Study design and setting
This was a community-based cross-sectional study, carried out over the 6 months period between March and June 2021 in the Buea health district. Buea is the capital city of the Southwest Region of Cameroon, one of the two English speaking regions of the country, situated in the eastern slope of mount Cameroon. The town occupies a surface area of 7000 km2 with a population of approximately 1,481,433 inhabitants. Most inhabitants practice agriculture as the main economic activity. From a heath perspective, the Buea health district is one of the 18 health districts of the Southwest Region of Cameroon, consisting of 25 health facilities (including a regional hospital, serving as one of the two referral hospitals of the region) spread out over seven health areas (Bokova, Molyko, Muea, Bokwango, Buea Road, Bova and Buea Town) for an estimated population of 168,366 inhabitants. The main diseases among children in Buea are anaemia, diarrhoea, parasitic infections, malaria, typhoid, and undernutrition [7].
Study population and participant selection
The study targeted children of both sexes between the ages of 6–59 months living within the Buea health district. Children were recruited into the study if there were aged 6–59 months and lived within the Buea health district, and there was at least one adult aged 18 years or more to provide consent. Excluded from the study were children with a health condition (e.g., lumbar scoliosis) that could falsify anthropometric measurements or children whose caretaker denied consent. Caretakers selected for interview were preferentially the mother. In case the mother was not available, the father or other adult (aunt, grandparent etc.) directly responsible for the child was interviewed.
Sampling
Sample size
A minimum of 278 participants calculated using the formula: n = \({Z}^2\times \frac{P\left(1-P\right)}{d^2}\times k\) [8], were required for the study, where: n = minimum sample size, z = confidence value = 1.96 for a 95% confidence interval, p = estimated prevalence of childhood malnutrition from a study done in a similar crisis setting = 9.0%% [9], k = design effect = 2, d = error margin = 0.05 and 10% attrition added.
Sampling technique
Surveyed households were selected using a multi-stage sampling technique. First, the health areas that make up the Buea health district were considered, and three of them (Bova, Bokwango, Buea Road) were selected by simple random sampling. Next, geographically accessible households with children 6–59 months were identified and surveyed for each health area. In each selected household, the mother, father or any adult present in the house at the time of the survey was then interviewed. In the event where consent was denied from a household head/ caretaker, the data collectors continued to the next eligible household. The number of households surveyed in each selected health area was proportionate to the estimated number of eligible households within the health area. In case, an eligible households with 2 or more eligible children, one child was selected randomly by ballot.
Data collection
Data collection tool
Data was collected via kobo collect on android phones, using a validated structured questionnaire designed as a kobo collect form. The questionnaire captured information on the socio-demographic characteristics of both the child and caretaker; water, sanitation and hygiene practices of participants and the household; dietary diversity of the children 24 hours prior to the survey using the dietary diversity questionnaire [10]; household food insecurity, assessed using the household food insecurity access scale (HFIAS) [11], a nine questions tool used to distinguish food insecure from food secure households, and to estimate the prevalence of household food insecurity; and the medical history of the children/caretakers with particular focus on chronic diseases such as HIV/AIDS that could influence their nutritional statuses. The questionnaire was pretested in the Molyko health area to ensure clarity of language, appropriateness, and sufficiency. This allowed for adjustments and corrections to be made as necessary before effectively beginning the data collection process.
Measurement of variables
Anthropometric parameters
Height was measured using a UNICEF height board to the nearest 0.1 cm, following standard procedures to ensure readings were accurate. Children aged 24 months and younger were measured lying down with infant’s head against the top of the headboard of the infantometer (recumbent length), while those older than 24 months old were measured standing up straight (height) with the child’s buttocks, shoulder blades, and heels together touching the back of the stadiometer. Weight was measured using a battery powered portable Seca 216 digital floor scale to the nearest 0.1 kg. At the beginning of each day, scales were calibrated with a standard 5 kg weight and validated as accurate before use. For children younger than 24 months or those older than 24 months who were unable to stand, tared weighing was done. For children 24 months or older who could stand still, the child was weighed alone. Mid upper arm circumference (MUAC) was measured using a colour coded MUAC tape to the nearest 0.1 cm following standardized procedures to ensure accuracy [12]. The height, weight and MUAC anthropometric components were standardized. All measurements were done twice by the same study personnel and the average taken. If the two measurements were not within 2 units (0.2 kg for weight and 0.2 cm for height and MUAC), the measurer was instructed to repeat the measurement until there were at least two measurements within 2 units.
Undernutrition
Stunting and underweight were defined as Length/height-for-age ≤ − 2 standard deviations (SD) of the median, and weight-for-age ≤ − 2 SD respectively. Wasting was defined as either a weight-for-height Z score ≤ − 2 SD or a MUAC ≤12.5 cm [2].
Dietary diversity
Food items consumed by the children 24 hours prior to the survey were recorded and grouped into the seven essential food groups for children as recommended by the World Health Organization (WHO) notably breast milk, cereals and tubers, legumes and nuts, dairy products, flesh foods (meats/fish/poultry), eggs, vitamin A-rich fruits and vegetables, other fruits and vegetables [13]. A child was considered to have consumed a particular food group if they consumed at least one food item from the food group. Each food group was scored 1 if consumed by the child and 0 if not. The dietary diversity score (DDS) was then computed for each child by adding up all the 1’s from the different food groups consumed by the child. The total DDS ranged from a minimum of 0 (the child consumed none of the food groups) to a maximum of 7 (the child consumed all the food groups). Children who consumed at least four of the seven food groups (DDS ≥ 4) were considered to meet the minimum dietary diversity requirements, while those with a DDS < 4 were on the other hand considered to have poor dietary diversity [13].
Household food insecurity
Household food insecurity (HFI) was assessed using the Household Food Insecurity Access Scale (HFIAS) for the 4 weeks period preceding the survey. To obtain the HFIAS score, the answer to each HFIAS question was coded as follows: If the respondents answer to a question was ‘no’, the answer to that question was coded as ‘0’. In case the respondents answer to a question was ‘yes’, the answer was coded based on the frequency reported by the respondent as 1 = Rarely (once or twice in the past 4 weeks), 2 = Sometimes (three to ten times in the past 4 weeks), 3 = Often (more than ten times in the past 4 weeks) [11]. The total HFIAS score was then obtained by summing the score to all the different questions. Consequently, the score ranged from a minimum of 0 (the answer to all questions was ‘no’) to a maximum of 27 (the answer to all questions was ‘yes, often’). Higher scores indicated higher levels of food insecurity and vice versa. HFI categories were then generated following previously defined guidelines [11]. HFI was classified into severely food insecure, moderate food insecure, mildly food insecure and food secure.
Data management
The data was exported from the kobo collect platform as a Microsoft Excel spreadsheet, cleaned and analysed using STATA version 16.0 for Microsoft Windows. The initial sample consisted of 334 observations. Four (04) observations were deleted from the database as they did not correspond to the selection criteria (children aged below 06 months of age). The proportion of missing data for each explanatory variable varied from none to a maximum of 1.8%. As such, we assumed that the data were missing completely at random, and that deleting observations with the missing data did not yield a considerable change in the dataset. Hence, list-wise deletion was employed, with 09 observations dropped. A total of 321 observations with no missing data, were retained for use for statistical analysis (Fig. 1).
For the retained 321 observations, Z scores were generated using the ‘zanthro’ function of the STATA software, by comparing the recorded weight and height measurements of each child, to the WHO 2007 standard growth charts for children of the same sex and age [14]. Continuous variables were summarized as means with corresponding standard deviations, while categorical variables were presented as counts with percentages.
Data analysis
The prevalence of each form of undernutrition (stunting, wasting and underweight) was compared between the different categories of each explanatory variable using the Chi square test or the fisher’s exact test as appropriate. Explanatory variables with p < 0.20 in the univariate analysis were retained for use as factors in multivariate analysis. The decision to use explanatory variables with p < 0.20 in the univariate analysis as factors in the multivariate model, was to maximize the chance of capturing variables that might influence the association studied or explain some of the variance in the outcome, even though they were not significantly associated to it. Multivariate logistic regression was used to determine characteristics independently associated with increased risk of stunting, underweight and wasting respectively. A 5% probability of a type I error was deemed acceptable. In all instances, two-sided p values were reported.