food secure or when it just experiences worry, but rarely.39 Epidata version 3.1 and IBM SPSS statistics 20 were used for data entry and analysis, respectively. Findings were described using frequencies, percentages, means, and standard deviations. Binary logistic regression analysis was done in steps to identify independent predictors of ANC attendance. Initially, bivariate analysis was done to select candidate variables for multivariable regression analysis. Independent variables having an association with the outcome variable at a P-value<0.2540 and with public health importance based on prior strong significance were selected as candidate variables to fit the final multivariable analysis model. Variance Inflation Factor >10 was used as a criterion to check the existence of multicollinearity between the candidate variables. Lastly, the final multivariable analyses model was fitted between the candidate independent variables and the outcome variable in order to control for possible confounding and determine the presence of statistically significant association. The fitness of the final model was checked using Hosmer and Lemeshow goodness of fitness test. A backward elimination method was used to select the variables in the regression model. Statistical significance was declared at a P-value of <0.05, and the degree of association between independent variables and the outcome variable was measured by AOR with its 95% CI. Seventeen explanatory variables (women’s age, educational status, marital status, occupation, habit of listening to the radio, household wealth and food insecurity status, gravidity, parity, plan of pregnancy, age at first pregnancy, place of previous birth, history of still birth, history of abortion/miscarriage, knowledge on danger signs and perceived risk from danger signs, partner’s educational status) were fitted to bivariate logistic regression model to select potential candidates for multivariable analysis. Women’s age, gravida, age at first pregnancy, and partner’s educational status did not fulfill the selection criteria and hence were not selected for multivariable analysis. A household wealth/economic status of the study women was set from 13 items on common permanent assets in the study area using a principal component analysis method of data reduction. The items were removed at various iteration steps after checking the fulfillment of assumptions for principal component analysis using Kaiser-Meyer-Olkin measure of sampling adequacy and Bartlett’s Test of Sphericity.