Objective To determine the effect of early childhood development interventions delivered by healthcare providers (HCP-ECD) on child cognition and maternal mental health.
Design Systematic review, meta-analysis.
Setting Healthcare setting or home.
Participants Infants under 1 month of age.
Interventions HCP-ECD interventions that supported responsive caregiving, early learning and motor stimulation. MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, Health Technology Assessment Database, Database of Abstracts of Reviews of Effects and Cochrane Database of Systematic Reviews were searched until 15 November 2021. Studies reporting prespecified outcomes were pooled using standard meta-analytical methods.
Main outcome measures Cognitive development in children at 0–36 months.
Results Forty-two randomised controlled trials with 15 557 infants were included in the narrative synthesis. Twenty-seven trials were included in the meta-analyses. Pooled data from 13 trials suggest that HCP-ECD interventions may improve cognitive outcomes in children between 0 and 36 months (Bayley Scales of Infant Development version IIII (BSID-III) mean difference (MD) 2.65; 95% CI 0.61 to 4.70; 2482 participants; low certainty of evidence). Pooled data from nine trials suggest improvements in motor development (BSID-III MD 4.01; 95% CI 1.54 to 6.48; 1437 participants; low certainty of evidence). There was no evidence of improvement in maternal mental health (standardised MD −0.13; 95% CI −0.28 to 0.03; 2806 participants; 11 trials; low certainty of evidence).
Conclusions We report promising evidence, particularly for cognitive and motor outcomes, of the effect of HCP-ECD interventions. However, effect sizes were small, and the certainty of evidence ranged from very low to moderate. Additional high-quality research is required.
PROSPERO registration number CRD42019122021.
- Child Development
- Primary Health Care
Data availability statement
Data sharing not applicable as no datasets generated and/or analysed for this study. Systematic review.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Globally, more than 40% of disadvantaged children under 5 years have neurodevelopmental problems resulting in social, emotional and educational functioning deficits into adulthood.1–3 The WHO defines early childhood development (ECD) interventions as physical, socioemotional, cognitive and motor development interventions implemented between birth and 8 years of age.4–10 The importance of the family and social environment in influencing children’s neurodevelopment is well-known. However, the impact of health services on the neurodevelopment of children, particularly primary care (the first level of the health system), is less well understood.11 12 Healthcare providers (HCPs) working in primary care, including community health workers, generalist nurses, health visitors, midwives, child health nurses and general practitioners, are uniquely positioned to augment early child development. However, many lack skills and confidence in neurodevelopmental care and few receive appropriate training, education and resources.13 14 HCP-delivered ECD interventions (HCP-ECD) include: WHO’s Care for Child Development package, family partnership working and motivational interviewing.15–20
Four systematic reviews have examined the effectiveness of ECD interventions to improve early child development.21–24 Most recently, a systematic review of 102 studies reported that parenting interventions improved a range of ECD outcomes at 3 years.24 However, these reviews had various individuals delivering the ECD interventions such as peer counsellors, family support workers, HCPs and researchers. To our knowledge, there have been no systematic reviews that have examined the effect of ECD interventions delivered solely by an HCP (HCP-ECD) to families in high-income country (HIC) and low/middle-income country (LMIC) settings.
There is a growing body of evidence that babies develop important communication and social behaviours within the first days and weeks of life, especially eye contact, visual locking, auditory responses, responsiveness and self-quietening behaviour.4 25–29 Systematic reviews have assessed the effect of interventions delivered in the antenatal period.23 24 However, to our knowledge, there have been no reviews of the effects of ECD interventions in a subgroup of babies who received ECD interventions in the neonatal period from 0 to 28 days (‘neonatal ECD’).23 24 The optimal number of visits or contacts (‘dose’) and types of ECD interventions delivered in the neonatal and infant periods is also not known.
The primary objective of this review was to assess effects of HCP-ECD on cognitive outcomes in children at 0–36 months. Secondary objectives were to assess effects on (1) childhood neurodevelopmental domains (speech, language, fine motor, gross motor, social emotional, behaviour) at 0–36 months; (2) maternal mental health at 0–36 months and (3) in prespecified subgroups (number and timing of infant and neonatal contacts, type of intervention, income level of country).
The protocol was registered in PROSPERO (CRD42019122021), and the detailed protocol is published separately.30 Preferred Reporting Items for Systematic Reviews and Meta-Analyses-Protocol (PRISMA-P) guidance was followed.31 Modifications made from the original protocol are provided in online supplemental appendix 1.
We searched the following databases with no restrictions to time periods and language: Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, the Cochrane Database of Systematic Reviews, Health Technology Assessment Database and the Database of Abstracts of Reviews of Effects. We also searched clinical trial registries. Reference lists from included studies and relevant systematic reviews were inspected for additional citations. The search was completed on 15 November 2021. The search strategy is presented in online supplemental appendix 2.
The HCP-ECD interventions had to be delivered by primary-level HCPs (eg, generalist nurses, health visitors, midwives, child health nurses, general practitioners, primary care doctors, community health workers). The interventions could commence in the hospital but had to include community-based post-discharge follow-up.32 Interventions were required to be face to face in nature, for example, delivered through home visiting, mobile health team visits, clinic visits, child health checks or group programmes. The comparator group was ‘no HCP-ECD interventions’, that is, any other care, standard care that did not include ECD or no care. Only individual, cluster and quasi-randomised controlled trials (RCTs) were eligible for inclusion.
We used WHO definitions and classified the ECD interventions into three categories: responsive caregiving, early learning support and motor stimulation.10 We also classified interventions as: any responsive caregiving, no responsive caregiving, and ECD predominant and ECD non-predominant. ECD predominance was defined as ECD implemented for more than 50% of the contact time (table 1).
The primary outcome measure was cognitive development in children at 0–36 months of follow-up. Secondary outcomes were: (1) speech, language, fine motor, gross motor, social, emotional, behaviour, executive functioning and adaptive functioning; and (2) maternal mental health. Studies were included in the systematic review regardless of the type of outcomes. However, only standardised measures, for example, the Bayley Scales of Infant and Toddler Development or the Griffiths Mental Development Scales for cognitive development, were used in the meta-analyses.
Our a priori primary analysis was the period between 0 and 36 months where an infant received assessment for outcomes ‘at latest follow-up’. We did this to ensure that the maximum amount of data could contribute to the primary outcome, that is, that all studies with follow-up could be included regardless of the duration of follow-up. We expected that the duration of follow-up would vary across studies so we reported the mean (SD) and median (IQR) duration of follow-up for each outcome and presented this in each forest plot. For completeness, we also assessed effects at 12, 24 and 36 months of follow-up. However, these time points were not prespecified as primary or secondary outcomes. A priori, we expected that these results would be underpowered and imprecise with wide CIs.
We assessed effects on cognitive development in children at 0–36 months in seven prespecified subgroups: (1) number of contacts in the neonatal period (one contact, two contacts, three or more contacts); (2) timing of contact (first week, second week or later); (3) antenatal period exposure (intervention delivered in the antenatal period, intervention not delivered in the antenatal period); (4) type of intervention (responsive caregiving, early learning support and motor stimulation) (any responsive caregiving, no responsive caregiving) (intervention predominantly ECD, intervention not predominantly ECD); (5) type of HCP (child health workers, nurse (including general nurse and child health nurse), child health workers and others); (6) income level of the country (HIC, LMIC); and (7) risk of bias (high risk of bias, some concerns of bias).
Study selection and data collection process
All titles, abstracts and full-text articles were reviewed and extracted independently by two review authors. Discussions with a third author were used to resolve any disagreement. Standardised pretested data collection forms were used. Data collected were: study design, study setting, intervention components, participant demographics and outcomes.
Risk of bias assessment
Two independent review authors used the Cochrane risk-of-bias assessment tool to assess the risk of bias.33 We also assessed meta-biases, including publication bias and selective reporting. No studies were excluded based on risk of bias assessment.
Data management and statistical analysis
We searched for both continuous and dichotomous data for all outcomes (online supplemental appendix 3 and 4). In the meta-analyses, we reported mean differences (MDs) for continuous data if they were measured on the same scales and standardised MDs (SMDs) for outcomes that were reported on different scales. Relative risks were reported for dichotomous data. We contacted authors where possible to request data.
Random-effects models were used with restricted maximum likelihood estimates and Knapp-Hartung SEs. Where possible, we imputed data using standard methods. We used the I² statistic to measure heterogeneity among the primary and secondary outcomes of all included trials. An I² value of >50% was considered to represent substantial heterogeneity. For outcomes with at least 10 studies, funnel plots and Egger’s test were used to assess publication bias and small study effects, respectively. We completed an unadjusted random-effects meta-regression with Knapp-Hartung SEs on the primary outcome for the number of expected visits (‘doses’). Statistical analyses were performed using STATA V.16.1 statistical software (Stata, College Station, Texas, USA).
Grading of evidence
We used the principles of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system to assess the quality of the body of evidence associated with specific outcomes which included assessment of risk of bias, consistency of effect, imprecision, indirectness and publication bias.34
Role of funding source
After the removal of duplicates, 9401 papers were eligible for inclusion (PRISMA flow diagram, online supplemental appendix 5). After assessing inclusion and exclusion criteria, 97 papers reporting on 42 trials were included in the narrative synthesis, of which 27 trials were included in the meta-analyses (online supplemental appendix 5).15 35–130 Online supplemental appendix 3 displays the outcomes and scales reported by each study and those included in the meta-analyses. Online supplemental appendix 6 shows the ongoing studies.
Of the 42 trials, 38 were individual RCTs,37 38 44–49 54–77 79–122 124–129 and 4 were cluster RCTs15 35 36 39–43 50–53 78 123 130 (table 2). Thirty-three trials were conducted in HICs44–49 54–67 69–77 79–97 100 103 105–108 111–119 121 122 125 126 128 and eight were conducted in LMICs (Pakistan, Bangladesh, South Africa, Columbia, Jamaica, Brazil, India and Zambia).15 35–38 41–43 50–53 68 78 96 98 99 101 102 104 109 110 120 123 124 127 129 130
A total of 15 661 infants participated in the 41 trials; 7857 intervention and 7804 comparison. There were 12 118 infants from HICs, 3136 from MICs and 407 from LMICs. All infants were from families experiencing some level of adversity such as low socioeconomic status, maternal drug abuse, adolescent mothers or were premature (table 2).
Thirty-six trials used a single HCP to implement the intervention (4 trials used child health nurses; 17 used generalist nurses; 4 used health visitors; 17 used community health workers; and 6 used multidisciplinary healthcare teams including child health nurses, general practitioners and generalist nurses) (online supplemental appendix 4). Forty trials used home visits and two used community clinics to implement their ECD intervention. The number of contacts in the trials varied from 6 to 312 (median 25, IQR 9–52), with the interventions lasting between 6 weeks and 36 months (mean 19.7, SD 3.21). The number of contacts in the neonatal period ranged from one to four. Twelve trials included contact in the antenatal period. Most interventions were classified as responsive caregiving only (15 trials) or early learning support only (18 trials). Nineteen trials provided responsive caregiving along with other interventions and 19 were classified as predominantly ECD (online supplemental appendix 4).
Risk of bias
For assessor-reported outcomes, 5 trials15 74 113 119 121 had moderate risk of bias and the remaining 20 trials had high risk of bias (online supplemental appendix 7). For patient-reported outcomes, 2 trials15 119 had moderate risk of bias and 27 had high risk of bias (online supplemental appendix 7). There was no evidence of publication bias or small study effects shown for any outcome including the cognitive development outcome (Egger’s test p=0.17) and maternal mental health outcome (Egger’s test p=0.10) (funnel plots, online supplemental appendix 7).
Data for the primary analysis are presented in table 3 and figure 1. The GRADE summary of findings is presented in online supplemental appendix 8. Pooled data from 13 trials suggest that HCP-ECD compared with usual care improved cognitive outcomes in infants at follow-up of 0–36 months (Bayley Scales of Infant Development version IIII (BSID-III) MD 2.65; 95% CI 0.61 to 4.70; 2482 participants; low certainty of evidence). We downgraded one level for heterogeneity (I2=63%) and one level for risk of bias (six trials had a high risk of bias in the selection of the reported result and two trials had a high risk of bias in outcome measurement). No publication bias was reported. There was little to no evidence of an effect of HCP-ECD interventions at 12 months, 24 months and 36 months of follow-up (table 3 and online supplemental appendix 9). However, these analyses had small sample sizes and wide CIs and were downgraded for imprecision and risk of bias.
Data from the secondary analyses are presented in table 3 and online supplemental appendix 10. Pooled data from nine trials suggest that HCP-ECD improves motor outcomes in infants aged 0–36 months (BSID-III MD 4.01; 95% CI 1.54 to 6.48; 1437 participants; moderate certainty of evidence). Pooled data from eight trials suggest that HCP-ECD improves home environments for children at 0–36 months (HOME inventory scales131 MD 1.37; 95% CI 0.29 to 2.45; 1534 participants; low certainty of evidence).
There was little to no effect on maternal health (SMD −0.13; 95% CI −0.29 to 0.03; 2806 participants; 11 trials; low certainty of evidence); speech and language (SMD 0.30; 95% CI −0.53 to 1.13; 1551 participants; 3 trials; very low certainty of evidence); socioemotional (Ages and Stages Questionnaire-Social Emotional scales MD −0.91; 95% CI −27.72 to 25.89; 369 participants; 2 trials; very low certainty of evidence) or infant behaviour outcomes (SMD 8.34; 95% CI −31.20 to 47.88; 1769 participants; 3 trials; very low certainty of evidence). No studies reported on executive or adaptive functioning.
There was no evidence of differences in the effect of HCP-ECD on the primary outcome (cognitive development) in any subgroup (number of contacts, timing, type of intervention, type of HCP, income level of country, risk of bias) except for ECD predominance (ie, ECD implemented for more than 50% of the contact time between HCP and family) (table 3 and online supplemental appendix 11). The effect of ECD-predominant interventions (BSID-III MD 3.31; 95% CI 0.74 to 5.88; 1672 participants; 10 trials) was greater than the effect of interventions that were not ECD predominant (BSID-III MD 0.27; 95% CI −1.62 to 2.16; 810 participants; 3 trials) (Χ2 statistic 4.16, p=0.04). No other differentials in effect were found for any other subgroup analysis. In particular, there was no evidence of a ‘dose–response’, that is, an effect of HCP-ECD by number of expected HCP visits (β coefficient 0.018; 95% CI −0.07 to 0.11, 1811 participants; 12 trials; table 3).
Few studies reported dichotomous outcomes. These analyses had wide CIs and were limited by imprecision. Results are presented in online supplemental appendices 9 and 10.
Our systematic review of 15 557 infants aged 0–36 months in 42 trials showed that HCP-ECD interventions may improve cognitive and motor outcomes and the quality of the home environment for infants aged 0–36 months across HICs and LMICs. No effect was seen on speech, language, socioemotional, behaviour or maternal mental health outcomes.
Our effects on cognitive outcomes (MD 2.65; 95% CI 0.61 to 4.70) at 36 months appeared greater than the four recent parenting reviews which reported SMD scores ranging from 0.25 to 0.42.23 24 132 133 We prespecified the combined period of follow-up of 0–36 months as our primary outcome to ensure that the maximum amount of data could contribute to the primary outcome, that is, all studies could be included regardless of the duration of follow up. The other analyses at 12 months, 24 months and 36 months were downgraded for imprecision due to small sample sizes and wide CIs and showed little to no evidence of an effect of HCP-ECD interventions at 12 months, 24 months and 36 months of follow-up.
Effects on motor development were similar to other reviews.22 23 No effects were seen on language, behaviour and socioemotional development domains. However, few trials assessed these outcomes (speech (two trials, 354 infants), language (two trials, 369 infants) and social and emotional development (three trials, 1769 infants)). The trials also had wide CIs and we downgraded the certainty of the evidence two levels for imprecision. We found no impact of HCP-ECD on maternal mental health. This is similar to most other reviews of ECD interventions,22–24 134 and could be because ECD interventions do not include techniques that directly address parental mental health, such as behavioural activation and cognitive–behavioural therapy. However, we did show that HCP-ECD interventions improved home environment scores. Forty of the 42 studies used home visits as the main delivery channel which may be an important mechanism, though further research is needed.
Trials that fulfilled the definition of ‘ECD predominance’ (ECD implemented for more than 50% of contact time) had a greater effect on child neurodevelopment than trials with ECD implemented for less than 50% contact time. However, caution is needed in interpreting these results due to unexplained heterogeneity, especially in the ECD-predominant group (I2=68%). There was no differential effect by type of intervention (responsive caregiving, learning support or other), antenatal contact or timing of neonatal interventions. However, these subgroup analyses had small sample sizes and limited power to detect effects.
There were a number of methodological limitations in the trials included in our meta-analysis. Using the GRADE system,34 we judged that the evidence for our primary outcome was low certainty due to risk of bias and heterogeneity. Many different scales were also used for measurement of child neurodevelopment and maternal mental health. However, we found sufficient data for pooling using SMDs or MDs for the follow-up period of 0–36 months. We also did not find publication bias or small study effects for our primary and secondary outcomes. All the ECD interventions in our systematic review were delivered to infants facing adversity including: poverty, maternal drug abuse and preterm birth. However, these situations are unfortunately not uncommon, and children facing these types of adversities are most in need of ECD interventions. Our study also had a number of other strengths. We included 12 013 infants and 27 trials in our meta-analyses. Our search was intentionally broad to capture all relevant studies, and we did not limit our search geographically, by language or by intervention approach. The interventions were delivered by a range of healthcare workers, including community health workers, generalist nurses, general practitioners and health visitors, making the findings relevant across many settings.
To our knowledge, this is the first systematic review and meta-analysis that has examined the impact of HCP-ECD interventions across HICs and LMICs. We report evidence of impacts on child neurodevelopment. Importantly, our review shows ‘what the health system can do’ to improve neurodevelopmental outcomes in the first 3 years of a child’s life. This is especially important as HCPs (such as midwives and child health nurses) have multiple contacts with the mother and child in the first 3 years of life and are well placed to integrate and support maternal health as well as ECD.
We believe a sustained long-term commitment to ECD from governments and donors that focuses on three core ECD interventions (responsive caregiving, early learning support and motor stimulation) could quickly accelerate the gains we reported in our meta-analysis. More investment is also needed to train and build the skills and confidence of HCPs in neurodevelopmental care.13 14 Many countries have committed to reaching the 2030 United Nations Sustainable Development Goal for ECD.5 Our findings suggest that the health system has a potentially important role to play in achieving this goal, especially in the early years.
Data availability statement
Data sharing not applicable as no datasets generated and/or analysed for this study. Systematic review.
Patient consent for publication
We gratefully acknowledge Associate Professor Aisha Yousafzai who responded to our inquiries and sent us data.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
KME and NS are joint senior authors.
Contributors KME conceived the idea for the review. NS designed and undertook the search. RH, NS, CA, CBK, DM and LH completed the review of abstracts, full text and data extraction. RH, NS, CA, CBK and LH completed the risk of bias on all studies. RH completed the statistical analysis, figures and appendices with support from NS and KME. RH wrote the first draft of the manuscript with input from NS and KME. LH and KME provided content expertise. All authors reviewed and revised subsequent drafts.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; internally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.
- Global child health
- Highlights from this issue