Not water, sanitation and hygiene practice, but timing of stunting is associated with recovery from stunting at 24 months: Results from a multi-country birth cohort study

Objectives: To measure the role of water, sanitation and hygiene (WASH) practices on recovery from stunting and assess the role of timing of stunting on the reversal of this phenomenon Design: Data from the MAL-ED multi-country birth cohort study was used for the current analysis. Generalised linear mixed-effects models were used to estimate the probability of reversal of stunting with WASH practice and timing of stunting as the exposures of interest. Setting: Seven different countries across three continents. Participants: A total of 612 children <2 years of age. Results: We found that not WASH practice but timing of stunting had statistically significant association with recovery from stunting. In comparison with the children who were stunted at 6 months, children who were stunted at 12 months had 1.9 times (β = 0.63, P = 0.03) more chance of recovery at 24 months of age. And, children who were stunted at 18 months of age even had higher odds (adjusted OR = 3.01, β = 1.10, P < 0.001) of recovery than children who were stunted at 6 months. Additionally, mother's height (β = 0.59, P = 0.04) and household income (β = 0.02, P < 0.05) showed statistically significant associations with the outcome. Conclusions: The study provided evidence for the role of timing of stunting on the recovery from the phenomenon. This novel finding indicates that the programmes to promote linear growth should be directed at the earliest possible timepoints in the course of life. Exposure to poor WASH environment induces diarrheal diseases and other subclinical infections(7). This causal pathway is further modified by poverty and poor maternal education(18). The other covariates that are known to play a crucial role in WASH–stunting hypothetical framework are low birth weight and height, inadequate energy from protein, and low maternal height and weight(10). We developed a conceptual framework (Fig. 1) based on the abovementioned context, which was used for variable selection and data analysis. Underlying causes of stunting other than those diagrammed here are either primary causes or effect modifiers in the proposed pathway. Conceptual framework depicting the water, sanitation and hygiene (WASH)–stunting causal pathway Data for this specific analysis was collected from the MAL-ED (Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health) birth cohort study. The MAL-ED study was conducted from 2 November 2009 to 28 February 2014 at eight different sites across three continents. A total of 2145 children from Dhaka, Bangladesh (BG), Vellore, India (IN), Bhaktapur, Nepal (NP), and Naushahro Feroze, Pakistan (PK), in Asia; Fortaleza, Brazil (BR), and Loreto, Peru (PE), in the Americas; and Venda, South Africa (SA), and Haydom, Tanzania (TZ), in Africa were enrolled within 17 d of their birth and followed uniformly up to 24 months of age. Enrolment took place over a 2-year period with the goal of enrolling 200 children per site. The detailed study design is described elsewhere(19). Variables used in the current analysis are: access to improved water (yes/no), access to improved sanitation (yes/no), treat water to make it safe (yes/no), caregiver washes her hands after using the toilet (never/rarely or sometimes/always), caregiver washes her hands before preparing food (never/rarely or sometimes/always), caregiver washes her hands after helping the child to defecate (never/rarely or sometimes/always), mother’s height, weight and educational status, asset index, household income, energy from protein intake, birth LAZ, birth weight-for-age z-score, diarrhoea episodes, exclusive breastfeeding days, and minimum dietary diversity (yes v. no). Demographic and socioeconomic status (SES) questionnaires were adopted from the DHS questionnaires, and water and sanitation sources were defined as improved (or not) based on the WHO criteria(20). Data on WASH practice was collected at 6, 12, 18 and 24 months of child’s age. But, the data did not show any significant variations over time (see online supplementary material, Supplemental Fig. 1). Hence, to avoid additional complexity and to ensure temporality, WASH data collected at 6 months of age was used for the current analysis. The household asset index was constructed using household asset data obtained from the SES questionnaire. From these asset-related dichotomous variables, a common factor score for each household was generated using principal components analysis. Trained field workers visited the households twice in a week to collect intensive dietary and morbidity data. Twenty-four-hour food frequency data was collected monthly from 9 to 24 months of age for assessing child’s dietary and energy intake. The 24-h multiple-pass dietary recall approach was used for this purpose(21). A food composition table, which was locally adapted, was used to calculate the amount of energy taken from the documented diet(22). From the nutrient intake group, the amount of ‘energy from protein’ was selected because a multi-country analysis of the same data revealed ‘lower per cent of energy from protein’ to be an important factor contributing to the odds of being in a lower length-for-age category at 24 months(10). Moreover, all the specific food groups (carbohydrate, protein and fat) were highly correlated to each other. Minimum dietary diversity (MDD), a core indicator of infant and young child’s feeding practice, was used to measure the appropriateness of the complementary feeding practice of children(23). Data on MDD was collected on 6th, 7th and 8th months of child’s age. MDD was a binary variable – ‘yes’ was indicated by 1, and ‘no’ by 0. An MDD score was developed by adding the MDD values of 6th, 7th and 8th months. If the total score was ≥2, the child was mentioned as having MDD. To document the breastfeeding status, the data collector questioned the mother about the child’s food consumption over the past 24 h. If the response was similar to the WHO definition of exclusive breastfeeding (no other food or drink, not even water – except breast milk – including milk expressed, ORS, drops and vitamins, minerals and medicinal syrups), then the child was considered to be exclusively breastfed. Instead of exclusive breastfeeding status (yes v. no), exclusive breastfeeding days was used as it counts the specific number of days. Diarrhoea is defined as having ≥3 loose stools in a 24-h period or at least one loose stool with blood reported by the mother(24). A diarrheal episode is defined as being separated from another episode by at least ≥2 diarrhoea-free days(24). Using a common protocol, trained field workers measured anthropometric indices monthly up to 24 months of age. Measuring boards were used to measure the length to the nearest 0·1 cm, and digital scales were used to measure the weight of the children to the nearest 10 g(10). LAZ and weight-for-age z-score for each child was determined using the WHO 2006 Child Growth Standards(25). A child with LAZ <–2 was classified as stunted(25). Enrolment weight and length, which was taken within first week of birth, was used as the surrogate measure for birth anthropometry. Standard wooden height-measuring boards and bathroom scales were used for measuring maternal height and weight. Children who became stunted at any of the timepoints of 6, 12 or 18 months of age but not found to be stunted at 24 months of age were classified as having recovered from stunting. A total of 626 children became stunted at 6, 12 or 18 months of age; of them, 130 could recover. Children who were stunted on multiple occasions were counted under the first month of onset. Out of these 626, fourteen participants had missing values. After excluding the ID with missing data, a total of 612 children’s data were available for the current analysis. Out of these 612 children, 127 constituted the ‘recovery from stunting’ group, and the rest remained as the non-recovery group. We first described the overall and country-wise household, maternal and nutritional status of the children using mean, standard deviation and percentages. A comparison of LAZ trajectory between recovered and non-recovered children was done using line graphs, and the rate of recovery was reported as a percentage of children who were stunted at 6 months of age but not at 24 months of age and likewise. This multi-country dataset contains clusters of non-independent observational units, namely ‘country’. Measurements within a country might be more similar than measurements from different countries. Moreover, the cluster sizes were also unequal. To adjust this clustering effect, we used generalised linear mixed-effects models (GLMM) where the intercept of the variable ‘country’ was kept as random. This approach allows a robust estimation of variance in the outcome variable within and between the clusters(26). GLMM estimated the probability of recovery from stunting with WASH practice as the exposure of interest, adjusting for all the other possible covariates. We began with the base model (Table 2, model 1) that was built with the fixed effects of WASH variables. Then, keeping the variables of the base model fixed, other covariates were added one after another according to the conceptual framework. Hence, a model was nested in its next model as it contained all the predictor variables used to build the previous model, plus at least one additional variable. This means that variables of a model were a subset of the next model. Model selection was done based on information-theoretic model selection procedures – Akaike Information Criterion (AIC) and Bayesian information criterion (BIC). Information-theoretic methods are equally applicable for both nested and non-nested models and can provide better estimate statistics to quantify the extent of differences between models than other model selection methods(27,28). The model showing the lowest AIC and BIC values was selected as the final model (model 41, Supplemental File 2). Along with the WASH variables, the fully adjusted final model (Table 2, model 2) contains mother’s height, LAZ at birth, gender and income as fixed effects. During exploratory data analysis, we noticed that timing of stunting could modify the odds of recovery from stunting. Therefore, to see the effect of timing of stunting on recovery, we developed the third generalised linear mixed-effects model (model 3, Table 2). We created a variable named ‘timing of stunting’ with three criteria, stunted at 6 months, stunted at 12 months and stunted at 18 months, and added the variable to model 2. We excluded data from Pakistan because quality assurance procedures identified an unexplained bias in a subset of length measurements(10). Data analysis was conducted in R (version 3.5.1), and lme4 package was used for GLMM(29). Parameter estimates for the fixed effects of water, sanitation and hygiene (WASH) and timing of stunting on recovery from stunting from the fully adjusted model AIC, Akaike Information Criterion; BIC, Bayesian information criterion. *P < 0·05, **P < 0·01, ***P < 0·001.