The majority of the evidence about the effectiveness of early parenting and nutrition interventions pertains to 1 targeted index child in a given household. We evaluated whether nontargeted sibling children benefited from a bundled parenting and nutrition intervention.
We designed a sub-study within a broader cluster-randomized trial that evaluated the effects of engaging both mothers and fathers and bundling parenting and nutrition interventions in Mara, Tanzania. Trained community health workers delivered interventions to parents through peer groups and home visits. Interventions encompassed various content including responsive parenting, infant and young child feeding, and positive couples’ relationships. The main trial enrolled mothers and fathers and 1-index children <18 months of age in 80 clusters. Between June and July 2021, in 32 clusters (16 intervention, 16 control), we reenrolled 222 households (118 intervention, 104 control) from the main trial that had another child <6 years of age (ie, sibling to the index child). We compared caregiving practices and child development and nutrition outcomes among siblings in intervention versus control households.
Compared with control siblings, intervention siblings had improved expressive language development (β = 0.33 [95% confidence interval: 0.03 to 0.62]) and dietary intake (β = 0.52 [0.10 to 0.93]) and reduced internalizing behaviors (β = −0.56 [−1.07 to −0.06]). Intervention caregivers reported greater maternal stimulation (β = 0.31 [0.00 to 0.61]) and paternal stimulation (β = 0.33 [0.02 to 0.65]) and displayed more responsive caregiving behaviors (β = 0.40 [0.09 to 0.72]) with sibling children.
A father-inclusive, bundled parenting and nutrition intervention can achieve positive spillover effects on sibling children’s developmental and nutritional outcomes.
What’s Known on This Subject:
Bundled parenting and nutrition interventions are effective for improving child development and nutritional outcomes. However, most evaluations have focused on outcomes of 1 index child in a given household, thereby ignoring whether nontargeted sibling children also benefit from such interventions.
What This Study Adds:
A father-inclusive, bundled parenting and nutrition intervention improved dietary diversity, expressive language development, and internalizing behaviors among sibling children in intervention households, as well as mothers’ and fathers’ parenting practices with the sibling child.
Globally, 43% of children <5 years of age are failing to meet their developmental potential.1 Inadequate stimulation and nutritional deficiencies are key risk factors to healthy early child development (ECD).2 To simultaneously address these common early childhood risks, integrated parenting and nutrition interventions (eg, counseling on responsive caregiving alongside complementary feeding) have become prioritized in low- and middle-income countries (LMICs) as a strategy for improving both children’s developmental (eg, cognitive development) and nutritional status (eg, diet).3
Although this evidence base has expanded rapidly over the past decade, the vast majority of trials to date have evaluated effects on 1 index child.4–6 Yet social and behavior change interventions to support optimal caregiving for young children do not operate in a vacuum restricted to the caregiver-child dyad but instead have the potential to influence other nontargeted individuals, including other sibling children in the household.7,8 One pathway through which caregiving interventions can achieve positive spillover effects on sibling children’s outcomes is via improved caregiving practices that participants may newly extend with their multiple children. Based on social learning and family systems theory, parents may apply caregiving advice learned originally for the targeted index child with other nontargeted children especially when they experience success.9,10
Another potential pathway for sibling spillover effects could be through the influences that index children may have directly on their siblings. Theoretically, program-related improvements in index children’s early developmental skills (eg, language development, prosocial behaviors) could benefit their siblings if siblings began to model the positive behaviors acquired by the index child or through more enriched and improved relationship quality between siblings.11
Finally, both channels could have even greater potential for intrahousehold spillover effects in interventions involving couples (ie, both mothers and fathers) compared with those targeting 1 caregiver. These program impact pathways can operate potentially by not only increasing both mothers’ and fathers’ likelihood of extending positive behaviors with multiple children but also improving broader family caregiving relationships, such as coparenting and partner support, that benefit all children and the overall family environment.12
The authors of relatively few studies have evaluated whether caregiver interventions have spillover effects on nontargeted sibling children, and the global evidence to date is mixed. Although some studies have revealed certain positive spillover effects on maternal feeding practices,13 dietary intake,14 growth,15 and developmental outcomes of sibling children,16 other studies have documented null results across various sibling child outcomes.17,18 The observational assessments of caregivers’ interaction quality with children are limited in this particular evidence base exploring intervention spillover effects on sibling children. Moreover, most of this evidence comes from interventions in the United States, with little known from LMICs. Yet this question about the presence of intervention spillover effects on other children in the household is especially relevant in LMICs, in which there is a substantial burden of poor nutrition and development across the life course and in which families, on average, have more children in shorter birth intervals.19
To address this gap, we leveraged a cluster randomized controlled trial (RCT) that was originally aimed to evaluate the effects of engaging mothers and fathers in a bundled parenting and nutrition intervention on developmental and nutrition outcomes for 1 index child per household in rural Tanzania. We added a sub-study within this trial and reenrolled households in which there was a sibling to the index child. We explored whether there were intervention spillover effects on child development, nutrition, or family caregiving outcomes involving sibling children who were not explicitly targeted in the intervention but resided in study households.
Methods
The Engaging Fathers for Effective Care for Nutrition and Development in Tanzania Study
The Engaging Fathers for Effective Care for Nutrition and Development in Tanzania Study (EFFECTS) was a cluster-RCT that was aimed to evaluate the independent and combined effectiveness of engaging fathers versus mothers only and a bundled parenting and nutrition intervention versus a nutrition intervention. The coprimary outcomes were to index children’s dietary diversity and developmental outcomes. Full details about the EFFECTS study design and interventions are described elsewhere. Briefly, 80 villages were randomly selected in the Mara region of Tanzania and allocated to 1 of 5 study arms. Within each village, 12 households with an index child <18 months of age and in which both mother and father co-resided were randomly selected for enrollment. In total, 960 mother-father-index child triads were enrolled.
EFFECTS interventions consisted of 24 total sessions that were delivered by government community health workers. Interventions involved mostly peer-group sessions that convened peer groups of mothers and additionally engaged fathers (ie, couples) in the father-inclusive interventions. All interventions incorporated social and behavior change techniques to primarily promote optimal infant and young child feeding practices for the index child but also covered appropriate agriculture, water, sanitation, and hygiene practices, parental mental health, gender equity, and positive couple’s relationship dynamics. The bundled parenting and nutrition interventions further integrated messages pertaining to ECD milestones, involved interactive sessions to coach parents on responsive caregiving through play and communication activities with the index child, and included content on child behavior management.
Results from the parent EFFECTS study are reported elsewhere.20,21 All interventions were delivered with high fidelity. For example, in the study arm in which the bundled intervention was offered to couples, average attendance in a given peer group session was 8.8 mothers and 8.3 fathers out of the 12 enrolled households per village. Compared with the nutrition interventions, the bundled parenting and nutrition interventions improved index children’s cognitive and receptive language development, as well as maternal stimulation with the index child, home learning materials, and reduced maternal harsh discipline, maternal intimate partner violence victimization, and paternal parenting distress. Compared with interventions with mothers only, interventions that also engaged fathers improved paternal stimulation with the index child and involvement in domestic chores. Significant interaction effects were detected between interventions for improving index children’s dietary diversity and fathers’ gender-equitable attitudes.
EFFECTS Sibling Spillover Sub-Study
For this sub-study, we focused specifically on the bundled parenting and nutrition intervention offered to couples for its comprehensiveness and compared it with the control. After completing the endline evaluation for the parent EFFECTS study, we revisited households in these 2 selected study arms and reenrolled those that had a sibling to the index child into this EFFECTS Sibling Spillover Sub-Study (EFFECTS-Sibs). Households were selected if there was a sibling child <6 years of age who shared the same mother and father as the index child and resided in that household. If there were multiple eligible siblings, we prioritized the youngest child to maximize the analytic sample of children who could be assessed about not only child development (0–6 years) but also nutrition outcomes (6–23 months). If there were sibling twins, we randomly selected 1.
Data for EFFECTS-Sibs were collected in June 2021 or 9 months after the completion of the interventions. All measures were translated into Kiswahili, back-translated to English to ensure appropriateness, reviewed by bilingual study team members, and pretested before data collection. After an 8-day training, a team of 6 university-educated enumerators administered surveys to 1 primary caregiver of the child per household and administered a child development assessment with the sibling child in a private setting outside of the household. Although enumerators were independent of the implementation team, blinded to intervention assignment, and unaware of the intervention components, complete masking was not possible because of the nature of the intervention, and some caregivers could have disclosed their participation in the program. Caregivers provided informed written consent for participation in EFFECTS-Sibs. This study received approvals from the institutional review boards at the Tanzania National Institute for Medical Research and the Harvard T.H. Chan School of Public Health.
Measures
Child Development
The primary outcome was sibling children’s development, which was assessed by using the Mullen Scale for Early Learning tool.22 The Mullen Scale is an observational tool that can be applied to children aged 0 to 6 years. The original Mullen Scale provides 5 scale scores: visual reception, fine motor, gross motor, receptive language, and expressive language. We assessed 4 out of the 5 scales and excluded gross motor development to reduce the overall time of assessment and because we determined that it was the least relevant domain to the EFFECTS intervention. Internal consistency was acceptable for children’s visual-receptive (α = 0.83), fine motor (α = 0.80), receptive language (α = 0.83), and expressive language development (α = 0.81). Mullen raw scores were internally age-standardized for analysis. For any child >2 years of age, caregivers also reported on child behavioral development using the Strengths and Difficulties Questionnaire.23 We calculated subscale scores for children’s prosocial behavior (α = 0.82), internalizing problems (α = 0.62), and externalizing problems (α = 0.61).
Child Nutrition
For any child aged 6 to 23 months, caregivers also reported on child dietary diversity, or the number of food groups consumed out of 8 total food groups in the previous day per World Health Organization guidelines for infant and young child feeding indicators.24 We also calculated the proportion of children aged 6 to 23 months achieving minimum dietary diversity, or those who consumed ≥5 out of 8 food groups.24
Family Caregiving
One primary caregiver per household (ie, mostly mothers) reported on maternal (α = 0.78) and paternal stimulation practices (α = 0.82), or how often in the past week each parent engaged in early learning activities with the sibling child, using an adapted version of the Family Care Indicators.25 Maternal (α = 0.82) and paternal harsh discipline (α = 0.85) were indicated by whether each parent used any harsh disciplinary practices against the child in the past month.26 In addition to caregiver-reported measures of parenting, we used an observational measure to assess caregiver-child interactions among children >6 months of age. Caregivers were provided with a picture book and instructed to use it however they wished for a 4-minute assessment. Enumerators video-recorded caregiver-child interactions, and the behaviors of both the caregiver and child were later coded on a 5-point Likert scale by using the Observation of Mother Child Interactions tool (α = 0.89).27 Finally, to measure sibling relationship quality, caregivers rated on a 4-point Likert scale how often the index and sibling child interacted with each other in positive (α = 0.85) and negative manners (α = 0.84) using a tool adapted from the Sibling Inventory of Behavior Items28 and the Maternal Interview Concerning Children’s Sibling Relationships.29
Analysis
First, we assessed the balance between the EFFECTS-Sibs intervention and control samples by comparing sociodemographic characteristics collected from these households at the time of the EFFECTS baseline evaluation. Then, to estimate the difference between intervention and control villages, we used mixed effects linear regression models with a specification for restricted maximum likelihood for continuous outcomes and mixed effects logistic regression models for binary outcomes. All models included an indicator variable for village treatment assignment (ie, intervention vs control villages), with a fixed effect for district and a random effect for village with an exchangeable covariance structure (ie, 1 common variance for all random effects and 1 common pairwise covariance). Each outcome was modeled separately, and all regression analyses were based on intention-to-treat. Unadjusted analyses controlled only for sibling age and included dummy variables for child development assessor (when estimating effects on Mullen scores) or type of caregiver respondent (eg, mother or father, for all other caregiver-reported outcomes). Fully adjusted analyses additionally controlled for sibling child sex, absolute age difference between index and sibling child, an indicator for whether the sibling was younger or older than the index child, whether the sibling was a singleton birth, maternal and paternal education, maternal and parental age, marital relationship, household size, wealth quintile, and household food insecurity access. Models for dietary outcomes were also adjusted for breastfeeding in the previous day. Finally, we conducted a sensitivity analysis using inverse probability weights to reconstruct the study population and account for households with a known sibling at the time of data collection but were not represented or those that had missing data for a given outcome. All statistical analyses were performed in Stata 15.1.
Results
We reenrolled 32 clusters (16 intervention, 16 control) and 222 households (118 intervention, 104 control) into EFFECTS-Sibs, which represented 58% of the households that were originally enrolled in these study arms for the parent EFFECTS trial (Fig 1). Baseline sociodemographic characteristics were generally balanced between the intervention and control groups in EFFECTS-Sibs (Table 1), except that mothers and fathers were more educated and households were less food insecure in the intervention group, which were similarly observed between groups in the parent EFFECTS trial.
Participant flowchart of the EFFECTS-Sibs nested within the EFFECTS parent trial, Tanzania, 2018–2021.
Participant flowchart of the EFFECTS-Sibs nested within the EFFECTS parent trial, Tanzania, 2018–2021.
Baseline Characteristics of Households Reenrolled Into the EFFECTS-Sibs
| Control | Interventiona | P | |
|---|---|---|---|
| Study characteristics | n = 104 | n = 118 | |
| Index child age, mo | 11.8 [±5.9] | 11.3 [±6.0] | .59 |
| Female index child | 52 (51.0) | 55 (46.6) | .52 |
| Maternal age, y | 29.8 [±5.8] | 29.4 [±6.7] | .65 |
| Paternal age, y | 38.1 [±9.2] | 37.8 [±9.8] | .80 |
| Maternal education | |||
| No schooling | 9 (9.3) | 5 (4.5) | .05 |
| Pre-primary to standard 6 | 6 (6.2) | 4 (3.6) | |
| Standard 7 | 74 (76.3) | 91 (81.3) | |
| Form 1 to completed university | 8 (8.3) | 12 (10.7) | |
| Paternal education | |||
| No schooling | 7 (7.1) | 1 (0.9) | .01 |
| Pre-primary to standard 6 | 7 (7.1) | 10 (8.9) | |
| Standard 7 | 75 (76.5) | 79 (70.5) | |
| Form 1 to completed university | 9 (9.2) | 22 (19.6) | |
| Number of children <5 y | 2.3 [±0.9] | 2.2 [±1.0] | .48 |
| Household size, number | 7.5 [±2.5] | 7.2 [±2.8] | .31 |
| Marital status: married and monogamous or cohabitating | 84 (82.4) | 95 (80.5) | .73 |
| Poorest wealth quintile | 22 (22.5) | 24 (20.9) | .82 |
| Household food insecurity access | |||
| Food secure | 6 (5.9) | 18 (15.3) | .07 |
| Mildly food insecure | 16 (15.7) | 22 (18.6) | |
| Moderately food insecure | 33 (32.4) | 28 (23.7) | |
| Severely food insecure | 47 (46.1) | 50 (42.4) |
| Control | Interventiona | P | |
|---|---|---|---|
| Study characteristics | n = 104 | n = 118 | |
| Index child age, mo | 11.8 [±5.9] | 11.3 [±6.0] | .59 |
| Female index child | 52 (51.0) | 55 (46.6) | .52 |
| Maternal age, y | 29.8 [±5.8] | 29.4 [±6.7] | .65 |
| Paternal age, y | 38.1 [±9.2] | 37.8 [±9.8] | .80 |
| Maternal education | |||
| No schooling | 9 (9.3) | 5 (4.5) | .05 |
| Pre-primary to standard 6 | 6 (6.2) | 4 (3.6) | |
| Standard 7 | 74 (76.3) | 91 (81.3) | |
| Form 1 to completed university | 8 (8.3) | 12 (10.7) | |
| Paternal education | |||
| No schooling | 7 (7.1) | 1 (0.9) | .01 |
| Pre-primary to standard 6 | 7 (7.1) | 10 (8.9) | |
| Standard 7 | 75 (76.5) | 79 (70.5) | |
| Form 1 to completed university | 9 (9.2) | 22 (19.6) | |
| Number of children <5 y | 2.3 [±0.9] | 2.2 [±1.0] | .48 |
| Household size, number | 7.5 [±2.5] | 7.2 [±2.8] | .31 |
| Marital status: married and monogamous or cohabitating | 84 (82.4) | 95 (80.5) | .73 |
| Poorest wealth quintile | 22 (22.5) | 24 (20.9) | .82 |
| Household food insecurity access | |||
| Food secure | 6 (5.9) | 18 (15.3) | .07 |
| Mildly food insecure | 16 (15.7) | 22 (18.6) | |
| Moderately food insecure | 33 (32.4) | 28 (23.7) | |
| Severely food insecure | 47 (46.1) | 50 (42.4) |
Data are either mean [standard deviation] or n (%).
This sub-study focused specifically on 1 of the 4 interventions tested in the main EFFECTS trial (ie, the most comprehensive intervention that bundled parenting and nutrition content and engaged both mothers and fathers).
Approximately one-half of enrolled sibling children were girls (53.6%), with an average age of 24.1 months (range: 0.4–77 months). Three-quarters (76.6%) were younger siblings who were 12.4 months of age, on average, whereas older siblings were 63.2 months of age. The absolute mean age difference between EFFECTS-Sibs and EFFECTS sibling pairs was 28.0 months. Four of the siblings (1.8%) were nonsingleton births (ie, twins). Among older siblings aged 3 to 6 years, two-fifths (40.0%) attended an early childhood education program. Caregiver respondents for EFFECTS-Sibs were mostly the child’s mother (91.0%), with a small proportion of fathers (5.0%) or other primary caregivers (4.1%; eg, grandparent, eldest sibling).
Child Development
Compared with siblings in the control, intervention sibling children had higher adjusted expressive language development scores (β = 0.33 [95% confidence interval: 0.03 to 0.62]) and reduced internalizing problems (β = −0.56 [−1.07 to −0.06]; Table 2). Despite wide confidence intervals, moderate-sized reductions were also observed for sibling children’s externalizing behaviors (β = −0.36 [−0.98 to 0.27]) and prosocial behaviors (β = 0.34 [−0.21 to 0.90]). There were no significant differences in visual-receptive, fine motor, or receptive language development scale scores.
Sibling Child and Family Caregiving Outcomes Between Control and Intervention Groups
| Control | Intervention | Unadjusted Difference | Adjusted Difference | |||||
|---|---|---|---|---|---|---|---|---|
| Outcomes | Valid n | Mean (SD) or % | Valid n | Mean (SD) or % | Coefficient (95% CI) | P | Coefficient (95% CI) | P |
| Sibling child development | ||||||||
| Visual-receptive development | 104 | 17.28 (11.0) | 118 | 16.72 (10.07) | β = 0.08 (−0.18 to 0.33) | .56 | β = 0.10 (−0.20 to 0.39) | .52 |
| Fine motor development | 104 | 17.87 (9.50) | 118 | 17.15 (9.27) | β = 0.06 (−0.21 to 0.33) | .65 | β = 0.07 (−0.22 to 0.37) | .62 |
| Receptive language development | 104 | 17.77 (10.34) | 118 | 17.05 (9.61) | β = 0.13 (−0.11 to 0.37) | .27 | β = 0.09 (−0.18 to 0.36) | .51 |
| Expressive language development | 104 | 15.67 (8.80) | 118 | 15.64 (8.58) | β = 0.22 (−0.03 to 0.47) | .08 | β = 0.33 (0.03 to 0.62) | .03 |
| Prosocial behaviorsa | 32 | 4.38 (2.34) | 33 | 5.03 (2.73) | β = 0.31 (−0.20 to 0.82) | .23 | β = 0.34 (−0.21 to 0.90) | .23 |
| Internalizing problemsa | 32 | 7.72 (3.02) | 33 | 6.33 (2.35) | β = −0.58 (−1.04 to −0.12) | .01 | β = −0.56 (−1.07 to -0.06) | .03 |
| Externalizing problemsa | 32 | 10.69 (2.75) | 33 | 9.36 (2.66) | β = −0.54 (−1.03 to −0.06) | .03 | β = −0.36 (−0.98 to 0.27) | .26 |
| Sibling child nutrition | ||||||||
| Dietary diversity, number of food groups (0–8)b | 60 | 3.23 (1.23) | 64 | 3.83 (1.67) | β = 0.49 (0.11 to 0.87) | .01 | β = 0.52 (0.10 to 0.93) | .02 |
| Minimum dietary diversityb | 69 | 36.2% | 77 | 48.1% | OR = 2.16 (1.02 to 4.57) | .04 | OR = 3.65 (1.21 to 11.05) | .02 |
| Family caregiving with sibling child | ||||||||
| Maternal stimulation | 104 | 2.18 (1.80) | 117 | 2.62 (2.13) | β = 0.30 (0.03 to 0.56) | .03 | β = 0.31 (0.00 to 0.61) | .04 |
| Paternal stimulation | 104 | 1.10 (1.57) | 117 | 1.55 (1.83) | β = 0.30 (0.01 to 0.58) | .04 | β = 0.33 (0.02 to 0.65) | .04 |
| Maternal discipline | 104 | 68.6% | 117 | 61.7% | OR = 0.50 (0.22 to 1.14) | .10 | OR = 0.33 (0.08 to 1.42) | .14 |
| Paternal discipline | 104 | 58.7% | 117 | 50.8% | OR = 0.61 (0.29 to 1.28) | .19 | OR = 0.77 (0.27 to 2.21) | .62 |
| Observed caregiver-child interactions (total score)c | 80 | 40.66 (8.83) | 87 | 42.72 (8.27) | β = 0.27 (−0.06 to 0.60) | .11 | β = 0.34 (0.05 to 0.64) | .02 |
| Observed caregiver-child interactions (caregiver sub-score)c | 80 | 24.23 (4.12) | 87 | 25.82 (4.13) | β = 0.40 (0.07 to 0.72) | .02 | β = 0.40 (0.09 to 0.72) | .01 |
| Observed caregiver-child interactions (child sub-score)c | 80 | 16.43 (5.48) | 87 | 16.90 (5.24) | β = 0.12 (−0.21 to 0.45) | .48 | β = 0.23 (−0.08 to 0.54) | .14 |
| Sibling relationship quality (positive behaviors) | 101 | 2.67 (0.64) | 115 | 2.65 (0.74) | β = 0.05 (−0.28 to 0.38) | .77 | β = 0.09 (−0.26 to 0.44) | .62 |
| Sibling relationship quality (negative behaviors) | 94 | 3.17 (0.58) | 105 | 3.22 (0.61) | β = 0.08 (−0.21 to 0.38) | .58 | β = 0.04 (−0.26 to 0.35) | .77 |
| Control | Intervention | Unadjusted Difference | Adjusted Difference | |||||
|---|---|---|---|---|---|---|---|---|
| Outcomes | Valid n | Mean (SD) or % | Valid n | Mean (SD) or % | Coefficient (95% CI) | P | Coefficient (95% CI) | P |
| Sibling child development | ||||||||
| Visual-receptive development | 104 | 17.28 (11.0) | 118 | 16.72 (10.07) | β = 0.08 (−0.18 to 0.33) | .56 | β = 0.10 (−0.20 to 0.39) | .52 |
| Fine motor development | 104 | 17.87 (9.50) | 118 | 17.15 (9.27) | β = 0.06 (−0.21 to 0.33) | .65 | β = 0.07 (−0.22 to 0.37) | .62 |
| Receptive language development | 104 | 17.77 (10.34) | 118 | 17.05 (9.61) | β = 0.13 (−0.11 to 0.37) | .27 | β = 0.09 (−0.18 to 0.36) | .51 |
| Expressive language development | 104 | 15.67 (8.80) | 118 | 15.64 (8.58) | β = 0.22 (−0.03 to 0.47) | .08 | β = 0.33 (0.03 to 0.62) | .03 |
| Prosocial behaviorsa | 32 | 4.38 (2.34) | 33 | 5.03 (2.73) | β = 0.31 (−0.20 to 0.82) | .23 | β = 0.34 (−0.21 to 0.90) | .23 |
| Internalizing problemsa | 32 | 7.72 (3.02) | 33 | 6.33 (2.35) | β = −0.58 (−1.04 to −0.12) | .01 | β = −0.56 (−1.07 to -0.06) | .03 |
| Externalizing problemsa | 32 | 10.69 (2.75) | 33 | 9.36 (2.66) | β = −0.54 (−1.03 to −0.06) | .03 | β = −0.36 (−0.98 to 0.27) | .26 |
| Sibling child nutrition | ||||||||
| Dietary diversity, number of food groups (0–8)b | 60 | 3.23 (1.23) | 64 | 3.83 (1.67) | β = 0.49 (0.11 to 0.87) | .01 | β = 0.52 (0.10 to 0.93) | .02 |
| Minimum dietary diversityb | 69 | 36.2% | 77 | 48.1% | OR = 2.16 (1.02 to 4.57) | .04 | OR = 3.65 (1.21 to 11.05) | .02 |
| Family caregiving with sibling child | ||||||||
| Maternal stimulation | 104 | 2.18 (1.80) | 117 | 2.62 (2.13) | β = 0.30 (0.03 to 0.56) | .03 | β = 0.31 (0.00 to 0.61) | .04 |
| Paternal stimulation | 104 | 1.10 (1.57) | 117 | 1.55 (1.83) | β = 0.30 (0.01 to 0.58) | .04 | β = 0.33 (0.02 to 0.65) | .04 |
| Maternal discipline | 104 | 68.6% | 117 | 61.7% | OR = 0.50 (0.22 to 1.14) | .10 | OR = 0.33 (0.08 to 1.42) | .14 |
| Paternal discipline | 104 | 58.7% | 117 | 50.8% | OR = 0.61 (0.29 to 1.28) | .19 | OR = 0.77 (0.27 to 2.21) | .62 |
| Observed caregiver-child interactions (total score)c | 80 | 40.66 (8.83) | 87 | 42.72 (8.27) | β = 0.27 (−0.06 to 0.60) | .11 | β = 0.34 (0.05 to 0.64) | .02 |
| Observed caregiver-child interactions (caregiver sub-score)c | 80 | 24.23 (4.12) | 87 | 25.82 (4.13) | β = 0.40 (0.07 to 0.72) | .02 | β = 0.40 (0.09 to 0.72) | .01 |
| Observed caregiver-child interactions (child sub-score)c | 80 | 16.43 (5.48) | 87 | 16.90 (5.24) | β = 0.12 (−0.21 to 0.45) | .48 | β = 0.23 (−0.08 to 0.54) | .14 |
| Sibling relationship quality (positive behaviors) | 101 | 2.67 (0.64) | 115 | 2.65 (0.74) | β = 0.05 (−0.28 to 0.38) | .77 | β = 0.09 (−0.26 to 0.44) | .62 |
| Sibling relationship quality (negative behaviors) | 94 | 3.17 (0.58) | 105 | 3.22 (0.61) | β = 0.08 (−0.21 to 0.38) | .58 | β = 0.04 (−0.26 to 0.35) | .77 |
SD, standard deviation; CI, confidence interval; OR, odds ratio.
Unadjusted analyses controlled for child age.
Adjusted analyses controlled for child age, child sex, age difference between index and sibling child, whether sibling is younger/older than index child, singleton birth, parental age, parental education, marital status, household size, wealth quintile, and household food security.
Measured among children >2 years of age.
Measured among children 6–23 months of age.
Measured among children >6 months of age; 91% of caregivers were mothers.
Child Nutrition
Intervention sibling children consumed significantly more food groups in the past 24 hours (β = 0.52 [0.10 to 0.93]), particularly more legumes, vitamin A-rich foods, and eggs (Supplemental Table 3). Intervention sibling children were also more likely to achieve minimum dietary diversity than those in the control (48.1% vs 36.2% respectively; odds ratio = 3.65 [1.21 to 11.05]).
Family Caregiving
Moderate-sized intervention effects were detected on caregiver reports of maternal (β = 0.31 [0.01 to 0.61]) and paternal stimulation with the sibling child (β = 0.33 [0.02 to 0.65]) and observed scores of caregiver-child interactions (β = 0.34 [0.05 to 0.64]), especially for the subscale of caregivers’ responsive behaviors (β = 0.40 [0.09 to 0.72]). Although there was a wide confidential interval around this point estimate, intervention mothers had 67% lower odds in the use of harsh discipline (odds ratio = 0.33 [0.08 to 1.42]). There were no significant differences in paternal harsh disciplinary practices, sibling children’s observed behaviors with their caregivers, or sibling-child relationship quality. Overall results were robust to a sensitivity analysis that applied inverse probability weighting to correct for bias due to missing data in the EFFECTS-Sibs sample (Supplemental Table 4).
Discussion
We evaluated whether a bunded parenting and nutrition intervention that was delivered to both mothers and fathers (ie, couples) and targeted an index child <2 years of age had any intrahousehold spillover effects on nontargeted sibling children <6 years of age. By nesting a sub-study within a larger cluster-RCT, we found that sibling children in intervention households had improved expressive language development and dietary diversity and reduced internalizing behavioral problems. Moreover, parents in intervention households engaged in more optimal parenting practices with the sibling child.
Our results contribute to a nascent evidence base about sibling spillover effects in the context of early childhood interventions. Although there have been a few noteworthy evaluations of sibling spillover effects in child nutrition interventions, the literature to date has been mixed. For example, a cluster-RCT of an integrated agriculture production and nutrition behavior change program in Burkina Faso revealed reductions in child wasting and anemia for younger siblings <1 year of age.30 Another study of 2 cluster-RCTs in Ghana and Bangladesh did not reveal any differences in growth or undernutrition among nontargeted younger siblings <3 years of age from a maternal supplementation and nutrition counseling intervention during pregnancy and the postpartum period.18 Heterogeneity in types of nutrition interventions and ages of sibling children may preclude direct comparisons between these previous studies and results from a purely social and behavior change nutrition intervention that bundled parenting without any distributed commodities or treatments.
To our knowledge, there is only 1 prior study published to date about parenting intervention spillover effects on sibiing children’s developmental outcomes in an LMIC. In Uganda, Boivin et al evaluated a parenting intervention that was intended for HIV-infected mothers with preschool-aged index children and explored spillover effects on the developmental outcomes of older school-aged siblings children (between ages 5 and 12 years).31 They found a positive effect on sibling children’s working memory, a negative effect on behavioral development, and null effects on the remaining child outcomes (eg, cognition, attention, behavior, growth). Our study unites and extends these separate bodies of literature from nutrition and parenting interventions by evaluating a multicomponent bundled parenting and nutrition intervention and uncovering some positive sibling spillover effects on both ECD and nutrition outcomes.
The positive impact identified on sibling children’s diet is consistent and to a similar effect size as the benefits observed in the main EFFECTS trial on index children’s dietary diversity (results under review). In contrast, different program effects were observed on ECD outcomes among siblings (ie, improved expressive language and internalizing behaviors) versus index children (ie, improved cognitive and receptive language development).20 These varying results raise questions such as whether parents were able to scaffold the parenting lessons learned with the index and apply them in an age-appropriate manner with their other children in different developmental stages, which is especially necessary to impact children’s higher-order skills like cognition.32 It is also worth recognizing that different outcome domains and measurement tools were used to assess sibling versus index child development outcomes because of the varying age ranges of the respective samples, which could also limit the direct comparability between EFFECTS-Sibs and EFFECTS results.
Our study is one of the first known to explore intrahousehold spillover effects and measure both mothers’ and fathers’ parenting practices with nontargeted siblings and relationships between sibling pairs. The authors of previous studies in this literature on sibling spillover effects across LMICs and high-income countries have evaluated outcomes at the sibling-child level.17,18,30,31 Our broader consideration of family caregiving outcomes offers additional insights into potential mechanisms that may explain some of the observed sibling spillover effects. More specifically, we found evidence of improved parenting practices with not only the index child20 but also their siblings. Mothers’ and fathers’ increased engagement in play and communication activities with multiple children beyond the index child alone supports social learning and family systems theory. The practical program components of EFFECTS, for example, providing parents with opportunities to engage responsively in play, communication, and feeding of their index child and having community health workers provide feedback and coaching to enrich the quality of parent-child interactions, likely facilitated internalized reflections and experiential learning that caregivers could more broadly apply with their other children.
In addition to the pathway via improved caregiving practices then extended with other children, we also speculated the possibility of intervention spillover effects on sibling outcomes via improvements in sibling relationship quality (ie, a more direct sibling influence). This exploratory hypothesis was motivated by nascent literature indicating that changes in 1 siblings’ temperament can impact relationship dynamics between siblings and in turn promote developmental outcomes of siblings.33,34 However, we found limited support for this particular pathway in our study. More research in the measurement of sibling-child relationships among young children, including observational measures of sibling interaction quality35,36 and especially for use in low-resource global contexts, can expand our understanding of whether interventions have potential spillover effects on sibling-child relationships.37 Finally, other possible pathways relating to the family caregiving environment more broadly could have also potentially contributed to sibling spillover effects within the EFFECTS trial, such as via improvements in fathers’ parenting stress, couples’ relationship dynamics, or gender attitudes.20
Our study had several limitations. EFFECTS-Sibs was exploratory and conceptualized after the design of EFFECTS. Thus, our sample size was fixed to households already enrolled in the parent study. Certain outcomes were measured in subsamples of children because of specifications about child age and were further underpowered. This impacted statistical power for not only the primary analyses but also precluded exploratory subgroup analyses (eg, whether spillover effects vary for younger vs older-aged siblings, siblings closer vs further apart in age) and mediation analyses to test mechanisms underlying program spillover effects.38 We also conducted many statistical tests without adjusting for multiple comparisons because of critiques about the inappropriateness of such approaches.39,40 Thus, our results may reflect some chance findings. Although both mothers and fathers (ie, couples) were recruited and participated in the intervention, for the EFFECTS-Sibs evaluation, only 1 primary caregiver was sampled per household (ie, primarily mothers), who self-reported on most outcomes, including both maternal and paternal parenting practices. Consequently, results may be subject to bias (eg, common reporter bias, social desirability bias). Future evaluations seeking to disentangle program impacts on maternal versus paternal parenting should sample both mothers and fathers as part of data collection efforts to obtain more balanced and holistic perspectives. Finally, although we systematically translated and pretested measures that have been used previously in Tanzania, there are limited studies about the validity of these tools and specifically for use in the local context, which raises the possibility of other sources of measurement bias.
Conclusions
We found that a father-inclusive, bundled parenting and nutrition intervention improved dietary diversity, expressive language development, and internalizing behaviors among sibling children <6 years of age in intervention households. We also found intervention improvements in caregivers’ parenting practices specifically with the sibling child, suggesting that sibling spillover effects may be partly due to caregivers’ extension of positive caregiving behaviors with not only the index child but also other children. Taken together, our findings highlight broader family-wide impacts of caregiving interventions on nontargeted siblings of enrolled children. Explicitly encouraging and expanding guidance about how caregivers can support multiple children (eg, developmentally at different stages/ages, positive sibling relationships) in future intervention curricula is likely to have greater potential for sibling spillover effects. Key areas of future research include developing a theory of change for caregiving intervention spillover effects on sibling children’s outcomes, identifying the specific program messages, components, and pathways that facilitate such intrahousehold spillover effects, and measuring a comprehensive set of sibling child outcomes using validated tools for the local context and across the age range of assessed children. In particular, future studies should investigate how additionally engaging fathers may contribute to spillover effects. Finally, spillover effects of caregiving interventions should be explored not only within and between families but also in terms of community-wide changes.
Acknowledgments
We thank the study participants and field research assistants for their participation and contributions to this work. This study is based in part on the EFFECTS study, which was a larger implementation research study funded by grants from the Conrad N. Hilton Foundation, Eleanor Crook Foundation, 1000 Days, and Risk Pool Fund to Global Communities, Purdue University, the Africa Academy of Public Health, and the Harvard T.H. Chan School of Public Health. The findings and conclusions in this study are those of the authors and do not necessarily reflect the views of the funders or partners.
Dr Jeong conceptualized and designed the study, conducted the analyses, and drafted the manuscript; Ms Domonko and Ms Mendile contributed to acquisition of data and oversaw field research activities; Dr Yousafzai contributed to interpretation of findings; and all authors critically reviewed and approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
This trial has been registered at www.clinicaltrials.gov (identifier NCT03759821).
Deidentified participant data will be made available after publication to researchers who provide a methodologically sound proposal for use and on reasonable request. Proposals should be submitted to the corresponding author.
FUNDING: Research reported in the present article was supported by the Eric M. Mindich Research Fund for the Foundations of Human Behavior. Dr Jeong is supported in part by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (K99HD105984).
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no potential conflicts of interest relevant to this article to disclose.
Comments