Study Area
The nutrition surveillance covered 44 districts in the three regions of Northern Ghana comprising the Northern Region (NR), Upper East Region (UER) and Upper West Region (UWR). The three regions share some boundaries with each other. The under-five population for the three regions is estimated to be 1,308,742 [18].
Majority of the people (60.0 %) have agriculture as their main occupation while some are involved in trading. The main staple foods including maize, sorghum, millet and yam are usually harvested from October through December. Although the food security situation is usually good during the harvesting time, child care tends to suffer because of lack of time on the part of rural mothers.
Survey design, population and sampling
In this study, regionally representative nutrition surveillance data were collected in November, 2013 in a cross-sectional nutrition survey. In this stratified cluster survey, each of the three regions was considered a stratum, and a number of clusters per stratum selected randomly using the probability proportionate to size (PPS) technique. Women of reproductive age and their children 6–23 months old in the sampled households were included in the study.
The main outcome variable used to calculate the sample size was prevalence of chronic malnutrition which was 37.4 % in Northern Region, 31.5 % in Upper East, and 23.1 % in Upper West (UNICEF MICS, 2011). Aiming at an absolute precision of 5 % at the 95 % confidence level, further assuming a correction factor of 2.0 (the “design effect”) for cluster sampling, and allowing for 5 % refusals and incomplete questionnaires, the required minimum sample sizes (n) were 756, 697 and 573 for Northern, Upper East and Upper West regions respectively.
The sample size was estimated using OpenEpi software for epidemiologic statistics version 3.01.
Selecting Households
There is a minimum number of clusters that should be included in each stratum for the survey to be considered valid [19]. Usually, 25 clusters are considered a minimum.
In each cluster, a complete list of all households was compiled and systematic random sampling used in selecting study households. All the households in each cluster were serially numbered. To get the sampling interval, the total number of households in a cluster was divided by the sample size of 20. The first household was then randomly selected by picking any number within the sample interval. Subsequent selections were made by adding the sampling interval to the selected number in order to locate the next household to visit. If the selected household did not have a target respondent, then next household was selected using the systematic sampling procedure. This process continued until the required sample size was obtained. A minimum of 20 mother/child pairs were randomly selected from a cluster. Only one eligible participant was selected from each household for interview, using simple random sampling.
Data Collection
Quantitative data were collected using structured questionnaire in face-to-face interviews during house-to-house visits. Socioeconomic and demographic characteristics of participants, child’s age, gender, morbidity in the past week, child feeding practices, and child anthropometry data were also collected. Details of data collected are contained in the Additional file 1.
The data were collected by interviewers who had completed at least Senior Secondary School and who underwent intensive training for two days on the content of the questionnaire and on general approaches to data collection.
Independent and dependent variables
WHO IYCF indicators [minimum dietary diversity (MDD), minimum meal frequency (MMF), minimum acceptable diet (MAD)] and a child feeding index (CFI) were the main independent variables. The main dependent variable was child nutritional status which was treated as both continuous and categorical variables. The continuous variables were length -for-age Z-score (LAZ), weight-for- length Z-score (WLZ) and weight-for-age Z-score (WAZ). Categorical variables were stunting, wasting and underweight which reflect LAZ, WAZ and WLZ below −2 standard deviations below the population median.
Other confounders included (i) age and gender of the child; (ii) maternal education, and utilization of prenatal care; (iii) and household wealth status. A brief description of main independent and dependent variables is as follows:
Assessment of IYCF Practices
Infant and young child feeding indicators including minimum meal frequency, minimum dietary diversity and minimum acceptable diet were estimated by recall of food and liquid consumption during the previous day of the survey as per WHO/UNICEF guidelines [20].
Minimum meal frequency is the proportion of children who received complementary foods at least the minimum recommended number of times in the last 24 hours. A child was judged to have taken ‘adequate number of meals if he/she received at least the minimum frequency for appropriate complementary feeding (that is, 2 times for 6–8 months and 3 times for 9–11 months, 3 times for children aged 12–23 months) in last 24 hours. For non-breastfed children, the minimum meal frequency was 4.
The WHO defined minimum dietary diversity as the proportion of children aged 6–23 months who received foods from at least four out of seven food groups [7, 8]. The 7 foods groups used for calculation of WHO minimum dietary diversity indicator are:
(i) grains, roots and tubers; (ii) legumes and nuts; (iii) dairy products; (iv) flesh foods; (v) eggs; (vi) vitamin A rich fruits and vegetables; and (vii) other fruits and vegetables.
The dietary diversity score therefore ranged from 0–7 with minimum of 0 if none of the food groups is consumed to 7 if all the food groups are consumed.
Additionally, the individual dietary diversity score (IDDS) of the children was also determined based on 14 food groups as recommended by the Food and Agriculture Organization (FAO) [21]. The food groups are cereals, vitamin A rich vegetables and tubers, white roots and tubers, dark green leafy vegetables, other vegetables, Vitamin A rich fruits, other fruits, organ meat (iron rich), flesh meat, eggs, fish, legumes, nuts and seeds, milk and milk products, and oils and fats. Based on these food groups, the dietary diversity score therefore ranged from 0–14 with minimum of 0 if none of the food groups is consumed to 14 if all the food groups are consumed. For comparison reasons, these food items were re-grouped into seven food groups according to WHO infant feeding guidelines.
From the dietary diversity score, the minimum dietary diversity indicator was constructed using the WHO recommended cut-off point with a value of “1” if the child had consumed four or more groups of foods and “0” if less. Minimum dietary diversity is the proportion of children who ate at least 4 or more varieties of foods from the seven food groups in a 24-hour time period [7, 8]. Minimum acceptable diet is a composite indicator of minimum dietary diversity and minimum meal frequency. Breastfed children who meet both the minimum diversity and the minimum meal frequency are considered to have met the WHO recommended minimum acceptable diet.
Previous studies have described complementary feeding practice using single indicators but a combination of indicators is needed to better explain the role of complementary feeding practices in child growth. To adequately quantify appropriate complementary feeding, we used a composite indicator comprising three of the WHO core IYCF indicators that relate closely to complementary feeding. These are timely introduction of solid, semi-solid or soft foods at 6 months, meeting minimum meal frequency, and minimum dietary diversity. Appropriate complementary feeding was thus defined in this study as when the child met all the following three criteria:
-
i.
Complementary feeding commenced at 6th month of birth
-
ii.
Minimum dietary diversity
-
iii.
Minimum meal frequency
Nutritional Status Assessment
Anthropometric indicators of length -for-age (LAZ), weight-for-age (WAZ), and weight-for-length (WLZ) were determined as recommended by the World Health Organization [22].
The questionnaire and anthropometric assessment was carried out by well-trained health and nutrition personnel. The age of the child was determined based on the date of birth (obtained from child health records booklets, birth certificates and baptismal cards) and the date of the survey. Provision was made to use calendar of events to estimate age of the child in the absence of documentary evidence.
Length, weight, and upper mid-arm circumference were obtained using standardized techniques and equipment. Recumbent length was measured to the nearest 0.1 cm with subjects in a lying position. The crown-heel length was taken using an infantometer (a flat wooden surface with head and foot boards). The child was placed on its back between the slanting sides. The head was placed so that it is against the top end. The knees were gently pushed down by a helper. The foot-piece was then moved toward the child until it presses softly against the soles of the child's feet and the feet are at right angles to the legs. The weight in light clothes was obtained using a digital weighing scale (SECA 890) to the nearest 0.1 kg. The mid-upper arm circumference (MUAC) was measured in centimeters, to the nearest 0.1 cm, using standard MUAC measuring tape for children.
Assessment Socio-economic Status
Principal Component Analysis (PCA) was used to determine household socioeconomic status (wealth index) from modern household assets namely, the presence of electricity, type of cooking fuel, material of the dwelling floor, source of drinking water, type of toilet facility, and possession of household items including computer, radio, television, refrigerator, bicycle, motorcycle/scooter, car/truck and mobile phone [23–26].
Data Quality Control Measures
In an effort to collect quality data, a number of strategies were applied. A two-day training session aimed at ensuring the reliability and validity of data collected was organized for data collectors. The training ensured a good understanding and acquisition of skills for effective and efficient administration of the data collection tools. The content of the training included the aim of study, survey methodology including selection of eligible participants, data recording, administration of questionnaires and supervision. In addition, the training focused on the art of interviewing and clarifying questions that were unclear to the respondents.
The final stage in the training of data collectors involved field-testing of data collection tools. The main aim here was to refine the tools and to ensure the competence of the data collectors. The household questionnaires were pre-tested and revised before the main field work commenced.
In each team, there was a supervisor who ensured that all the methodological issues were being adhered to. Furthermore, field supervisors checked all data collected in a given day and made sure that all field challenges were attended to immediately in the field. Any errors noted were discussed with the concerned enumerators. Briefing meetings took place every day where teams shared their experiences in the field and necessary corrections and recommendations made to ensure smooth implementation of the survey. In addition, the Survey Coordinator visited teams in the communities at random to observe how interviews were conducted.
Statistical Data Analyses
The analysis of data took into account the complex design of multi-stage cluster surveys. All quantitative data were coded for statistical analysis using SPSS Complex Samples module for windows 18.0 (SPSS Inc, Chicago). This was done in order to make statistically valid population inferences and computed standard errors from sample data. Design weights were added to each region (that is, total population divided by number of respondents) to perform weighted analysis.
The Emergency Nutrition Assessment (ENA) for SMART software (2010 version) was used for the anthropometric data analysis and reported using WHO 2006 growth reference values with SMART cut-offs.
Both bivariate and multivariate analyses were performed to identify the determinants of stunting. Firstly, bivariate analyses for all the various risk factors were performed using Chi-square (χ2) tests and ANOVA. The association between dependent variables (stunting and wasting) and independent variables was determined using multiple logistic regression modeling, which included all potential socioeconomic, and demographic confounders that were significant at P values < 0.05 in the bivariate analysis. The logistic regression outputs were presented as adjusted odds ratios (AOR) with 95 % confidence intervals (CI).
Before testing for associations between independent variables and the dependent outcomes WLZ, LAZ and WAZ, the data were tested for dependencies, intra-class correlations and clustering effects between the different regions [27]. Additionally, the data were cleaned and outliers removed. Z-scores which were outside the WHO flags: WLZ −5 to 5; LAZ −6 to 6; and WAZ −6 to 5 were excluded from the data set.
Underweight was defined as (WAZ < −2 SD), acute malnutrition (WLZ < −2 SD) and chronic malnutrition (LAZ < −2 SD).
Ethics Consideration
The permission to carry out this study was sought from district health authorities and the study protocol was approved by the School of Medicine and Health Sciences of the University for Development Studies, Ghana. Informed consent was also obtained verbally after needed information and explanation. Participation was voluntary and each woman signed (or provided a thumb print if she was uneducated) a statement of an informed consent after which she was interviewed.