Introduction

Young children living in high-adversity settings experience high levels of poverty, undernourishment and stunting, inadequate access to early childhood education, and risks in their home environments including parental depression, domestic violence, and drug abuse (Black et al., 2017). Interventions to improve early childhood development (ECD) have proliferated in the last two decades. Typical interventions comprise a diverse set of practices that aim to help young children reach their full developmental potential by enhancing responsive care (Walker et al., 2011). Because environmental stimulation affects human brain development most critically in the early years (Grantham-McGregor et al., 2007), interventions implemented during this period can translate to short-term benefits—such as school readiness and reductions in harsh parenting—as well as longer-term benefits, including an increased ability to learn, better school achievement, emotional wellbeing, and cognitive outcomes into adolescence and adulthood (Kagitcibasi et al., 2009; Walker et al., 2018).

Interventions are often targeted at parents and/or other caregivers to promote developmentally appropriate nurturing care by providing learning opportunities for the young child and helping parents and/or caregivers to engage the child in stimulating play activities (Eshel et al., 2006; Jeong et al., 2018). However, there are gaps in how we understand these interventions to work. Notably, few interventions have been longitudinally followed, and there is evidence of “fade-out” where effects are not last-lasting. Existing evidence about intervention effectiveness in the short term is also mixed, as content and delivery strategies, as well as evaluative rigor, can vary substantially (Jeong et al., 2021). These interventions also contain distinct elements that are theorized to drive the same outcomes. Psychosocial interventions may include elements of positive parent–child interactions, including providing positive attention and responsiveness to cues and milestones, encouraging children’s autonomy and exploration of the environment, and promoting attachment (Yousafzai & Aboud, 2014). In practice, however, these interventions are designed and implemented in different ways, with lack of clarity about which components may work best.

Furthermore, much of the highest-quality evidence on early childhood interventions still comes from high-income countries (HICs) (Lipina & Colombo, 2009). While evidence-based practices for promoting parent–child interactions and encouraging cognitive development can ostensibly be widely applied, implementing these well-evidenced interventions in low- and middle-income countries (LMICs) often requires contextual, sociocultural, and resource adaptations (Britto et al., 2018). The group of countries often referred to as LMICs are highly diverse, reflecting a wide range of social, cultural, economic, and political realities. Consequently, the generalizability of findings across LMIC settings is often limited, although the need for low-cost cognitive development interventions that are feasible to implement in low-resource contexts is great (Richter et al., 2017).

As the evidence base grows, there is an increasing need to systematically study and identify the necessary and sufficient components for effective early interventions (Cavallera et al., 2019). Current practice is moving away from a pure effectiveness approach—where interventions are delivered and tested as manualized and sometimes ‘black box’ packages—towards a more flexible approach where key intervention components, sometimes referred to as common elements, are closely documented or matched to circumstances (of the child, context or level of facilitator expertise) to improve outcomes and scalability (Chorpita et al., 2007; McLeod et al., 2017). Isolating “active” ingredients of interventions could help reduce overlap across a saturated programming field and build consensus about what works. To date, this task has proven difficult due to inconsistency and poor quality in how interventions are described and reported in peer-reviewed literature (Yousafzai et al., 2018). In this review, we aimed to identify effective ECD interventions through conducting a systematic review of reviews and analyzing intervention descriptions to identify common elements that could support future practice.

Methods

To identify effective intervention programs, we reviewed existing systematic reviews and meta-analyses investigating the effectiveness of interventions in LMICs to improve ECD.

We searched three electronic databases (PsycINFO, Web of Science, and PubMed) and three subject-specific databases (EPPI Centre, WHO Global Health Library, and UNICEF Publications Database) covering the period until the end of March 2021. Reference lists of relevant reviews, including the recently published WHO ECD Guidelines (WHO, 2020), were consulted to identify any additional reviews not captured by the database searches.

Inclusion Criteria

We included systematic reviews that met the following criteria:

  1. 1.

    Population: Parents and/or primary caregivers of children under 5 years of age;

  2. 2.

    Intervention: Studies evaluating early childhood interventions promoting cognitive development, either as a stand-alone intervention or combined with other intervention strategies, such as nutritional programs (excluded interventions solely focused on material and/or structural interventions, e.g., cash transfers);

  3. 3.

    Comparator: Standard care or active interventions;

  4. 4.

    Outcome: Studies measuring child cognitive outcomes and/or parenting outcomes;

  5. 5.

    Setting: Studies in LMICs only, as defined by the World Bank criteria.

Search and Screening Strategy

Our search terms (see Table 1)were inclusive, enabling us to find reviews with a wide range of cognitive outcomes. No language limits were applied to searches. Hand searches of reference lists and the three additional subject specific databases were also conducted. Following identification of reviews through databases and reference list searches, records were screened for relevance based on title and abstract by two reviewers independently.

Table 1  Search strategy and terms
Full size table

Quality Assessment

Two raters (CL and SG) assessed the methodological quality of the included reviews using the AMSTAR2 tool for assessing systematic reviews (Shea et al., 2017). Discrepancies in assessments were resolved through discussion.

Data Extraction

Following review selection, we identified individual studies that (1) demonstrated effectiveness and (2) were included in reviews that were assessed as “moderate” or “high” quality on AMSTAR2. Effectiveness was ascertained through standardized mean difference (SMD) values, supported by review authors’ conclusions. Core data extracted included country, aim, number of studies, number of total participants, outcomes, and findings including SMD. We then contacted authors of these studies to request the original intervention protocols. Where intervention protocols were not available, published papers of the studies were used to extract information about the intervention content, with coding based on descriptions provided in the articles.

Common Element Coding

The PracticeWise manual was used for coding common elements (Chorpita et al., 2005). The original manual was developed for high-income contexts, and we adapted this to suit our focus on the early years in the LMIC context. Elements could relate to intervention content, delivery, or other intervention practices.

Two coders used the PracticeWise manual to review 62 common elements in the available protocols and intervention studies (PracticeWise, 2012). Codes that were not relevant for the purpose of the current review were excluded (n = 19). Additional relevant codes were added to the coding structure through the procedure outlined in the manual, adding new elements as free text, and then reviewing these for frequently occurring elements. Through this process, 20 new elements were added (see Table 2). Codes not identified in two or more interventions were omitted from the final list (n = 17), leading to a final count of 46 common elements. Several original codes were re-defined to relate to a caregiver, such as individual therapy-caregiver. Other definitions of codes were broadened to include early years specific aspects; for instance, the code caregiver-directed nutrition was widened to include advice about breastfeeding, weaning, and psychoeducation about portion sizes for young children. For studies identified in the updated search, the early set of codes were applied to a new set of interventions to streamline efforts. In cases where multiple active interventions were tested in a trial (multi-arm study), we included common elements found in the most “involved” or intensive arm (Table 2).

Table 2  Elements and descriptions
Full size table

After additional codes were included, two raters (CL, SG) coded elements on a sample of 9 protocols to test for inter-rater reliability (kappa = 0.82).

Results

Description of Reviews

We identified 54 records, excluding n = 37 reviews based on title and abstract, and assessed an additional 17 reviews in detail. Ten reviews fit inclusion criteria for the current review (see Fig. 1 for PRISMA flow chart). The findings from these ten reviews are detailed in Table 3.

Fig. 1
figure 1

PRISMA flow chart

Full size image
Table 3 Summary of included reviews
Full size table

Five of these ten reviews used meta-analyses. However, only three of these meta-analyses focused specifically on cognitive outcomes in children; the remaining two systematic reviews included psychosocial interventions but only conducted meta-analyses on nutrition-specific interventions, a topic outside of the scope of our current review. Several reviews evaluated outcomes on child cognitive and language development (n = 7) and parent–child relationships (n = 7), while one measured maternal mental health outcomes with a secondary focus on child cognitive development (n = 1).

Interventions

Interventions included: interventions in early infancy focusing on promoting maternal-child interaction; interventions with disadvantaged children and their families; combined interventions for cognitive development and nutrition programs; interventions to increase parental responsiveness related to feeding, attachment, and general parenting skills; and interventions to encourage learning and play activities, positive discipline, and problem-solving. Some interventions reviewed were parent-focused, with a primary focus on improving maternal mental health with children’s developmental and cognitive outcomes measured as secondary outcomes.

Intervention delivery formats included home visits, group sessions, and clinic appointments; delivered by paraprofessionals visiting the family weekly or monthly, or delivering the sessions to a group of mothers, to talk to and play directly with the child while the parents watched and coaching activities between the parent and child. Play materials were often provided or hand-made in the sessions. Effectiveness of interventions was measured by the effect size of child outcome or the changes in parenting behavior contributing to a more stimulating child environment.

Quality Assessment of Systematic Reviews

The results of the quality assessment using AMSTAR2 tool are summarized in Table 4. Two raters achieved acceptable inter-rater agreement, at Cohen’s kappa = 0.65 (95% CI 0.107, 1.207). The confidence in the results of five reviews was judged to be high, with three reviews assessed to be of moderate quality, and two considered to be of low quality. Most reviews considered the heterogeneity of the included interventions and noted the limitations this created for drawing conclusions. A common weakness across reviews was a lack of assessment of publication bias, and an insufficient consideration and discussion of risk of bias. We adopted a more lenient approach to assessing two of these criteria: we measured whether authors considered risk of bias and heterogeneity, without stipulating that authors provide more nuanced assessments. No reviews provided a list of excluded studies and reasons for this as dictated by “gold standard” criteria, nor was the source of funding of individual interventions considered. Few reviews provided a clear indication that the study selection and data extraction was carried out in duplicate. For the current review, these were deemed to be non-critical weaknesses in assessing the strength of confidence in the findings of reviews.

Table 4 AMSTAR quality appraisals
Full size table

Overview of Review Findings

Across reviews, several similarities emerged. However, authors also employed a diverse set of strategies to classify outcomes, and extracted different information from included studies. Overall, there were generally large effect sizes identified across reviews in child cognitive and language development, as well as in parenting practices. Aboud and Yousafzai (2015) reviewed and meta-analyzed 21 interventions, and effects were found for cognitive outcomes (SMD = 0.42 [95%CI: 0.36, 0.48]) and language outcomes (SMD = 0.47 [95%CI: 0.37, 0.57]), with high heterogeneity (Aboud & Yousafzai, 2015). Jeong et al.’s meta-analysis (Jeong et al., 2018) on the effects of psychosocial interventions of parent caregiving practices in LMIC settings found medium-to-large effects across multiple domains; these domains included improved home caregiving environment (k = 10; SMD = 0.57; 95%CI: 0.37, 0.77), mother–child interactions (k = 3; SMD = 0.44; 95%CI: 0.14, 0.74), and maternal knowledge of child development (k = 6; SMD = 0.91; 95%CI: 0.51, 1.31). Jeong et al.’s more recent global systematic review and meta-analysis (Jeong et al., 2021) incorporated a broader set of outcomes and interventions than prior interventions, which the authors noted as a strength. The review’s sub-group analysis of studies in LMICs (k = 41 of the total 102 trials evaluated) identified significant positive improvements in child cognitive development outcomes (k = 32, SMD = 0.41, 95%CI: 0.29, 0.53, P < 0.001), language outcomes (k = 25, SMD = 0.35, 95%CI: 0.21, 0.48, P = 0.002), and on parenting practices (SMD = 0.47; 95%CI: 0.34, 0.61, P < 0.001). No significant effects were identified across parenting knowledge, parent–child interaction, or parental depression. The authors note that these large effect sizes are likely closely linked to the strategies commonly utilized across interventions that target early play and learning opportunities.

Other reviews looked to distinct groupings of interventions. Britto et al., investigating parenting programs (k = 105), found that stimulation interventions were found to yield better results than interventions with a sole focus on nutrition (Britto et al., 2015). Rao et al. (2017) meta-analyzed studies by intervention focus area, with child-focused education and stimulation interventions comprising the largest grouping (k = 37), followed by parent-focused education and support interventions (k = 22) and comprehensive interventions integrating aspects of both child and parent stimulation (k = 4) (Rao et al., 2017). Comprehensive programs had the largest positive effects on cognitive development (g = 1.05, 95%CI: 0.5,1.61), followed by child-focused stimulation interventions (g = 0.64, 95%CI: 0.42, 0.85), and parent-focused education and support interventions (g = 0.44, 95%CI: 0.26, 0.61).

There was also some emerging evidence of differential intervention impact by socioeconomic status. Among the outcomes for which Jeong et al. (2021) found significant improvements, intervention effects were greater among interventions delivered in LMICs than those in HICs (Jeong et al., 2021). Examples include parenting practices (SMD = 0.47, 95%CI: 0.34, 0.61, P < 0.001 in LMICs, versus SMD = 0.08, 95%CI: − 0.01, 0.16 in HIC) and child cognitive development (SMD = 0.41, 95%CI: 0.29, 0.53, P < 0.001 in LMIC versus SMD = 0.17, 95%CI: 0.10, 0.22 in HIC). As the authors note, however, no studies directly compared the same intervention between LMIC and HIC settings. Rao et al. found significantly different program effects from child-focused educational interventions, with an effect size nearly double in LMICs (0.64) compared to HICs (0.35) (Rao et al., 2017). Engle et al.’s review, too, identified larger effect sizes in studies conducted among socioeconomically disadvantaged populations (Engle et al., 2011).

There was some inconsistency across reviews about the moderation of effect by intervention characteristics. Britto et al.’s review of parenting programs (k = 105), including interventions for parenting practices, parent–child interactions, and parenting knowledge, beliefs, and attitudes reviewed studies for intensity, delivery model, approach, staffing, and program type (Britto et al., 2015). Studies with no intervention impact (k = 14) were typically delivered in a didactic manner and in lower dosages; these were often interventions where parents did not have direct interactions with children during the intervention. More frequent program contact was also associated with greater effectiveness, a finding echoed by Baker-Henningham and Boo’s comprehensive literature review (Baker-Henningham & Lopez Boo, 2010). Based on these findings, the authors recommended implementation periods of 12 months or longer to achieve optimal improvements.

Engle et al. found larger effect sizes for interventions including both parents and children and encouraging parent–child interventions (median d = 0.46, range 0.04–0.97) compared to those targeting parents only (median d = 0.12, 0.03–0.34) (Engle et al., 2011). Programs reported as most effective in this review were those with a structured curriculum, systematic training for intervention implementers, and opportunities for parents to actively practice new strategies with children and receive feedback. While the review by Rahman et al. (2013) had a slightly different focus—maternal interventions for common mental health problems—it examined secondary outcomes of child cognitive and physical development and mother–child interactions (Rahman et al., 2013). There were additional benefits identified from maternal participation in interventions, which improved knowledge about children’s needs and higher sensitivity, and provided improved home stimulation environments, as well as encouraging mothers to spend time playing with their children. These findings were echoed by reviews by Khatib et al. (2020) and Baker-Henningham (2014).

Jeong et al. also identified that parenting interventions that promoted responsive caregiving had significantly greater effects on parenting knowledge, practices, and parent–child interactions, as well as children’s cognitive development. However, when considering factors such as child age, the duration of the intervention, mode of delivery, or setting, they found no significant moderation effects on more than one outcome (Jeong et al., 2021).

Common Elements

We extracted common elements from intervention studies that (1) demonstrated intervention effects on children’s cognitive or parenting outcomes and (2) were part of systematic reviews identified as high-quality reviews by our AMSTAR2 process.

In total, 36 effective interventions were identified. Because several trials tested the same intervention protocol, we streamlined this number to n = 28 distinct intervention protocols for common elements extraction (see Table 5).

Common elements were derived from intervention protocols and/or manuals specific to the intervention, where we were able to obtain them from the from authors (n = 13), as well as intervention descriptions contained in outcomes papers and doctoral theses (n = 15). Figure 2 summarizes elements by frequency across interventions.

Table 5 Overview of included interventions
Full size table
Fig. 2
figure 2

Frequencies of common practice elements, by umbrella theme

Full size image

We grouped common elements under six broad “umbrella” themes: facilitator-focused elements, session-based content, parent-focused elements, enhancing parent–child interactions, psychoeducation, and techniques. Facilitator-focused elements included skills that facilitators used within sessions to encourage parents. Session-based elements included materials or practices that supported the facilitators within intervention setting sessions. Parent-focused elements included components that supported parents as individuals, whether through enhancing their skills or supporting their wellbeing. Enhancing parent–child interactions, the largest category, included skills and practices that aimed to shape and improve interactions between parent and child. Psychoeducation included any education-related provision within the session. Finally, techniques referred to other cross-cutting parenting practices covered in sessions that did not fall neatly into other categories.

While caregiver psychoeducation was the most common element (100%), strategies to enhance parent–child interaction were similarly well-represented—including play/pretend (92.9%), talking to the baby (85.7%), responsive care/parenting (85.7%), use of toys (78.6%), and attachment building (75%) among the most common.

Psychoeducation included providing parents with information about stages of infant development (World Health Organization, 2012). Parents are advised about how to best engage the child given their level of ability. Similarly, interventions provided information on the learning benefits of children’s self-feeding, which included motor coordination skills; cognitive development through learning different shapes, textures, and tastes; and gaining a sense of mastery of doing things by themselves (Aboud & Akhter, 2011).

The play/pretend element was often described to parents as children’s “work,” through which they learn about themselves, others, and the world. Protocols provided recommendations for play activities most commonly for children under 2 years of age, and parents were often encouraged to continue these activities in their own time. Often this element was coded alongside caregiver psychoeducation, as caregivers learned about play activities as a means of promoting children’s healthy development through seeing, hearing, touch, movement, and taste (Yousafzai et al., 2014). Play/pretend was also often coded alongside use of toys, as many interventions (n = 23) made use of simple, often hand-made toys, sometimes crafted as part of the intervention activities from household or other common materials.

The responsive parenting element was observed across protocols in diverse ways (Cooper et al., 2009; Singla et al., 2015), where activities encouraged parents to engage in a two-way talk with their child about a picture and respond to the child’s interests and sounds, as well as follow their lead to create a mediated learning experience. This element also included responsive physical care, such as responsive feeding.

The talking to baby element, encouraging communication and narration from caregivers to infants, was described across protocols as aimed at enhancing learning and bonding (Chang et al., 2015). This element was also coded in cases where caregivers were encouraged to sing, babble, or make eye contact with babies, including verbal and non-verbal communication between the parent and the child. This element also linked with attachment building, including building trust, showing love, and providing safety and security.

Within broader categories, other elements emerged. Family enhancement was common among techniques used, as were strategies to support problem-solving. Session aids, live or video demonstrations, and peer-based delivery methods were common. Nearly two-thirds of interventions included strategies to increase motivational enhancement and just over half of interventions included facilitator-specific elements such as relationship/rapport building, modelling, coaching, and accessibility promotion, to support buy-in and uptake of core intervention content. Furthermore, nearly two-thirds of interventions involved facilitator praise for the caregiver, as well as more general parenting skills, with other parent-focused provisions more intermittently included.

Discussion

This review of reviews and accompanying common elements analysis evaluated the evidence and distilled elements underpinning ECD interventions in LMIC contexts to provide a clearer picture of what works to improve child cognitive and parenting outcomes.

Findings in Context

The ten included systematic reviews found evidence of effectiveness across psychosocial interventions implemented in LMICs. Comprehensive interventions of higher intensity and longer duration yielded better results (Baker-Henningham, 2014; Britto et al., 2015; Engle et al., 2011), whereas less effective interventions were typically delivered didactically, without parent–child interaction, and in lower dosages (Britto et al., 2015). Interventions addressing maternal mental health during the perinatal period also emerged as beneficial for children’s cognitive development (Rahman et al., 2013). While three reviews found stronger intervention effects when the most disadvantaged groups and younger children were targeted (Baker-Henningham & Lopez Boo, 2010; Engle et al., 2011; Khatib et al., 2020), these conclusions should be considered tentatively, given low confidence ratings in the quality assessment for two of these reviews.

Our analysis of common elements identified a core set of elements used across 28 early childhood interventions. The most commonly occurring elements were, in order, caregiver psychoeducation, play/pretend, talking to baby, responsive parenting, use of toys, and attachment building, each present in ≥ 75% of protocols of effective interventions. Synthesizing the common elements present across most effective interventions in LMIC settings can enable conversations about how these components may bolster parents’ skills across varied settings, as well as where gaps exist in applying the most up-to-date evidence.

Overall, these elements aligned with literature on best practices for child development and parenting, reiterating that many of the trials of interventions assessed were supported by evidence-based practices and translatable to LMICs. Our analysis elicits key considerations for implementing ECD interventions in LMIC settings and highlights several important provisions to support. In considering the combination of responsive care with a focus on learning opportunities and key milestones, we found that many interventions integrated the twin aims of ECD interventions as articulated by Aboud and Yousafzai: providing early learning opportunities for children, and teaching responsive parenting (Aboud & Yousafzai, 2019). Of the 25 interventions that included the element of responsive care, 84% (n = 21) included toys use, and 100% included play/pretend—indicating that interventions are providing opportunities for both early learning and play, as well as responsiveness and sensitivity to children’s cues.

Our analysis also draws out elements that support optimal child development in settings that are significantly resource-constrained. Firstly, elements such as attachment building and talking to baby may be particularly important for promoting socioemotional development in the context of additional environmental stressors. Additionally, it may be valuable to consider integrating a nutritional component—which many interventions did through direct nutritional supplementation or psychoeducation about child feeding (Andrew et al., 2020; Atukunda et al., 2019; Galasso et al., 2019; Gardner et al., 2005). In settings where child stunting is high, combined approaches to child development may be essential to safeguard gains from ECD interventions. These intervention approaches are often complementary (Black & Aboud, 2011). Lastly, where resources are constrained, task-shifting to non-specialists may be essential (Gilmore & McAuliffe, 2013; Seidman & Atun, 2017). Over half of the included interventions (n = 18, 62%) were designed for delivery by community-based peers who received program-specific training. Specific skills informed by evidence-based approaches may be particularly important in ensuring intervention acceptability and uptake in LMIC settings. For example, interpersonal “soft” skills, paired with an educational dimension, have been shown to be important in home visiting interventions in low-literacy settings (Laurenzi et al., 2019).

Implications for Practice and Future Directions for Research

This review has several implications for practice and policy. In the field of early intervention, new programs are routinely branded as “new interventions” resulting in a plethora of interventions. In resource-constrained settings, common element approaches that are more generic and transdisciplinary are key. The evidence base is mature enough to make a strong case that efficacy and effectiveness studies are no longer needed, especially to evaluate distinct, “new” interventions. Implementation science evidence could support the rollout of existing effective interventions at larger scale (Rotheram-Fuller et al., 2017).

Future research should investigate how intensity and duration of interventions may be linked with child outcomes. Several included reviews (Britto et al., 2015; Engle et al., 2011) emphasized how longer, higher-intensity psychosocial interventions can protect and buffer against cognitive developmental delays in the most disadvantaged children—including undernourished and low birthweight children, as well as those living with HIV. Common elements analyses should investigate “dose–response” relationships between interventions and outcomes, to enable cost-effectiveness considerations. Understanding the minimal intensity required to achieve intended effects is important, as longer exposure typically results in more consistent and larger effects on child development (Engle et al., 2007), yet may drive up costs. Importantly, dosage should be carefully considered alongside quality and participant engagement, without assuming that more exposure will result in better outcomes. Similarly, long-term follow-ups of child participants are needed to investigate specific long-term effects of psychosocial interventions.

Building on the idea that evidence-based interventions share common elements that can be distilled through this kind of analysis, future research should explore the “matching” part of the distillation and matching model developed by Chorpita et al. (Chorpita et al., 2005). The model suggests that following distillation, clinicians can select common elements that apply to a particular problem or target population (Chorpita et al., 2007). To do this effectively in LMIC contexts, common element analyses would need to expand beyond content to focus more specifically on intervention implementation, including frequency, delivery method (group, individual, home visits), duration, training and supervision of implementers, characteristics of the target populations, cultural adaptation, and context. The age and developmental stage of the child, too, may influence which components are applied, or how they may be adapted to be best matched. Common element research could also focus on the process of adaptation to specific contexts, to identify how interventions can be planned and replicated in new settings.

Limitations of This Review

This analysis has several limitations. The substantial heterogeneity of most systematic reviews limits the generalizability of the findings; however, the overall pattern of evidence strongly supports early interventions in improving children’s cognitive outcomes. Because many interventions were delivered as part of larger programs incorporating multiple elements to improve parental sensitivity and responsiveness, it is difficult to attribute direct effects of specific intervention elements on children’s cognitive outcomes. Our analysis of common elements suggests that benefits for children’s cognitive outcomes likely accrue when interventions target children’s needs holistically (including nutrition, stimulation, parental responsiveness, hygiene) (Black et al., 2017; Walker et al., 2011). Nevertheless, future research will benefit from the work undertaken here, by meta-analytically analyzing program effects according to the elements they contain. This, alongside other approaches, will help identify which program elements are critical for delivering optimal outcomes. Adding an age-specific dimension to these analyses may also help to further disentangle differential effects of components on children of different ages and developmental stages; however, as Jeong and colleagues note in their review, parenting and early stimulation interventions have been found to be beneficial no matter the age of the child nor the timing at which the intervention is introduced to the child (Jeong et al., 2021).

Our common element analysis also faced certain limitations. Using the PracticeWise manual as a key source, our coding scheme focused predominantly on content; however, we did not systematically extract data on frequency, session duration, or other measures of intervention intensity. Assessing intervention intensity may be important in understanding treatment effects at a more granular level. The distillation and matching model also has some inherent limitations, such as key elements being identified based on presence or absence in a given intervention, instead of dosage or how well an element is integrated (Chorpita et al., 2007).

We also relied on different data sources to populate the common elements database, including both intervention protocols/manuals and peer-reviewed publications. Diverse sources, including supporting evidence, enabled us to triangulate data about which elements were present in each intervention. However, these sources also reduced the level of standardization by presenting different reporting practices. Where we relied on published papers for description of the intervention, the coding was less detailed than where an intervention manual was available, and may not have been as rich in description, meaning that some information about the intervention contents may have been missed.

Despite these limitations, the results of the review offer valuable insights into elements of effective early psychosocial interventions conducted in LMICs.

Conclusion

As researchers and policymakers seek effective, and cost-efficient, ways to improve child development—and to safeguard gains from the past two decades as the COVID-19 pandemic alters the global economic landscape—it is critical to gather evidence about what works to support positive outcomes for child development in LMICs. Given resource constraints, continuing to develop novel interventions is unnecessary and expensive. Continued efforts to build evidence around the most effective ingredients in interventions to improve child development should seek to optimize and streamline these elements to be more accessible to a greater number of people.