Study design and setting
A case-control study was conducted in two purposively selected districts (Kersa and Omo Nada) of the Jimma Zone, Oromia Regional State, Ethiopia, from December 2018 to January 2019. The Zonal capital, Jimma town, is located in southwest Ethiopia, 357 km away from Addis Ababa. The Zone extends between 7013’ – 8056’ North latitudes and 35049 -38038’ East longitudes and the altitude ranges from 1740 to 2660 above sea level. Agriculture is the major economic activity and it includes mainly growing coffee and cattle rearing. According to the Jimma Zonal Health Office annual report (2019), the populations of Kersa and Omo Nada were 227,959 and 208,517, respectively. About 81.65% of residents of the Kersa district and 71.7% residents of the Omo Nada district rely on improved drinking water sources in 2018. Currently, the improved latrine coverage of the districts is 40% for Kersa and 39% for Omo Nada 27.
Source population
All children from 6 to 59 months, living in Kersa and Omo Nada districts with Z-scores of weight-for-height (wasting) < -2 (SD), weight-for-age (underweight) < -2 SD, height-for-age (stunting) < -2 SD, mid-upper arm circumference (MUAC) less than 11.5 cm or have edema were the source population of cases, whereas all children from 6 to 59 months with Z-scores of weight-for-height from -2 to +2SD, weight-for-age from -2 to +2SD, height-for-age from -2 to +2SD, or MUAC above 12 cm or have no edema based on growth reference of WHO28 were the source population of controls.
Inclusion and exclusion criteria
All children aged 6-59 months who had malnutrition and whose mothers resided in the study districts for at least a year were included as cases. All children aged 6-59 months of age and had no malnutrition based on the WHO growth reference 28 and whose mothers resided in the studied districts for at least a year were included in the control group. Children whose mothers were seriously ill and could not communicate information were excluded from the study.
Sample size determination and sampling procedure
The sample size was determined using a formula for calculating the double population proportion by assuming estimates of the proportion of well-nourished children (P2) as 43.8% (taking water sources as a major factor29), α=1.96 at 95% confidence level, odds ratio 1.89; (from literature, children living in households that had been using unprotected water sources were 1.89 times more likely to be acutely malnourished than those who had been using protected water sources 30), power: 80% (0.84), the ratio of cases to controls was 1:2. This implies that 117 cases and 234 controls were required. After adding 10% for the non-response rate, the final samples were 128 for cases and 256 for controls.
Five health centers, which have centers for the treatment of undernourished children, were selected purposively from the two districts. Then, three kebeles (the smallest administrative unit) were selected from each district based on the level of malnutrition reported as high, medium and low kebeles. Census was conducted to identify the number of under-five children and their nutritional status in the kebeles. Then, the cases were randomly selected from the malnourished children and controls were randomly selected from well-nourished children.
Anthropometric measurements
Bodyweight, length/height, mid-upper arm circumference and presence of edema of the children were measured based on the WHO references 28. The weight and height of the children were measured using the Salter scale and measuring board, respectively. The body weight of all children was measured without shoes to the nearest 0.1g, whereas the height/length of children was measured to the nearest 0.1 cm. Each measurement was done twice, and the mean of the two readings was recorded. The weighing scale was calibrated regularly with a known weight. The ace scales' indicator was checked against zero reading after weighing every child. To convert raw anthropometric data (weight, height, and age of children) into an anthropometric Z-score (weight-for-age, height-for-age, and weight-for-height), emergency nutrition assessment (ENA) for standardized monitoring and assessment of relief and transition (SMART) was used. Thumb pressure was applied to the upper side of both feet for three seconds to diagnose the presence of edema. The presence was diagnosed if a bilateral depression (pitting) remained after the press release. Mid-upper arm circumference was measured in centimeters using MUAC tape on the left arm and was recorded to the nearest 0.1 cm. The nutritional status of children was identified as case or control using cutoff points recommended by the World Health Organization based on the Z-score, edema, and MUAC values28. Confirmed malnourished children were linked to the health center after consultation with the data collector.
Data collection instrument
Data were collected by interviewing mothers/caretakers of the children using a pretested questionnaire. The questionnaire was adapted from the WHO/UNICEF Joint Monitoring Program for Water Supply, Sanitation, and Hygiene 2017, core questions on water, sanitation, and hygiene for the household survey 31. Some questions were revised to suit the context of the study by the principal investigator.
The questionnaire consisted of variables related to socio-demographic, child, water supply, sanitation, and hygiene practices. The wealth index was developed from assets and other housing characteristics. Handwashing at critical times was assessed through information about their handwashing behavior after defecation, before handling food/water, before feeding a child, or after cleaning the child stool. If they responded in the affirmative about these critical times of handwashing, we gave the remark; always; if at least one was missed, we gave the remark; sometimes. Mothers/caregivers were asked about the ages of their children or this information was collected from the immunization cards if present. If they did not know the age or did not have immunization cards, data collectors asked them whether the child was born before or after known holidays and/or local market days. They were also asked about any occurrence of diarrhea and vaccination status based on the age of the child to identify the past two weeks of diarrhea and vaccination of children.
Data collection and quality
The data were collected by health professionals through face-to-face interviews with mothers/caregivers. The questionnaire used for this data collection was originally prepared in English and then translated into the local language (Afan Oromo) and back retranslated into English to check its consistency by public health and linguistics professionals. Then, the necessary correction and modification of the instrument were made.
The mother’s interview and anthropometric measurements were done by data collectors following two days of intensive training, which included orientation, demonstration, and field procedures. Pretest of the instrument and the procedure was conducted on 5% of mothers or caregivers of the children in the selected households before actual data collection. Anthropometric measurements were done by using calibrated and pretested scales. The overall day-to-day data collection process and completeness of the collected questionnaires; was checked and any other amendments were made by supervisors and the principal investigator.
Data analyses
Data were entered, cleaned, and checked for correctness using EpiData version 4.2. After exportation, all statistical analyses were carried out using SPSS version 24. Data were described by frequency, percentage, and mean (for continuous data) to compare the cases and controls. The wealth status of the household was computed from the household’s asset ownership and housing characteristics using principal component analysis32. It was categorized as poor, middle, or rich. The logistic regression model was used to assess whether water supply, sanitation, and hygiene practices are associated with childhood undernutrition. The crude odds ratio (COR) and 95% CI were used to identify the unadjusted strength of association between independent and dependent variables in bivariate analysis. To adjust for confounders, multivariable analysis was used. All variables that had a p-value of 0.25 or less in the bivariate analysis were included in the multivariable analysis after multicollinearity among variables was assessed by calculating the variance inflation factor (VIF). Adjusted odds ratios (AOR) with 95% CI were computed to assess the strength of the association and a p-value < 0.05 was used to declare statistical significance in the multivariable analysis.