Effect of a home-based health, nutrition and responsive stimulation intervention and conditional cash transfers on child development and growth: A cluster-randomised controlled trial in Tanzania

Introduction Evidence on the effects of community health worker (CHW) interventions and conditional cash transfers (CCTs) on child growth and development in sub-Saharan Africa remains sparse. Methods We conducted a single-blind, cluster-randomised controlled trial of an integrated home-visiting health, nutrition and responsive stimulation intervention alone and in combination with CCTs to promote antenatal and child clinic attendance from 2017 to 2019 in rural Morogoro Region, Tanzania. Pregnant women and caregivers with a child <1 year of age were enrolled. Twelve villages were randomised to either (1) CHW (n=200 participants), (2) CHW+CCT (n=200) or (3) control (n=193). An intention-to-treat analysis was conducted for the primary trial outcomes of child cognitive, language and motor development assessed with the Bayley Scales of Infant and Toddler Development and child length/height-for-age z-scores (HAZ) at 18 months of follow-up. Results The CHW and CHW+CCT interventions had beneficial effects on child cognitive development as compared with control (standardised mean difference (SMD): 0.15, 95% CI 0.05 to 0.24, and SMD: 0.18, 95% CI 0.07 to 0.28, respectively). The CHW+CCT intervention also had positive effects on language (SMD: 0.08, 95% CI 0.01 to 0.15) and motor (SMD: 0.16, 95% CI 0.03 to 0.28) development. Both CHW and CHW+CCT interventions had no effect on HAZ in the primary analysis; however, there were statistically significant positive effects in multivariable analyses. The CHW+CCT group (mean difference: 3.0 visits, 95% CI 2.1 to 4.0) and the CHW group (mean difference: 1.5 visits, 95% CI 0.6 to 2.5) attended greater number of child health and growth monitoring clinic visits as compared to the control group. Conclusion Integrated CHW home-visiting interventions can improve child cognitive development and may have positive effects on linear growth. Combining CHW with CCT may provide additional benefits on clinic visit attendance and selected child development outcomes. Trial registration number ISRCTN10323949. We conducted a single-blind, cluster-randomised controlled trial of a home-based CHW-delivered health, nutrition and responsive stimulation intervention alone and combined with CCTs to promote antenatal care and routine well-child visits in rural Ifakara, Tanzania. The full trial protocol is detailed elsewhere.19 This report presents the effect of the interventions on the primary outcomes of the study (endline child development and HAZ), and we also present the secondary anthropometric outcomes of WAZ and weight-for-height z-scores (WHZ). We plan to present the results for all other secondary trial outcomes in forthcoming reports. The trial was conducted in 12 selected villages in the Ifakara Health Institute Health and Demographic Surveillance System (HDSS) in Kilombero and Ulanga districts in Morogoro Region.20 The Ifakara HDSS area is located approximately 450 km by road from Dar es Salaam and has a population of ~400 000 people. The HDSS area is predominately rural, and most residents are subsistence farmers. A recent study conducted among children 18–36 months of age in the study area found a 16.9% prevalence of low birth weight (< 2500 g) and a 36.2% stunting prevalence.21 The trial recruited pregnant women and mother/caregiver–infant pairs who lived in the study villages through a house-to-house survey. The inclusion criteria were (1) permanent residence in a study village, (2) pregnant (self-reported) or had a child <1 year of age at the enrolment visit and (3) provided written consent. The exclusion criteria were (1) enrolment in any other clinical trial or intervention study or (2) child with severe physical or mental impairment. Potential participants were not aware of their village’s randomisation arm at the time of seeking consent. In each study village, enrolment continued until all pregnant women and mothers/caregivers with a child <1 year of age were enrolled or until 50 participants were enrolled, whichever was reached first. If the mother had twins, one child was randomly selected for the trial and the same child was assessed at each time point. Written informed consent was obtained from all participants. Participants were referred to clinics at the time of enrolment and during outcome assessment visits if the research assistant identified an emergency maternal or child health issue or if the mother reported immediate risk of self-harm in the depression assessment. Village clusters were randomly allocated in a 1:1:1 ratio to one of the three trial arms: (1) CHW, (2) CHW+CCT or (3) control. The full details of the CHW, CHW+CCT and control arms are described in the trial protocol and online supplemental table 1.19 Briefly, the CHW intervention was the same in the CHW and CHW+CCT arms and therefore the CCT was the only difference between the two intervention arms. The control arm had access to the existing clinic-based maternal and child services (no CHW or CCT). Randomisation was stratified by semiurban (six villages) and rural (six villages) villages to increase the likelihood of baseline balance between arms. Randomisation was done by a non-study statistician using a computer-generated randomisation list with sequence blocks of three. bmjgh-2021-005086supp001.pdf The same CHW intervention was delivered in the CHW and CHW+CCT arms. CHWs delivered an integrated health, nutrition and responsive stimulation intervention in the home every 4–6 weeks for the trial duration of 18 months. Online supplemental table 1 presents a full description of the CHW intervention.22 Briefly, the trial CHWs received the year-long basic training on the national curriculum by the government before being hired by the project. The national CHW programme was not implemented in the study area during the trial due to resource constraints, and the CHW salaries were paid by the project; the monthly salary for the CHWs was ~600 000 Tanzanian shillings (~$230). Each CHW was assigned and delivered the intervention to two villages (~100 participants). There was no CHW turnover over the course of the trial and therefore participants received the intervention from the same CHW for the duration of the trial. The CHW intervention included activities and duties of the standardised Tanzanian CHW curriculum with the addition of a responsive stimulation component (online supplemental table 1).19 The CHW intervention’s main maternal and child components included (1) identification and referral for under-5 childhood illness per the Integrated Management of Childhood Illness; (2) antenatal and postnatal pregnancy, delivery and essential newborn care counselling and danger signs identification; (3) family planning; and (4) emergency and routine referrals to facilities. The CHWs will be assigned to the study villages at the start of the trial. The CHWs provided counselling and referrals but did not directly provide treatments, medicines, nutritional supplements and immunisations or provide child growth monitoring. The Tanzanian government CHW curriculum does not include responsive stimulation or other direct early child development promoting activities. The study team adapted the UNICEF and WHO Care for Child Development package to the local context and provided a 1-week training in September 2017 on the responsive stimulation intervention that included integrated classroom and practical sessions.23 The responsive care component of the CHW intervention included essential early childhood development knowledge, promotion of caregivers’ sensitivity and responsiveness, and promotion of developmentally appropriate play and communication activities, toy making and problem-solving. In addition, caregivers tried responsive stimulation activities with their young child and received feedback and coaching from the CHW. One field coordinator supervised CHWs through biweekly one-on-one meetings with each CHW, a monthly meeting with all CHWs, and monthly home visit spot checks where the field coordinator accompanied CHWs to home visits (~5% of visits). A CCT intervention was also provided to participants in the CHW+CCT group every 4–6 weeks at the time of the CHW visit. The conditions for the CCT were attendance of routine antenatal care or routine well-child health and growth monitoring clinic visits. The CCT was intended to increase access and utilisation of antenatal and child health services, including child growth monitoring, treatment of health conditions and other interventions not provided by the CHWs. CHWs assessed antenatal care and child health cards at each home visit and provided mothers with cash payments of 10 000 Tanzanian shillings ($4.30) per antenatal care visit or 5000 Tanzanian shillings ($2.15) per routine child health and growth monitoring visit that was completed since the last study visit. The average daily income per person for smallholder farmers in Tanzania is US$1.90.24 CHWs communicated that the CCT payments could be used in any way without penalty, but suggested that mothers use the money for resources to support the health and development of the child. Independent fieldworkers who were blinded to the randomised arm conducted home interviews with mothers at enrolment (baseline) and at 18 months after enrolment (endline). The fieldworkers were randomly assigned to villages each survey round and were not to ask participants about the intervention they received. The baseline visit occurred before implementation of the intervention, so it was not possible for the fieldworkers to know the randomised group. Nevertheless, it cannot be ruled out that at the endline visit fieldworkers came to know the intervention status of a household from conversations with the mother. Standardised questionnaires were administered to collect demographic and socioeconomic data. In addition, the Hopkins Symptom Checklist (HSCL-25) was administered to assess symptoms of depression and anxiety25; symptoms consistent with depression were defined using the Tanzanian validated HSCL-25 cut-off.26 Functional social support was assessed using the Duke University-University of North Carolina Functional Social Support Questionnaire.27 The Caregiver Knowledge of Child Development Inventory (CKCDI) was administered at baseline.28 Child anthropometric measures were taken in triplicate in the home at baseline and endline visit. Child weight was measured to the nearest 100 g using a digital scale (Seca, Hamburg, Germany). Child length (children <24 months of age) was measured to the nearest 0.1 cm using a length board (Seca), and child height was measured to the nearest 0.1 cm using a stadiometer (Seca). Anthropometric z-scores were calculated using the 2006 WHO Child Growth Standards.29 At endline, the fieldworkers administered process evaluations to each group. At endline, fieldworkers travelled outside the study area to conduct interviews and take anthropometric measurements for participants who moved temporarily or permanently outside the study area. Female research nurses who were blinded to randomisation arm administered a Tanzania-adapted and Swahili-translated version of the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III).30 31 The BSID-III was administered in quiet rooms at two health facilities serving the study area. The BSID-III nurses completed a 3-week BSID-III training led by experts from Boston, USA and completed the BSID-III assessments for a prior research study.30 For participants who moved outside of the study area, BSID-III assessments were not conducted. The two assessors differed in mean BSID-III domain composite scores (online supplemental table 2) and therefore all analyses were adjusted for assessor. However, the BSID-III showed high internal consistency for all domains (Cronbach’s alphas ≥0.91) in the full sample and separately for each assessor (online supplemental table 3). Sample size calculations were based on randomisation of 12 village clusters, 50 mother/caregiver–child pairs per cluster, a nominal type I error rate of 0.05 and an intracluster correlation of 0.01. We assumed 7.5% fetal loss or child death, 5% loss to follow-up (unknown vital status) and 15% missing data on length-for-age z-score (LAZ). For child development, we originally planned to randomly select 60% of participants to have the BSID-III assessed, of which we assumed 10% would not complete the assessment (resulting in approximately 50%). Based on these assumptions, we had 80% power to detect a standardised mean difference (SMD) of 0.40 in HAZ and 0.53 SD of development z-scores. However, to increase statistical power, we decided to invite all participants for BSID-III assessment. In a post-hoc power analysis based on the actual number of BSID-III assessments and observed within village correlation, we had 80% power to detect effect sizes of 0.48, 0.86 and 0.46 SD for cognitive, language and motor scores, respectively. The magnitude of the detectable differences for linear growth and development was large; however, this was a proof-of-concept efficacy trial and we hypothesised that there was potential for the intervention package and the combination of CHWs and CCT to provide a relatively large impact on child growth and development. The intention-to-treat (ITT) principle was used for all primary analyses; participants who moved to neighbouring villages were analysed in the village originally randomised. All analyses accounted for clustering by village and urban/rural residence due to the stratified randomisation scheme. Generalised linear regression models were used to assess the effect of the CHW and CHW+CCT interventions on the primary outcomes at endline: BSID-III subscale z-scores and HAZ. BSID-III z-scores were calculated using the internal mean and SD. For the primary minimally adjusted analysis, BSID-III z-scores were adjusted for child age, sex and BSID-III assessor; HAZ similarly adjusted for child age and sex. We also present BSID-III composite scores as a secondary outcome; however, applying US norms for BSID-III scores to children in other settings can result in misclassification and cross-cultural bias.32 33 The secondary outcomes of child WAZ, weight-for-length/height z-scores (WHZ), and the number of child health and growth monitoring clinic visits were also assessed using generalised linear models. We did not analyse antenatal care visit attendance by randomised group since only 30% of women were pregnant at baseline and 60% of enrolled pregnant women delivered within 60 days of enrolment. Log-Poisson models were used to examine relative risks of stunting (HAZ <−2), wasting (WHZ <−2), underweight (WAZ <−2) and overweight (WHZ >2). The Benjamini-Hochberg procedure, which is a preferred method for accounting for multiple testing of correlated outcomes, was used to control for the potential false discovery rate for the eight primary analysis tests.34 35 Benjamini-Hochberg procedure-adjusted p values <0.05 were considered statistically significant. We conducted sensitivity analyses that (1) adjusted for baseline factors which showed some degree of imbalance between randomisation arms based on p<0.20, and (2) used stabilised inverse probability weights to account for dependent censoring (ie, loss to follow-up).36 We also present the effect estimates collapsing the CHW and CHW+CCT intervention arms. In addition, we explored the potential for effect modification by predefined baseline variables.19 The statistical significance of interaction was assessed with the likelihood ratio test and we did not adjust for multiple testing. Statistical analyses were performed with Stata V.16.