Study area, study design, and study period
The community-based cross-sectional study design with focus group discussion (FGD) was conducted in Konso District from 1st December 2018 to 31st January 2019. Konso District is located 595 Kilometers from Addis Ababa (capital city of Ethiopia) and 365 km from Hawassa (the capital city of the south region). Based on the Konso district administrative population profile the district has a total population of 270,837 from this 132,710 male and female 138,127. The District has 50 health posts, 11 health centers, and 1 primary hospital, and 25 private clinics which are providing health services for the community.
Population
The source population of this study was all pregnant women in Konso District. The study population was all pregnant women in the selected Kebeles (small administrative unit in Ethiopia) of Konso District. All pregnant women who were residents of the Konso district for at least 6 months or above were included in the study whereas those pregnant women who were seriously ill, those who had hand deformity were excluded.
Sample size determination and sampling technique
The sample size was computed using single population proportion formula; considering 95% confidence level, 5% margin of error, 1.5 design effect, 31.8% hypothesized prevalence of undernutrition among pregnant women which has been taken from a study conducted in the central rift valley of Ethiopia [15] and by adding 5% none response rate the final sample size for this study was 527. A multi-stage stratified sampling which was followed by a systematic sampling technique was applied to reach each study participant. The kebeles in the district were stratified into rural and urban. One urban and eight rural Kebeles were selected by using the lottery method from a total of forty-one kebeles in the Konso district. Then by using health extension workers housing registration, the total number of households with pregnant women (2277) was accessed and the sampling interval was calculated. Finally, households with eligible pregnant women were selected using a systematic random sampling technique. For those households with more than one pregnant woman, one pregnant woman was selected by using the lottery method. Data collators visited the house on the next day when the pregnant women were not available at home and the pregnant women who were not available during the second visit were recorded as non-response, and then nearby household was considered.
Qualitative sampling procedure
Two FGD were conducted with 12 subjects in each group who were purposively selected. A situational analysis was done before conducting the focus group discussion (FGD) to minimize errors in the selection of participants. The key informants were a case team leader for maternal health and a supervisor assigned for each Kebele and agricultural expert (agriculture and natural resource head). In each FGD pregnant women, their husbands, and health extension workers participated. To minimize the possible bias in a selection of the study participants, we made sure to emphasize that we want a group of people that can express a range of views, to be able to have a proper discussion. A smooth discussion environment was created and we tried to encourage communication and interaction during the focus group discussion in every possible way. FGDs were held in a neutral setting which encourages participants to express their views freely. We made sure that there were no disturbances, adequate lighting, and ventilation as it was the hottest season during the data collection period, and also there were cold beverages including water. Materials that were necessary to conduct the focus group discussion (FGD) were prepared before a discussion like the FGD guideline, voice recorders, notebooks, pen, and pencils. To create a conducive discussion area, the chairs were arranged in a circle.
Study variables
The dependent variable in this study was undernutrition among pregnant women. The independent variables were: Socio-demographic factors:-age, marital status, husband education maternal education, family size, polygamy. Maternal related factor: – parity, family planning utilization before current pregnancy, birth interval, receiving iron supplementation, ANC follows up, ANC satisfaction, Nutrition knowledge, Illness, History of abortion, History of stillbirth, dietary diversification (24-h recall), meal frequency, Socio-Cultural factors: -food taboo and food restriction during pregnancy, decision making on household assets, family stable food. Economic factors:-Family source of food, farmland ownership, employment (maternal& husband job) status, household income, wealth index. Hygiene and sanitation-related factors:-access to water and sanitation facilities, such as latrine availability & utilization, family source of water, distance to get water. Food Security-related factors – food accessibility and availability.
Data collection tool and procedure
An interviewer-administered structured questionnaire was used to collect the data for the quantitative part of the study and qualitative data was collected using two focus group discussions (FGDs). The tool included: Socio-demographic factors adapted from the Ethiopian demographic and health survey (EDHS 2016) [36], Maternal related factor, Socio-Cultural factors, Economic factors, Hygiene and Sanitation related factors, and food Security related factors.
To determine the nutritional status of pregnant mothers, MUAC (mid-upper arm circumference) was used. Mid upper arm circumference of the left arm was measured triplicate using a non-stretchable standard MUAC tape to the nearest 0.1 cm with no clothing on the arm. The mean of triplicate measurement was taken. The value of MUAC below 23 cm was considered as undernourished and MUAC ≥23 cm was considered as normal nutritional status [34, 37].
Dietary diversity information of individual respondents was collected using the 24-h recall method and women dietary diversity score model questionaries’ of nine food groups with food listing method in which list of food items replaced by common foods in local context included in the questioner [38].
Household food security was measured by the household food insecurity access scale (HFIAS) which is an adopted approach in estimating the prevalence of food insecurity in the united states (USA) and was used to estimate the food insecurity among study participants [39]. HFIAS prevalence indicator categorizes households into four levels of food security as food secure, mildly, moderately, and severely food insecure [25]. HFIAS yes or no questions were used to collect information on the food security status of the household followed by the occurrence of the situation if the response is yes [39]. For the occurrence of once or twice, it was recorded as rarely if the occurrence is 3–10 times it was categorized as sometimes and if the occurrence was more than ten times in the past 4 weeks it was categorized as often [26, 40].
Purposively selected subjects participated in the two focused group discussions for qualitative data. The composition of focus group discussion participants was, pregnant mothers, elders or mother’s in-low, and their husbands, health extension workers were engaged in the discussion separately to facilitate the expression of opinions without fear and a key informant interview was held with the health department head and agricultural expert (agriculture and natural resource head). The discussion was conducted at community meeting places, and the information was collected using open-ended questions. Note-taking and tape recording were used to document the appropriate information and detect redundant responses. Identified redundant responses were considered to be saturated and removed every evening after triangulation the day work during preliminary analysis.
Data quality assurance
To assure data quality, training was given to data collectors and supervisors before data collection. The data collection tool was pre-tested in 5% of the sample size. The pretest was conducted on individuals having similar characteristics of the study in Kebele which was not selected in this study. After the pre-test, the instrument was modified accordingly. Supervisors supervised the data collection process and checked the completeness of the questionnaire daily. Besides, principal investigators carefully cleaned the data and entered collected data into computer software.
Data processing and analysis
Epi-data version 3.1 statistical software was used for data entry and exported to SPSS version 21 for analysis. Descriptive statistics like mean, standard deviation, frequencies, and percentages were computed. Bivariable and multivariable logistic regression was used to determine the degree of association between independent and dependent variables. All variables with a p-value less than 0.2 in the bivariable analysis were entered into a multivariable logistic regression. The presence of an association between dependent and independent variables was checked with an adjusted odds ratio with 95% confidence intervals. Then the statistical significance was declared at a p-value less than 0.05 and adjusted odds ratio interval which excluded one. Assumption of logistic regression such as; meaningful coding, multicollinearity, and outliers checking was done before logistic regression model analysis. Multicollinearity was also checked by using Variance inflation factors and Tolerance test. The Hosmer-Lemeshow tests were checked to assess the goodness-of-fit model and it was a good fit(p-value > 0.05). The wealth index of individual respondent families was also analyzed by using principal component analysis.
