Impact of adversity on early childhood growth & development in rural India: Findings from the early life stress sub-study of the SPRING cluster randomised controlled trial (SPRING-ELS)

Introduction Early childhood development is key to achieving the Sustainable Development Goals and can be negatively influenced by many different adversities including violence in the home, neglect, abuse and parental ill-health. We set out to quantify the extent to which multiple adversities are associated with impaired early childhood growth & development. Methods This was a substudy of the SPRING cluster randomised controlled trial covering the whole population of 120 villages of rural India. We assessed all children born from 18 June 2015 for adversities in the first year of life and summed these to make a total cumulative adversity score, and four subscale scores. We assessed the association of each of these with weight-for-age z-score, length-for-age z-score, and the motor, cognitive and language developmental scales of the Bayley Scales of Infant Development III assessed at 18 months. Results We enrolled 1726 children soon after birth and assessed 1273 of these at both 12 and 18 months of age. There were consistent and strongly negative relationships between all measures of childhood adversity and all five child growth & development outcome measures at 18 months of age. For the Bayley motor scale, each additional adversity was associated with a 1.1 point decrease (95%CI -1.3, -0.9); for the cognitive scales this was 0.8 points (95%CI -1.0, -0.6); and for language this was 1.4 points (95%CI -1.9, -1.1). Similarly for growth, each additional adversity was associated with a -0.09 change in weight-for-age zscore (-0.11, -0.06) and -0.12 change in height-for-age z-score (-0.14, -0.09). Discussion Our results are the first from a large population-based study in a low/middle-income country to show that each increase in adversity in multiple domains increases risk to child growth and development at a very early age. There is an urgent need to act to improve these outcomes for young children in LMICs and these findings suggest that Early Childhood programmes should prioritise early childhood adversity because of its impact on developmental inequities from the very start. SPRING-ELS was a sub-study of the Wellcome Trust funded SPRING cluster randomised controlled trial in India analysing stress and adversity in young children. Details on SPRING are presented elsewhere[15] but in brief SPRING in India developed an innovative, feasible, affordable & sustainable community-based approach to delivering a home visiting programme through a new cadre of community-based worker with the aim to improve early childhood growth and development. SPRING was designed from the outset to be feasible and scalable through the governmental health system. A parallel trial was done in Pakistan with the same aim but working through existing health system structures with an existing cadre of worker. SPRING was evaluated by parallel cluster randomised controlled trials with clusters in India defined as the catchment area of functioning health sub-centres, the lowest level of the Indian primary healthcare system. There were 24 clusters. Primary outcomes were height-for-age, the best early childhood predictor of human capital[16], and Bayley Scales of Infant Development III (BSID-III), the gold standard assessment of a child’s development in the early years[15]. These impact outcomes were complemented by in-depth economic analysis, process-evaluation and a broad range of intermediate outcomes selected based on a conceptual-framework. This additional work will inform unpacking of the SPRING causal pathway, provide deeper understanding of mechanisms of trial impact, and inform lessons for scale-up and incorporation into health systems. SPRING took place in 120 villages across three administrative blocks of Rewari district, Haryana state, India. The total population was around 200,000. Rewari district is predominantly rural and has health and demographic indicators around average for the state. The overall literacy rate in Haryana is 76%, with female literacy of 67%[17]. The sex ratio is 879 females per 1000 males[17]–amongst the lowest ratio in India. Infant mortality is 41/1000 live births[18]–around the national average. More than one third of under-five year old children are stunted[19]. The SPRING trial is registered with ClinicalTrials.gov, number {“type”:”clinical-trial”,”attrs”:{“text”:”NCT02059863″,”term_id”:”NCT02059863″}}NCT02059863. Participating SPRING mothers and their children were identified by an ongoing trial surveillance system whereby trained resident fieldworkers visited every household in the study area every 8 weeks to identify pregnancies and births, and follow-up pregnant women & children already identified. Surveillance system fieldworkers collected the demographic & socioeconomic data used in this study on custom-programmed mobile phones at enrolment. A separate group of fieldworkers did detailed SPRING assessments with children and their mothers when children turned 12 & 18 months of age. Adversity assessments were done at 12 months with mothers and children where the child was planned to have outcome assessments at 18 months of age. This was at least the first 50 children born in each of the 24 clusters after the trial start date (18 June 2015). These detailed assessments were spread out over two days in order to reduce the burden to participants and took a total of around 2.5 hours. All questionnaires were asked of the mother, and observations were done with both the mother and child. Other assessments during this two day visit that are not described in this paper include anthropometry, a feeding questionnaire, a maternal knowledge questionnaire and saliva and hair sampling for stress biomarker analysis. To develop the set of adversities, we carried out formative research with local mothers and grandmothers, took advice from child development experts and reviewed the literature on existing evidence and tools. We selected 22 adversities with a focus on those adversities operating at the household level and did not consider those operating more widely because young children in this setting spend most of their time and interact most closely with family members inside the home and these adversities are therefore of most importance to these young children. Three of the adversities were assessed at enrolment, and the other 19 were assessed at 12 month assessment (Table 1). prevalence and proportion of imputated values. a SES score calculated with principle components analysis using data on mother, household demographics and animal & asset ownership b Answered yes to question: “Since you became pregnant, have you or your immediate family who live with you been in debt?” c Answered yes to question: “Since you became pregnant, have you ever been hungry because you could not afford to buy food?” or similar related to child d Using WHO multi-country study on women’s health and domestic violence against women e If woman reported husband drinking alcohol, answered yes to question: “does this cause any problems for you” f Question: “When [person] found out your baby was a girl were you/they happy, unhappy or didn’t mind whether you had a girl or a boy?” g Assessed using observed feeding index. Very low quality means < = 1 positive verbalisations, and < = 1 games played and < = 1 responsive actions, plus > = 1 negative actions by mother towards child during feeding session h The Home Observation for the Measurement of the Environment Inventory i Not exactly 20% because cut-off made at HOME score of 27 *E All items were assessed at 12 months of age except those marked *E which were collected at enrolment into the surveillance system To explore further we examined the relative importance of particular groups of adversities based on a conceptual framework starting with direct adversities for a child with links to more distal adversities including maternal stress and difficulties in the carer-child relationships and the more overarching socio-economic factors (Fig 1). For the second part of this analysis, the 22 adversities were therefore grouped as follows: 1) household-level socio-economic factors, 2) maternal stressors, 3) child-carer relationships and 4) child-related factors and are described within these groups below. Group 1—Socioeconomic—Consists of five factors: 1) Asset index—being in the lowest quintile for the population at enrolment (calculated with principle components analysis using data on mother, household demographics and animal & other asset ownership) 2) Low parental education–no education or primary-schooling only (asked at enrolment) 3) Father occupation—father did not work, was seasonably employed or was a casual labourer at 12-month assessment 4) Mother married under the legal age of 18 years (reported at 12 month assessment) 5) Family debt—mother reported family debt or being unable to afford to buy food for herself or her child at any point between becoming pregnant and the 12 month assessment. Group 2—Maternal stress—Consists of six factors: 1) Death of one or more of mother’s close family members since becoming pregnant reported at 12-month assessment 2) Mother seriously injured or ill 3) Any violence towards mother from husband (assessed using WHO multi-country study on women’s health and domestic violence against women[20]) or any other person since becoming pregnant reported at 12-month assessment 4) mother screens positive for mild, moderate or severe depression on PHQ9 or answers ‘yes’ to PHQ9 question on suicidal ideation (at 12-month assessment). PHQ9 is one of the most commonly used screening tools for depression and has been used widely in India[21] 5) Low level of support or high stress from others around the mother using the Duke social support & stress scale[22] reported at 12-month assessment 6) Problematic husband alcohol use reported by mother at 12-month assessment Group 3—Relationship—Consists of four factors: 1) Any family member was unhappy when they found out that the child was a girl 2) Moderate or high concern level on Mother Object Relations Scale–short form (MORS-SF) at 12-month assessment. MORS-SF is a screening tool consisting of 14 short statements which a mother is asked to rate on a Likert-type scale to identify potential problems in early mother-infant relationship[23] 3) Very low quality interactions observed during a feeding episode at the 12-month assessment (assessed by non-specialist fieldworkers using the observed feeding index, a tool developed in this project where feeding is scored using tick-boxes. This tool will be published in due course). Very low quality means that the following was observed during the feeding episode: < = 1 positive talk by mother towards child, and < = 1 episodes of playful feeding and < = 1 responsive feeding actions, plus one or more negative actions such as force feeding, holds child’s head still to give food, shaking, threatening, shouting or berating observed by the mother towards child during feeding session 4) Lowest quintile score on HOME inventory measuring quality of the home environment through observations of the home and questions to the mother (total of 45 items, each scored 0 or 1) over the course of one hour [24] at 12 month assessment–the cut-off for the quintile fell between 27 & 28 points and the lowest of these (27 points) was chosen to create a conservative estimate of this factor. Group 4—Child—Consists of six factors: 1) Child born prematurely (asked at 12-month assessment) 2) Child hospitalised in first year of life 3) Separation of mother & child for more than a week in the first year of life 4) Inadequate care–child left alone or with a child under 10 years for more than one hour in the past week (assessed at 12-month assessment) (From [25]) 5) Older children in the house say anything to make child cry or unhappy (in last week) (at 12-month assessment) (From [25]) 6) Older children who live in house: hit/punched/kicked/bit child on purpose to make them unhappy (in last week) (assessed at 12 month assessment) A systematic cultural adaptation process based on Khan & Avan[26] was used. This comprised of six steps and aimed to ensure that each item was assessing the construct it was attempting to understand. Each item was first written in English, and the process of adaptation was: 1) Translation into Hindi independently by two trained research associates 2) Comparing these translations & assessing technical equivalence of these, then producing final translations by consensus for testing 3) Field research with project staff and mothers of young children to test understanding of translations and to improve them 4) Finalisation of tool for pretesting 5) Pretesting in the community to assess usability, 6) Assessor training, establishing inter-rater reliability and Pilot-testing. For the 10 mothers with twins, questions relating to the child (e.g child hospitalisation) were asked for each child, and those relating to the mother herself (e.g maternal depression) were asked only once and answers applied to each child. We trained assessors to do child development assessments at 18 months of age using the motor, cognitive & language scales of the Bayley Scales of Infant Development 3rd Edition (BSID-III) in the home[27]. Assessors did two BSID-III assessments per day in pairs. Each assessment took 2–3 hours to complete. Each BSID-III scale consists of a series of progressively more difficult activities which children are asked to do whilst interacting with an assessor. Each item was scored 1 if the activity was demonstrated, otherwise it was scored 0. Assessment on each scale started at the item marked ‘K’ (start point for 16.5–19.5 month old children). Children not able to achieve three activities at that level were assessed as far as two levels back (the item marked ‘I’, which is the start point for 11–13.5 month old children) before the assessment was stopped. The assessment on each scale ended when the child scored 0 on five consecutive activities. We did comprehensive cultural adaptation and inter-rater reliability (IRR) checks, finding mean agreement between assessors of greater than 97%. The same fieldworkers did anthropometrical measurements of children. Weight was measured to the nearest 0.01Kg using SECA-384 electronic scales which were calibrated weekly. Weighing was ideally done with the child’s clothes removed. If this was not possible, the child was weighed fully-clothed, then the clothes were removed and weighed. The difference between the weight of the fully-clothed child and the weight of the clothes was calculated to give the child’s weight. Length was measured to the nearest 0.1cm using the SECA-417 infantometer by two assessors as follows. The child was laid down on the infantometer board. The first assessor cupped their hands over the child’s ears and held the head against the end of the measurement board. The second assessor then ensured that the child’s body was straight on the board, placed one hand on the child’s legs to stabilise them and brought the footpiece upwards towards the child’s feet which were held perpendicular to the board. This assessor then read aloud the length board reading and this was recorded by the first assessor. There were therefore three development outcomes & two growth outcomes assessed at 18 months of age. One sample size calculation was done for the whole SPRING-ELS substudy, and the minimal sample size was exceeded in the work presented in this paper. A minimum sample size of 25 children per cluster was chosen for the overall SPRING-ELS substudy to give 90% power at the 5% level of significance to explore a range of adversities with prevalence of 20% to 80% and to detect effect sizes between 0.4SD & 0.5SD (assuming an intra-cluster correlation of 0.05) using an established formula[28]. Table 1 shows that five adversities had missing data. Four were missing less than 4%, and the fifth was missing 32.8%. We assumed these were missing-at-random and used multiple imputation by chained equations (MICE)[29], including all explanatory and outcome variables in each analysis. We used 30 imputations. We calculated descriptive data using a combination of all imputations. We summed the 22 adversities described to create a total adversity score following a cumulative-adversity model [30–32] which recognises that children can be resilient to single adversities, but that combinations of these may be more harmful and overwhelm protective factors in a child’s life. In addition to this overall measure, we summed adversities within each of the four categories. This gave a total of five primary explanatory variables. To ensure that summing adversities in this manner was not ‘double counting’ adversities (because children who had one adversity may be more likely to have several other related adversities) we used principle-components-analysis (PCA) to capture the linear combination of adversities which creates the maximum variance in the data, a similar manner to calculation of wealth indices[33]. We converted the raw PCA score into adversity quintiles giving five groups of children for analysis, giving a sixth explanatory variable. Raw scores for the BSID-III scales were converted to composite scores for each child following the BSID-III manual, based on the child’s age at assessment. This is done because BSID-III scores change quickly with age at this stage of development. We converted child length and weight to height-for-age and weight-for-age z-scores using the zscore06 package for Stata15[34] which is based on the 2006 WHO child growth standards[35]. Using z-scores for these growth measures allowed child length & weight to be compared with international standards based on healthy breastfed children who on a population-level grow with the same distribution and trajectories wherever in the world they live. We used Stata 15 for all statistical analyses (StataCorp LLC: College Station, TX, USA). We used mixed-effects linear regression, accounting for trial cluster as a random effect and trial arm allocation as a fixed effect to calculate the adjusted mean growth and development values at each level of cumulative adversity and adversity quintile. This allowed us to examine the change in these outcomes as children were exposed to incrementally greater adversity. Because only 4% of children had a total adversity score of nine or more, we combined these with the children with a score of 8 to create an 8+ group for this analysis. Scatter diagrams suggested a linear relationship and we next we created models treating each of adversity score and adversity quintile as continuous variables to calculate the change in each of the five outcomes for a one-unit change in cumulative adversity. We used a similar model to explore the relationship between each of the four individual adversity domains and outcomes adjusted only for clustering and trial arm allocation. We then analysed the four domains together in a mutually adjusted multivariate model to examine the interrelationships between them with respect to outcomes. Ethics approval was obtained from the London School of Hygiene & Tropical Medicine research ethics committee (SPRING: 23 June 2011, approval number 5983; SPRING-ELS substudy 19 May 2015, approval number 9886) and the Sangath Institutional Review board (IRB) (SPRING: 19 February 2014; SPRING-ELS substudy 27 May 2015). Approval was also granted by the Indian Council of Medical Research’s Health Ministry Screening Committee (HMSC) (SPRING: 24 November 2014; SPRING-ELS substudy: 6 October 2015). The SPRING trial is registered with clinicaltrials.gov, number {"type":"clinical-trial","attrs":{"text":"NCT02059863","term_id":"NCT02059863"}}NCT02059863. Informed written consent was obtained from mothers at enrolment into the trial surveillance system and again before a child’s first birthday for detailed developmental assessments. The work was funded by the Wellcome Trust through two awards: a Wellcome Trust Research Training Fellowship to Sunil Bhopal (107818/Z/15/Z) & a Wellcome Trust Strategic Award for the SPRING Programme (0936115/Z/10/Z) for which Betty Kirkwood is the principle investigator. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. SB & BK have complete access to the study data and are responsible for the reported study findings, and made the decision to submit for publication.