Background
Socioemotional, cognitive and physical development indicators by the age of five in low- and -middle-income countries (LMIC) have been linked to poor academic performance and poverty in adulthood, which can perpetuate poverty throughout generations [
1]. Responsive caregiving has emerged in multiple settings as a key parenting domain that is linked to improved physical, cognitive and psychosocial health in children in both high and lower income countries [
2,
3]. Responsive caregiving can be defined as interactions in which a caregiver provides the child with proper feedback to their behaviors and signals [
4,
5]. Examples of responsive caregiving include behaviors that positively encourage focus on a task with the child, as well as positive affect and positive verbal statements. The quality of these interactions is linked to child-development knowledge [
4,
6], as well as the emotional availability of the caregiver [
4,
7]. Responsive caregiving is thought to be essential for forming a secure attachment relationship and has been linked to improved cognitive, health, psychosocial, disease and mortality outcomes in children [
3,
8]. The positive effects of responsive caregiving behaviors have been found to extend long into childhood in high income settings. For example, more frequent displays of responsive caregiving in early childhood have been linked to lower amounts of behavioral issues at three years of age [
9], higher intelligence at four and twelve years of age [
10,
11], and positive academic outcomes at seven years of age [
3,
12]. There is also research in high income countries that suggests responsive caregiving is linked to improved child health outcomes such as decreased hospitalizations [
13]. Several findings have also identified low levels of responsive caregiving as a predictive factor of worse developmental outcomes. For example, one study found that a lack of responsive caregiving behaviors in the early months of life was linked to worse socioemotional outcomes at age three [
14]. Responsive caregiving might be especially important in the context of maternal depression in that low-quality responsive caregiving may be one of the pathways through which maternal depression is predictive of non-optimal child outcomes [
15]. Results from a meta-analysis of 46 studies in high-income countries indicated that maternal depression was associated with negative parent behavior (d = 0.40) and disengagement (d = 0.29) with the child [
16].
Within LMIC, there is a dearth of research designed to examine responsive caregiving in detail. A key reason is that the measurement of responsive caregiving has traditionally been based on the coding of videotaped interactions between the mothers and their infants in lab environments [
14,
17]. Using these procedures standardizes coding and environmental variance, but using them is a time and resource intensive endeavor that poses significant challenges to accomplish in LMIC. In addition to challenges associated with resources, lab-based recording approaches may not be culturally appropriate or possible due to human resource shortages and a lack of funding to train individuals [
4]. A handful of investigators have successfully observed mothers in a naturalistic setting, and these results linked specific responsive caregiving behaviors to positive outcomes such as improved parent-child communication and better child vocabulary in relation to mother’s verbal responsive caregiving [
18,
19].
The majority of research related to responsive caregiving in LMIC relies on the Home Observation for Measurement of the Environment (HOME) [
12]. This tool, designed to assess home environment and stimulation quality, has been used frequently in LMIC [
20‐
22]. Two subscales of the HOME (Responsivity and Involvement) are often used to assess responsive caregiving; however, these subscales are rarely reported separately from the total HOME score, which makes it difficult to parse out the effects of responsive caregiving from the effects of the physical home environment [
4,
20].
The Observation of Mother-Child Interactions (OMCI) measure was recently created and used in rural Pakistan to observe and code structured responsive caregiving interactions without the use of video-recording [
4]. The OMCI tool was developed utilizing a theoretical framework by Landry, Smith, and Swank [
23] to measure the frequency of four domains of maternal responsive behaviors (contingent responding, emotional-affective support, support for infant foci of attention, and language inputs) through a structured interaction activity. In the present study, responsive caregiving was characterized using a 5-min observation of mothers and children interacting with a picture book. At the beginning of the interaction, the mother was told that the observer was interested in watching her play and talk with the child using the picture book. During the interaction, the observer simultaneously coded 11 maternal and 4 child behaviors. The interaction was intended to last five minutes. If the mother stopped early (< 4 min had passed), she was asked to continue; however, if the interaction had lasted longer than 4 min, the assessor allowed the interaction to be ended prematurely. Child behaviors were also observed to assess their responsivity to the caregiver’s behaviors [
4]. Tool development occurred in a series of five steps: construction of items, field testing, expert review, data collector training, and pilot use by child development assessors based in the community [
4]. During tool development, OMCI scores were found to be normally distributed, to have a high inter-observer reliability (
r = 0.85), and to have predictive validity for child outcomes including language development (
r = 0.62) and motor development (
r = 0.57) [
4]. Additionally, the OMCI scores were found to be significantly correlated with the Responsivity (
r = 0.27) and Involvement (
r = 0.33) subscales of the HOME and associated with child growth [
4]. The ability to assess structured interactions without the use of video may provide additional information on responsive caregiving behaviors, since direct observation has been shown to provide less biased information on responsive caregiving behaviors than self-report measures [
24]. Findings from a recent study indicated that responsive caregiving, as assessed by the OMCI, mediated the effects of responsive stimulation interventions on children’s cognitive outcomes [
25]. To date, few other published studies have used the OMCI to assess its relationship with child outcomes.
This study was designed towards three goals. The first goal was to describe how maternal depression and sociodemographic variables were associated with responsive caregiving as measured by the OMCI. The second goal was to assess how responsive caregiving was linked with socioemotional and growth development indicators in the second year of life. Thirdly, since there is prior evidence that the OMCI and HOME are correlated [
4], this study compared results using the OMCI and the HOME Responsivity and Involvement subscales to see if there is any new information to be gained through the use of the observational strategy.
Discussion
This study was designed to investigate the relationship between responsive caregiving and child outcomes using the recently developed Observation of Mother-Child Interaction (OMCI) tool. The results of this study suggest that maternal depression, maternal education, household assets, and number of children are significantly related to maternal responsive caregiving scores, pointing to the importance of sociodemographic factors in determining responsive caregiving in LMIC. Additionally, the results of this study highlight responsive caregiving as a significant factor associated with positive socioemotional outcomes for children in rural Pakistan. This finding remained statistically significant after controlling for sociodemographic factors and maternal depression, indicating that improved responsive caregiving is associated with improved child outcomes even in low-resource environments. Finally, we found that the OMCI uniquely predicts socioemotional outcomes and is distinct from the responsive caregiving HOME subscales.
Our findings regarding the sociodemographic links with responsive caregiving corroborate the current body of literature. A recent study that examined data from 44 LMIC countries found that maternal education was linked to improved child outcomes, and that a key mechanism in this association was responsive caregiving behaviors, such as positive stimulation and other behaviors that supported learning [
34]. Associations have shown that improved child outcomes are linked to higher assets partially through the mediating factors of responsive caregiving behaviors [
35]. Our finding between maternal depression and responsive caregiving aligns with previous bodies of literature that have connected maternal depression to negative behaviors towards the child [
16] and have also linked participation in depression interventions to better physical and cognitive health outcomes for children later in life [
36].
This study adds to a growing body of research that links responsive caregiving to positive child outcomes and is among the first to use a measure of direct observation in LMIC [
25,
37,
38]. Tools that complement parent reporting with live observational strategies may help to capture a more complete picture of responsive caregiving behaviors and of children’s response to those behaviors. We found that responsive caregiving was associated with better child outcomes above and beyond sociodemographic factors, which has been found previously [
38]. Furthermore, interventions that focus on responsive stimulation have been shown to improve child outcomes [
37,
39‐
41] indicating that responsive caregiving is a modifiable domain of focus. Therefore, the OMCI may be an ideal tool to guide interventions and studies geared towards optimizing child development through responsive caregiving.
The study found no meaningful association between OMCI and 24-month child height-for-age z-score (HAZ). The current literature provides a mixed picture on this relationship. While some studies have identified a relationship between responsive caregiving and HAZ [
4,
42,
43], others have found no effects of responsive stimulation on child growth [
39,
40,
44,
45]. In a 2 × 2 factorial randomized controlled trial of responsive stimulation and enhanced nutrition in Pakistan, improvements in child HAZ was only found in the enhanced nutrition arms [
44] in which supplements were provided in addition to nutrition education. Other studies in Bangladesh and Uganda have also found no effects of responsive caregiving interventions with emphasis on responsive stimulation and caregiving on linear growth [
39,
40,
45]. In our sample, the average HAZ at 24 months was − 1.23 and the prevalence of stunting was around 25%, highlighting significant levels of overall disadvantage for children. In our population, it may be the case that environmental factors, such as latrine use and handwashing, and structural factors, such as food security and wealth, are more important for child HAZ than responsive caregiving alone. In addition, responsive caregiving as captured through mother-child interactions with a picture book in the OMCI may not be an adequate measure of the responsive caregiving behaviors that are pertinent for child nutritional status, such as maternal behavior during feeding.
The strength of the association between the OMCI and the ASQ-SE, as well as the links to sociodemographic variables previously connected to responsive caregiving, demonstrate that the OMCI is predictive of relationships supported by previous research and is worthy of further use in LMIC in assessing responsive caregiving. Based on the comparison with the HOME Inventory, which is often used to assess responsive caregiving in LMIC, the OMCI is capturing unique information. Coded and structured interactions, assessed by the OMCI rather than the unstructured HOME, may be useful in determining specific domains that are linked to development. This information could better serve research and interventions that investigate how specific, and potentially teachable, responsive caregiving behaviors are related to development in ways that have yet to be possible without the addition of this measurement strategy. Further research should compare the OMCI and the HOME, as this study is limited in that the measures were taken at different time points and there could be significant temporal variations between the effects of responsive caregiving at 12 and 24 months. It is likely that using the OMCI and the HOME in tandem could yield a more robust picture of responsive-caregiving and the environment in which it occurs.
Strengths and limitations
This study has several strengths. First, we used a unique, low-cost tool to assess responsive caregiving that does not rely on maternal self-report or require extensive resources (e.g., video recording equipment, lab space, etc.). Moreover, assessing responsive caregiving in its ‘natural’ environment may be more valid than lab-based settings. Second, we leveraged population-representative longitudinal data to explore the impact of responsive caregiving on child outcomes in an understudied, vulnerable population. Several of the strengths provided by the OMCI also create limitations for this study. Though the OMCI’s direct observation feature is helpful in LMIC where video recording is logistically infeasible or socially inappropriate, it is also limiting in that it is subject to greater rates of observer bias. Though the tool has been shown to demonstrate good inter-observer reliability between expert and trained assessors [
4], it is still a limitation that the responsive caregiving interactions are unable to be assessed multiple times in video analysis to ensure consistent coding across multiple assessors. Additionally, the observed responsive caregiving interactions may have been different than their regular responsive caregiving interactions, due to observer effects (though this is also a limitation in other video-based assessments). However, the OMCI tool was designed to reduce observer bias as much as possible [
4]. For example, mothers were familiar with the person observing them and the observation occurred in their home. Additionally, we were limited in that nutritional factors such as exclusive breastfeeding status, complementary feeding practices, and dietary diversity, were not available for further analysis. An important aspect of responsive caregiving involves the child feeding behaviors and practices of mothers. Future tools to characterize responsive caregiving should take into account both structured play observation as well as mother-child interactions during feeding. Another limitation is that the responsive caregiving interaction was carried out using a picture book, which may have been a preference for more educated mothers. While the book had no words, it still may have been biased towards mothers who were more familiar with books. Using some other culturally-appropriate toy during the responsive caregiving interaction may be more successful in eliminating bias towards educated mothers. Since the OMCI is a relatively recent tool, it has been used in limited studies since its development in 2013 and warrants further investigations, particularly given the promise it demonstrated with the present analyses.
Acknowledgements
The authors would like to thank the team at the Human Development Research Foundation (HDRF) including Rakshanda Liaqat, Tayyiba Abbasi, Maria Sharif, Samina Bilal, Quratul-Ain, Anum Nisar, Amina Bibi, Shaffaq Zufiqar, Sonia Khan, Ahmed Zaidi, Ikhlaq Ahmad, and Najia Atif for their meaningful contributions to the study’s design and implementation. We also gratefully acknowledge the larger Bachpan and SHARE CHILD study teams. Lastly, we are deeply grateful to the women, children, and communities that are a part of the Bachpan cohort.