Menu
Menu
Menu
Menu
Background: Evidences indicate that the risk of linear growth faltering is higher among children born from young mothers. Although such findings have been documented in various studies, they mainly originate from cross-sectional data and demographic and health surveys which are not designed to capture the growth trajectories of the same group of children. This study aimed to assess the association between young maternal age and linear growth of infants using data from a birth cohort study in Ethiopia. Methods: A total of 1423 mother-infant pairs, from a birth cohort study in rural Ethiopia were included in this study. They were followed for five time points, with three months interval until the infants were 12 months old. However, the analysis was based on 1378 subjects with at least one additional follow-up measurement to the baseline. A team of data collectors including nurses collected questionnaire based data and anthropometric measurements from the dyads. We fitted linear mixed-effects model with random intercept and random slope to determine associations of young maternal age and linear growth of infants over the follow-up period after adjusting for potential confounders. Results: Overall, 27.2% of the mothers were adolescents (15-19 years) and the mean ± SD age of the mothers was 20 ± 2 years. Infant Length for Age Z score (LAZ) at birth was negatively associated with maternal age of 15-19 years (β = – 0.24, P = 0.032). However, young maternal age had no significant association with linear growth of the infants over the follow-up time (P = 0.105). Linear growth of infants was associated positively with improved maternal education and iron-folate intake during pregnancy and negatively with infant illness (P < 0.05). Conclusion: Young maternal age had a significant negative association with LAZ score of infants at birth while its association over time was not influential on their linear growth. The fact that wide spread socio economic and environmental inequalities exist among mothers of all ages may have contributed to the non-significant association between young maternal age and linear growth faltering of infants. This leaves an opportunity to develop comprehensive interventions targeting for the infants to attain optimal catch-up growth.
Data used for this study were obtained from the Empowering New Generation in Nutrition and Economic opportunities (ENGINE) birth cohort study, which was conducted from January 2014 to March 2016 in the Oromia Region of Ethiopia. The region was selected purposively being the largest in the country targeted by ENGINE with relatively high rates of stunting. Three districts, namely Goma, Woliso and Tiro Afeta, were further selected based on (i) an expected population of more than 3000 pregnant women to account for loss to follow up, (ii) geographical similarities in agro-ecology and agricultural production practices and (iii) proximity and accessibility. Administratively, each district was further divided into units called kebeles and all recruitment took place at the kebele level. A minimum sample size of 1281 subjects was required to detect a moderate effect size of 0.3 standard deviations (SDs) in infant linear growth over 12 months of follow up (equivalent to an effect size of 0.025 SDs in monthly changes of infant linear growth), assuming an auto correlation of 0.3, 80% statistical power, a one to three adolescent to adult pregnant women ratio, a type I error of 5% and a 30% attrition rate [23]. We therefore included a total of 1423 mothers from the original study who met the inclusion criteria for this study. Inclusion criteria were women of age 15–24 years with a singleton live birth without any congenital anomalies. The mother-infant pair was followed for 12 months after delivery. However, the analysis was based on 1378 subjects with at least one additional follow-up measurement to the baseline. A team of trained data collectors including nurses followed the pair during the study period. There were five time points, each three months apart, at which measurements were taken. Data on household characteristics, socio-economic and demographic information, antenatal exposures and dietary information were collected using a structured, pre-tested, interviewer-administered questionnaire using Android tablet computer at recruitment. Infant length was measured in a recumbent position to the nearest 0.1 cm using a length board (Weigh and Measure LCC, USA) and birth weight was measured to the nearest 10 g using a digital weighing scale (SECA 876, Hannover, Germany) with cloths removed. Low birth weight was defined as birth weight < 2500 g. Maternal height was measured using a stadiometer (Weigh and Measure LCC, USA) to the nearest 0.1 cm with no shoes on and with the five points touching the vertical stand of the stadiometer. A wealth index variable was constructed using principal components analysis based on data on housing conditions, ownership of durable assets and availability of basic services [24]. Infant illness was measured by maternal reporting of symptoms like fever, cough, diarrhea or other symptoms. The outcome variable LAZ was generated using the WHO standards [25] over time, as a continuous variable. Important precautions have been undertaken in order to ensure quality of the data at various stages of the study. Enumerators and their supervisors were recruited based on their prior experiences of engaging in such large scale surveys, fluency in speaking Afan Oromo, familiarity in using electronic data collection tools and their academic backgrounds. Afterwards, adequate training was given on each item of the data collection tool and how to take all the measurements needed in the study using practical applications through role playing. A three days long pretesting was also conducted in order to understand any variations in administering questions and taking measurements among the enumerators before commencing actual data collection. Electronic data collection method was used to minimize errors in data collection and entry by using android tablet computers. The collected data was checked by the supervisors in regular basis before the data was sent to a centrally located server. Additionally, a data manager regularly checked quality of the data and took back mistakes and incomplete data to the field to be corrected. A research team from Jimma University also closely followed up for technical and administrative supports. Refreshment trainings on data quality were given on a regular basis during the two years of study period. We used linear mixed-effects model with random intercept child and random slope time to fit child linear growth curve (change in LAZ) over the study follow-up period. Fixed-effects in the model included LAZ at birth, maternal age, time of study follow-up, a quadratic term of time and the interaction term between maternal age and time that compares maternal age categories on the evolution of child LAZ (linear growth) over time. A quadratic term of time was considered in the model to capture the nonlinear change in the growth curve. The use of other possible models using polynomial and spline functions of time were also considered for a better fitting model by comparing the AIC and BIC estimates of model performance (Additional file 1: Estimates table for the linear, quadratic and cubic-spline models). An unstructured correlation matrix was chosen for the correlation among repeated LAZ measurements per child after comparison of models using other correlation matrices. We considered the use of additional covariates of child growth to maternal age including maternal education, wealth index, iron-folic acid intake and infant illness. Model building was performed through several steps and the selection of important covariates in the final model was decided based on results of the regression outputs and consideration of the literature. We also assessed for effect modification by checking the interactions between the different covariates on child LAZ whenever found to be relevant. We performed different regression diagnostics assessing models goodness of fit (normality and heteroscedasticity of the residuals at different levels), model specification and other numerical problems like multicollinearity, and the sensitivity of the findings to potential influential observation. All the analysis was performed using STATA version 14 (StataCorp, Texas, USA) and all the tests were two-sided with a statistical significance considered at p < 0.05. P values were adjusted for multiple testing of hypothesis using Benjamini-Hochberg method [26]. Ethical approval was granted from the Institutional Review Board of Jimma University in Ethiopia (RPGC/264/2013) and Tufts University in USA (Tufts Health Sciences Campus IRB reference number:11088) before commencement of the study. Informed consent was obtained from the participants after a detailed explanation of the objectives of the study. Data was registered and stored in a secured server and access to the data was upon permission of the principal investigators with personal identifiers removed. During the study women or infants who had health problems were referred to a nearby health facility to seek proper medical care.
