Study location
The study was conducted in Malawi, a small landlocked country of about 118,484 km² in Sub Sahara Africa. Malawi is bordered by Mozambique to the South and East, Zambia to the West, and Tanzania to the North. The current population of Malawi is about 17,563,749 million and nearly 85% depend on agriculture for their livelihood (12). This research was conducted in six outreach clinics in rural communities that are within the radius of 5 to 10 kilometers away from Mvera Mission Hospital in Dowa district (Fig. 1) during the months of May to September 2018.
Figure 1. The research setting
>>>>>>Insert Fig. 1 here<<<<<<
Study sample
I employed a descriptive cross-sectional study design. A multi-stage cluster sampling technique was used to select representative research participants (13). Specifically, I randomly selected six out of eight outreach postnatal clinics that operate under Mvera Mission Hospital. The selected outreach clinics were Mkhalanjoka, Gogo, Mvera, Kalinyengo, Mphande, and Ching’amba. During the time of the study, Mvera hospital was serving a population of 27,719 people. Out of the total population, 4820 under-five year old child/mother dyads were clients of postnatal health services in the selected six-outreach clinics. A Raosoft online software program was used to calculate a sample size (14).The margin of error was set at 5%, with 95% confidence level, and a response distribution of 50%. The minimum sample size was found to be 356 under five-years-old child/mother dyads. I chose to increase my sample size to 538 dyads with an aim of strengthening the study’s reliability and reducing the margin of error from 5–4%. A systematic sampling technique was used to select 538 out of 4,820 under-five child/mother dyads from postnatal registers by selecting every ninth pair starting with a randomly selected pair.
Participant recruitment
I contacted the selected children and their mothers through their postnatal clinics when they were attending a regular monthly health assessment. Health workers asked the mothers if they would be interested in participating in the study following their health assessment in a private consultation room. Five mothers did not consent to take part in the study, and they were subsequently replaced by randomly selected child-mother dyads that were not chosen during the initial sampling stage. For those who consented, a research assistant who was also a health worker orally administered the questionnaire in the same consultation room, away from the other health workers and clients. Through this set-up, the research team was able to protect the confidentiality of children and their mothers.
Measures
Outcome variables
The primary outcome variable of this study was child undernutrition in the form of stunting, underweight, and wasting. Stunting refers to a chronic form of undernutrition that entails prolonged periods of insufficient nutrient intake and assimilation (15). In general, stunting is an indicator of overall community social-economic wellbeing, where there is enough accessible food for the child to consume (16). Wasting refers to an acute form of undernutrition and entails a deficit of body tissue and fat mass. The condition arises when a child fails to access adequate food nutrients within a short period of time (15). Underweight is a composite measure of child undernutrition that encompasses both stunting and wasting. SECA gmbh & co. Model 874 mother-infant weighing scale was used to measure and collect the weight of the children and mothers. All standard practices for weighing children were followed, which included children wearing minimal clothing and no shoes when they were weighed (17). For accuracy and consistency, a 2 kg weight was used to adjust the measuring scale to zero after weighing each mother-child pair. The results of child and maternal weight were calculated to the nearest 100 grams. Length measuring boards were used to take children’s height and were recorded to the nearest 0.1 cm (18). Children who were two or more years old were measured in a standing position while those that were less than two years old were measured in a recumbent position.
I calculated height-for-age, weight-for-height, and weight-for-age Z-score values from measured children’s heights, weights, age in months and reported child sex. A WHO Anthro software version 3.2.2, January, 2011 (19) was used to calculate the Z-scores with reference to the WHO child growth standards (20). Child stunting, underweight, and wasting were denoted as those that were less than or equal to 2 standard deviations of height-for-age, weight-for age, and weight-for-height Z-scores (18).
Independent variables
Based on UNICEF’s “care for nutrition” conceptual framework, I included numerous explanatory variables that have been linked to child undernutrition in developing countries (21). The following immediate child risk factors were considered as potential explanatory variables. Normal childbirth weight (≥ 2.5 kg) was coded zero, and low birth weight (˂ 2.5 kg) was recorded = one. Common child morbidities in the study areas were diarrhea, malaria, and acute respiratory infection (ARI). Mothered were asked to report the common sicknesses that the child suffered from in the two weeks prior to the research. Mothers’ reported child morbidity was ascertained in the child health passport book as examined by medical personnel on the day that corresponded with the survey. I coded 1 = diagnosed/reported sicknesses and 0 = no sickness in the past two weeks. Child gender was coded as 0 = female, and 1 = male. Children who were treated with deworming drugs such as albendazole within the past year were coded = 1 and those who did not receive the treatment were coded = 0. Age of children were categorized as 1 = 1–5 months, 2 = 6–11 months, 3 = 12–23 months, and 4 = 24–59 months old.
The first underlying risk factors of child undernutrition were maternal demographic characteristics. They included age which was coded as 1 = 16–24 years old, 2 = 25–34, and 3 = 35–49. Maternal education was categorized as 0 = no education, 1 = primary school, and 2 = secondary school. Maternal faith was categorized as 1 = Church of Central African Presbyterian (CCAP), 2 = Catholic, and 3 = Pentecostal. The number of children that a mother gave birth to were coded as 1 = one, 2 = two, 3 = three, 4 = four, and 5 = five and more. I also considered whether the pregnancy for the child participant was planned or unplanned. I coded 0 = planned, and 1 = unplanned. Mothers’ nutritional status was categorized as 1 = underweight (BMI < 18.5 kg/m2), 2 = normal (BMI 18.5–24.9 kg/m2), 3 = overweight (25.0–29.9 kg/m2), and 4 = obesity (≥ 30.0 kg/m2) (18). Mothers who received nutritional counselling during pregnancy were coded 1 = yes and 0 = no. I also assessed whether a mother was exposed to IPV which was coded as 1 = yes and 0 = no. IPV was assessed by using a WHO Multi-country study questionnaire on women’s health and life experiences that was previously validated and administered in Malawi (22,23).
The second underlying risk factors of child undernutrition were attributed to the father. Amongst them was education which was coded 0 = no formal education, 1 = primary, and 2 = secondary and tertiary education. Age was coded as 1 = 15–24 years old, 2 = 25–34, and 3 = 35–49. We included fathers’ risky health behaviours such as alcohol consumption, smoking, whether he was involved in polygamous family, infidelity, and/or divorced. Each risk behavior was coded as 0 = no and 1 = yes.
Poverty threshold as a household level underlying risk factor for child undernutrition was defined based on the international poverty headcount ratio of USD 1.90 a day (24). Household food security was measured by a Household Food Insecurity Access Scale (HFIAS), which is calculated based on responses to questions about the frequency of occurrence of nine experiences characteristic of food security in the past month (25). Whereas the HFIAS yields four categories of food security, I coalesced the four HFIAS categories into food secure (food secure and mildly food insecure) and food insecure (moderately and severely food insecure) households to produce a binary set of categories (0 = food secure and 1 = food insecure household). The nutritional quality of diets was measured by Household Dietary Diversity Scale (HDDS) (18,26). I coded household as 1 = inadequate household dietary diversity when the mother reported that they consumed four or less food groups in the past 24 hours and 0 = adequate household dietary diversity when a mother reported that they consumed five or more food groups in the past 24 hours (18,27). I also inquired about household domestic water source. I coded 1 = borehole (potable water) and 0 = river/wells (unsafe water). I also considered time that the mother took to fetch a pail of water. This was coded as 0 = ˂ 30 minutes, and 1 = ≥ 30 minutes. I considered child delivery place as hospital = 1, and home or traditional birth attendant = 0. I did not include toilet facility and household type because almost 99% and 98% of participants reported having an open pit latrine and grass thatched houses respectively.
The survey was administered on Android tablets. The Open Data Kit (ODK) was used to upload the digital version of the questionnaire into the tablets. An ODK is an Android application that can administer surveys, collect, and organize the survey data (28). This application allows for immediate data validation in the field. My team (the research trainers) included a clinical officer, a medical doctor, and a PhD social work candidate. Nine married female health surveillance assistants (research assistants) that were trained using the WHO protocol for conducting studies of IPV were responsible for data collection (29,30). It took five days to train the research assistants and to pretest the survey questionnaire. The research assistants and trainers all had professional training in child and maternal health. The researcher stakeholders also had vast experience in measuring child weight and height and diagnosing common child illnesses in Malawi. Since I examined a sensitive topic relating to IPV, stakeholders decided that the questionnaire should be administered in the clinic’s consultation room (31,32). The procedure for recruitment was as follows: when a mother and her child completed their monthly postnatal primary healthcare check-up, a health worker on-duty informed the mother about my nutritional study. If the mother was interested to take part in the research, the health worker invited the survey enumerator to administer the questionnaire. The arrangement was agreed upon by the survey team in consultation with hospital officials in order to maximize confidentiality of the research participants. Research assistants explained the objectives of the research and when the mother agreed to participate, she was requested to affirm her consent verbally.
Research Ethics Review
I obtained ethics approval to conduct this study from McGill University’s Research Ethics Board in Canada protocol number (REB File #: 503–0518), and University of Livingstonia research committee in Malawi protocol number (UNILIA-REC-4/18). I also obtained written permission from the authorities at the Dowa district Commissioner’s office, the Dowa District Health Office, and Mvera Mission Hospital. Oral consent was also obtained from local health leaders.
Data analysis
All 538 selected child-mother pairs were assessed for nutritional status. Cronbach’s α analysis was conducted to calculate the internal reliability of the questionnaire items that were used to determine HFS, dietary diversity, and IPV (33). The HFIAS’ 9 items ranged in severity from worrying about household access to food to going the whole day and night without consuming any food due to lack of resources (34,35). The HDDS had 12 food categories which were used to probe the type of food that household members consumed in the past 24 hours. The scales on IPV controlling behavior had five items, psychological abuse had four items, physical abuse had six items, and sexual abuse had three items. I considered an α level of 0.70 or higher as satisfactory (33). The calculated Cronbach’s α for the HFIAS was 0.87, 0.83 for the HDDS, while the Cronbach’s α for the controlling behavior items was 0.81, psychological violence was 0.75, physical violence was 0.83, and sexual violence was 0.87.
I used the Kolmogorov-Smirnov test to determine the normality of the distribution of numerical variables that included age, number of children, and HFIAS. It was found that my data was not normally distributed and opted to construct numerical variables into categories according to standard procedures (36). I calculated the determinants of stunting, wasting, and underweight based on socio-demographic characteristics of children, mothers, fathers, and household factors. A chi-square test was used to examine the association between indicators of each form of child undernutrition and given explanatory variables. Univariate logistic regressions were also performed to determine significant risk factors of child undernutrition from the selected independent variables. Three separate multivariable logistic regression analyses were performed to explore predictors of child stunting, underweight, and wasting. Variables that were significant at a bivariate level were entered into the final multivariable logistic regression models using forward method.
Multicollinearity of explanatory variables was tested and a variance inflation factor (VIF) of 5.342 was obtained. The VIF was less than ten and I was confident that the included independent variables were not similar. Therefore, my regression coefficients estimates were reliable (37). I have reported the results of each of the three-child undernutrition multivariable analyses models as crude and adjusted odds ratios (AORs) with a 95% confidence interval (CI). A p value was considered statistically significant when it was less than 0.05. An IBM Statistical Package of Social Sciences (SPSS) for Windows version 23.0 (IBM Corp., Armonk, NY, USA) was used to analyze the data.