Maternal nutritional status, decision-making autonomy and the nutritional status of adolescent girls: a cross-sectional analysis in the Mion District of Ghana

A mother’s nutritional status and participation in household decision-making, a proxy for empowerment, are known determinants of improved nutrition and health outcomes for infants and young children; however, little is known about the association among adolescents. We examined the association between maternal nutritional status, decision-making autonomy and adolescent girls’ nutritional status. We analysed data of 711 mother-adolescent girl pairs aged 10-17 years from the Mion District, Ghana. Maternal nutritional status and decision-making autonomy were the independent variables while the outcomes were adolescent girls’ nutritional status as defined by anaemia, stunting and body mass index-for-age Z-score categories. Girl-level (age, menarche status and the frequency of animal-source food consumption), mother-level (age, education level, and monthly earnings) and household-level (wealth index, food security status and family size) covariates were adjusted for in the analysis. All associations were examined with hierarchical survey logistic regression. There was no association between maternal height and adolescent girls being anaemic, underweight or overweight/obese. Increasing maternal height reduced the odds of being stunted [adjusted odds ratio (OR) 0.92, 95 % CI (0.89, 0.95)] for the adolescent girl. Maternal overweight/obesity was positively associated with the girl being anaemic [OR 1.35, 95 % CI (1.06, 1.72)]. The adolescent girl was more than five times likely to be thin [OR 5.28, 95 % CI (1.64-17.04)] when the mother was underweight. Maternal decision-making autonomy was inversely associated with stunting [OR 0.88, 95 % CI (0.79, 0.99)] among the girls. Our findings suggest that intergenerational linkages of a mother’s nutritional status are not limited to childhood but also during adolescence. We analysed baseline data from the Ten2Twenty-Ghana study; the study design, setting, and population have previously been described in detail elsewhere(29). In brief, Ten2Twenty-Ghana was a randomised controlled trial evaluating the efficacy of multiple-micronutrient fortified biscuits compared with unfortified biscuits on micronutrient status, height and cognition of adolescent girls aged 10–17 years in the Mion District, in north-eastern Ghana. The study began with a large survey (n 1057), which led to a trial (n 621). The survey was conducted in November/December 2018; it includes data on the nutritional status of the girls, their time use, aspirations, and dietary intake, socio-economic status, household demographics, and structure, and maternal factors including nutritional status, participation in household decision-making and a life-history calendar that captured maternal fertility, education and occupation. Participation was entirely voluntary, and the girl gave her assent, after receiving signed/thumb-printed informed consent from her guardian or parent. The study protocol was approved by the Navrongo Health Research Centre Institutional Review Board (NHRCIRB323). The research was carried out in the Mion District of Ghana’s. The climate of the district is tropical, with two distinct seasons: a dry season from November to March and a rainy season from April to October. According to the 2010 Ghana population and housing census, Mion District has a population of 81 812, with 91⋅1 % of that population residing in rural areas; about 19⋅5 % of the district’s female population is aged 10–19 years; the illiteracy rate is high in the district and the population is mostly dependent on agriculture as its livelihood(30). The study participants were adolescent girls aged 10–17 years and their mothers, residing in the Mion district, in Ghana’s Northern region. The adolescent girls were selected from nineteen different elementary schools across the district. The sampling included four clusters, where four schools in the urban area were all selected, and fifteen larger rural schools were selected. A 16-item screening questionnaire ensured that all participating adolescent girls were pre- or post-menarche, healthy with no apparent signs of poor health, not pregnant and not lactating at the time of the survey(29). Adolescent girls with missing data on haemoglobin (Hb) status (n 4) and mothers with missing data on decision-making (n 53) were excluded from the final analysis. Thus, a total of 711 mother–daughter pairs were used in the present study (Fig. 1). Flowchart showing the sample selection of the present study. The data collection methods included one-on-one interviews with a semi-structured questionnaire, anthropometry, Hb status assessment by finger prick, a qualitative 24-hour dietary recall (24HR) and a one-month food frequency questionnaire (FFQ), conducted in November/December 2018. The questionnaire was pre-tested in the neighbouring Yendi Municipality. Given some questions like menarche were sensitive, interviewers were trained ladies, recruited from the University for Development Studies (UDS). Supervisors verified and validated all questionnaires for consistency and completeness throughout fieldwork. Standardised anthropometric guidelines(31) were followed in measuring the height (cm) and weight (kg) of the mother in duplicates to the nearest 0⋅1 decimal using a Seca stadiometer and a digital weighing scale, respectively. The average of the duplicate measures was used in the analysis. Body mass index (BMI, kg/m2) of mothers was computed and BMI categories were defined: underweight (BMI < 18⋅5), normal weight (18⋅5 ≤ BMI < 25) and overweight/obese (BMI ≥ 25)(31). The attained height of the mother was the mean height. The mothers’ participation in household decision-making autonomy, herein referred to as maternal autonomy, was assessed with the demographic and health survey 8-item final decision-making index(32). These questions included final say on: (1) ‘how respondent's money is spent’, (2) ‘health care’, (3) ‘making large household purchases’, (4) ‘making household purchases for daily needs’, (5) ‘family visit’, (6) ‘food to be cooked each day’, (7) ‘what to do with money husband earns’ and (8) ‘the number of children to have’. We assigned a score of 1 to a mother involved in decision-making alone or with any other person in the household, whereas a score of 0 was given if she did not participate in decision-making. The scores from the 8-item questions were summed, ranging from 0 to 8, with a higher score denoting higher participation in decision-making in the household. The height (cm) and weight (kg) of the adolescent girl were also measured following the same guidelines described for the mother. We computed the height-for-age Z-score (HAZ) and body mass index-for-age Z-score (BAZ) of the girl with WHO AnthroPlus using the WHO growth reference for 10–19 years of age adolescent girls. We defined stunting as HAZ < −2sd, whereas BAZ was categorised as thinness (BAZ < −2sd), normal weight (−2sd ≤ BAZ ≤ + 1sd) and overweight/obese (BAZ > + 1sd)(33). The attained height of the girl was the mean of the duplicate height measured. Phlebotomists from the Tamale Teaching Hospital assessed Hb by finger prick using a HemoCue 301 (Angelholm, Sweden; 0⋅1g/dl precision). The photometer was calibrated with certified quality control samples from the CDC/Atlanta, and the readings of ten patients were repeated each day for quality control. We defined anaemia as having Hb < 12 g/dl for girls aged ≥12 years and Hb < 11⋅5 for girls <12 years(34). A single qualitative 24HR assessed the dietary diversity score (DDS) of the girls using a 10-food group indicator(35). In the 24HR, the girl was first asked to mention all foods, including drinks and snacks that she consumed in and outside the home (including school) the previous day. She was then asked to describe the ingredients of any mixed dishes. Based on a pre-defined table with a list of all possible food items in the ten food groups, a score of 1, else 0 was given if a girl consumed at least one food item from any food group. A summated score was computed by summing the scores for all the food groups, resulting in a maximum attainable score of 10. The ten food groups included: grains, white roots, tubers and plantains (1), pulses (beans, peas and lentils) (2), nuts and seeds (3), dairy (4), meat, poultry and fish (5), eggs (6), dark green leafy vegetables (7), other vitamin A-rich fruits and vegetables (8), other vegetables (9) and other fruits (10). We next defined minimum dietary diversity (MDD-W) as DDS ≥5(35). The effect of dietary diversity as a continuous score (DDS) and as a dichotomous variable (MDD-W) was also explored. Additionally, the girls’ dietary patterns were assessed with a 1-month FFQ using the ten food groups(35). The data also included the ethnicity, religion, class and age of the girl using a household roster and as well, menarche status based on recall. Maternal covariates in the data included education (none, basic, secondary/higher), occupation (not currently working, farmer, trader and others), literacy (dichotomous) and age. A life-history calendar also tracked the mother's parity and earnings in a month. The international wealth index (IWI) was used to assess the socio-economic status of the households(36). The IWI ranks households based on the ownership of durable assets including TV, refrigerator, phone, bicycle, car, household utensils categorised as cheap (<$40) and expensive (>$250), access to electricity, the type of water and toilet facilities accessed by the household and as well as the floor material of the household. The IWI was created purposely for assessing the socio-economic status of households in LMICs using principal component analysis (PCA) on data from 97 LMICs(36). We adopted and used the IWI SPSS Syntax to run the calculations; the IWI ranges from a minimum of 25 to a maximum of 100. Households were subsequently ranked into quintiles of wealth based on their IWI score. The Food Insecurity Experience Scale (FIES)(37) was used to measure the food security of the girls’ households. The FIES is an 8-question survey that uses yes/no responses to assess the degree of food insecurity. When the answer is ‘yes’, the questions are given a score of 1; otherwise, they are given a score of 0. We computed the FIES score by summing the scores of the eight items; the score ranged between 0 and 8. A higher score indicated a more severe level of food insecurity, whereas a lower score indicated a less severe level of food insecurity. The sum score was used to assign the girls to one of the following categories: food secure (FIES = 0), mild food insecure (FIES score 1–3), moderate food insecure (FIES score 4–6) and severe food insecure (FIES score 7–8). A household roster captured data on paternal education (none, primary, secondary/higher), occupation (none, farmer, trader/self-employed, formal employee) and literacy (dichotomous). We computed and included in our analyses household dependency ratio, sex and literacy ratios similar to the Ghana Statistical Service(30). Count variables for household size and the number of children under 5 years were also explored. The statistical software programs SPSS (version 26) and R-studio (version 4.0.0) were used to analyse the data. Categorical descriptive variables were expressed as percentages and frequencies, while continuous variables were presented as means and standard deviation (mean ± sd). Data normality was examined visually with the normality histogram curves and Q-Q plots. We assessed the association between maternal nutritional status, autonomy and the nutrition of the adolescent girls using survey logistic regression (binary and multinomial); including a random intercept for the study design (School). The outcome variables in the binary logistic regression analysis were anaemia (anaemic or not) and stunting status (stunted or not), while the BAZ category (normal, thin, overweight/obese) of the girl was the outcome variable for the multinomial logistic regression. The mother’s attained height, BMI category (normal, underweight and overweight/obese) and autonomy (as a continuous variable) were the exposure variables. We categorised the height of the mother into a dichotomous variable as short stature (<145 cm height) and normal stature (≥145)(31) but short stature prevalence (0⋅6 %) was low; hence, we analysed height (cm) only as a continuous variable. In the analysis, maternal height and autonomy were analysed together and the BMI category of the mother replaced height in a repeated analysis. The crude and adjusted odds ratios and 95 % confidence intervals with their corresponding P-values were presented. A two-tailed P-value ≤ 0⋅05 at a 95 % confidence interval was considered statistically significant. Potential confounding variables were selected a priori based on literature and included girl-level (age, menarche status, dietary diversity and/or animal-sourced food intake), maternal (age, education, monthly earnings) and household (food security, wealth index, household size) factors(14,17,18,20–25). Multicollinearity between explanatory variables was assessed using tolerance values (TOL) < 0⋅1 and the variance inflation factor (VIF) < 10 in a linear regression step. Aside from the basic model (model 1), three multivariable models were developed. Model 2 was adjusted for adolescent girl-level characteristics such as the girl's age, menarche status, DDS and frequency of animal-source food intake in a hierarchical order. Model 3 took into account other maternal factors such as age, education and monthly wages. Finally, household-level factors such as household food security, wealth index and family size were adjusted for in model 4. We looked for pair-wise interaction terms between maternal decision-making and the adolescent girl's other explanatory variables, such as DDS and animal-sourced food intake, but none was found to be significant. Mathematically, the models are expressed below. Model 1 (Crude model): Model 2: adjusted for potential child-level covariates. Model 3: adjusted for potential maternal covariates. Model 4: adjusted for potential household-level covariates. We repeated all the analyses using linear mixed model analysis, including a random intercept of the study design (school) in which the Hb status (g/dl), HAZ and the BAZ of the girl were the outcomes (Supplementary Tables S1 and S2).