Description of the condition
A large proportion of children under five years of age in high-income countries experience significant non-parental care. Specifically, an estimated 80% of children in the United States regularly attend day care (NICHD ECCRN 2006), with similar figures reported in the United Kingdom and Australia (CCCH 2009; Smith 2010). In addition, almost 50% of children aged three to four years in the United States are in full-time day care (at least 35 hours a week) (Capizzano 2005). Parents often choose day care for economic reasons (for example, to enable them to work, look for work, or study) (Smith 2010). For the estimated 20% of single-parent families in high-income countries (OECD 2010), this may be particularly pertinent. In addition, parents might choose day care to improve their child’s social and academic skills before entry into formal schooling (Lamb 2006). Finally, centre-based day care can also be the setting for early interventions to target and enhance the social, cognitive, and academic development of disadvantaged young children (Campbell 2001).
Description of the intervention
Day care for children in high-income countries takes various forms, and may serve multiple purposes. For children under three years old, services tend to be care-oriented and include care by professionals, childminders, and day nurseries. For children aged three years and older, day care provision has more focused educational aims and includes preschools, playgroups and nursery classes (Melhuish 2004). The quality, quantity, and type of day care service appear to be differentially related to child outcomes (NICHD ECCRN 2005; Belsky 2007). Therefore, whilst there is an increasing overlap between education- and care-focused services, it is important to distinguish between studies assessing children under three years and children older than three years, as the nature of the services provided and results may differ. It is also important to distinguish between formal centre-based care (for example, in nursery schools or preschools) and informal care (for example, care by an ex-partner or grandparent) and consider whether the effectiveness of provision differs across settings, including rural versus urban settings or areas of high versus low socioeconomic status (Smith 2010). In addition to differences across settings and type of provision, the explicit purposes for day care can also differ. First, day care enables parental, and specifically maternal, employment. Provision of day care is correlated with increased female labour force participation in high-income countries (Gelbach 2002; Esping-Andersen 2009). Second, day care may also impact on the long-term cognitive and socioemotional development of children, particularly for children from deprived homes (Dearing 2009).
How the intervention might work
Research over many decades has highlighted that the reasons why day care is provided and the mechanisms by which it affects children’s development vary. First, a considerable body of research has investigated whether day care disrupts secure mother-child attachment, which could have long-term developmental consequences for children (Ainsworth 1978; Sroufe 1999). However, day care may serve to enhance the early physical and emotional environment of children, which could be beneficial for children growing up in otherwise impoverished or adverse circumstances. Some studies have found that day care directly affects children’s psychosocial outcomes, including improved social competence (for example, Clarke-Stewart 1994; Balleyguier 1996), which may have lasting effects for mental health outcomes over a lifetime (Shonkoff 2011)). However, other longitudinal studies have found that more time in day care predicts higher levels of externalising behavior, including aggression and non-compliance (NICHD ECCRN 2006; Philips 2006; Belsky 2007).
Second, day care may serve to directly enhance child outcomes via cognitive development, which may have collateral effects on future educational attainment and adult outcomes. Specifically, school readiness and cognitive capacities appear to be enhanced by structured activities, psychosocial stimulation (NICHD ECCRN 2006), and responsive, verbally articulate staff within day care settings (Melhuish 2004). High quality centre-based care has also been linked to improved language development (Clarke-Stewart 1987; Schliecker 1991). In particular, language learning appears to be facilitated in day care settings when children are afforded more opportunities to interact verbally with adults and peers. Longitudinal studies have also demonstrated that early experience of high quality day care as part of multi-component interventions, particularly by low income children, predicts better academic outcomes, higher rates of employment, and less adult criminal activity (for example, Schweinhart 1993; Campbell 2001). At the same time, some studies have also found poorer language development to be associated with increased day care experience (Brooks-Gunn 2002; Bainbridge 2003). Indeed, low quality day care settings may see large numbers of children cared for by highly stressed staff, thereby reducing the potential effectiveness of day care as an intervention to improve child psychosocial or cognitive outcomes (Raver 2004).
Finally, day care is typically offered as a means to improve household income. Mothers may be able to fully participate in the labour market when they feel their children are secure and cared for (Vandell 2002; Melhuish 2004). Indeed, provision of day care is correlated with increased female labour force participation in high-income countries, and an earlier return to the workforce after pregnancy (Brooks-Gunn 1994; Gelbach 2002; Esping-Andersen 2009). There is also likely to be a range of indirect effects of improved household income on child outcomes, including improved nutrition and a more enriched home environment.
Why it is important to do this review
A significant proportion of children in high-income countries under the age of five years experience significant non-parental day care within formal and informal settings (Melhuish 2004; NICHD ECCRN 2006; Smith 2010). It is therefore important to evaluate the effects of day care on the cognitive, linguistic, educational, socioemotional, attachment, and physical health outcomes of children and its impact on parental employment and family income. The only Cochrane review on the topic was carried out a decade ago (Zoritch 2000) and an up to date assessment of the state of the evidence is needed. Furthermore, the previous review included co-interventions beyond day care (such as parent training and home visits), limiting conclusions regarding the effects of day care services alone. To best understand its effects, the intervention of centre-based day care must be isolated from additional programme components such as home visits and other interventions that are not centre-based. The effects of additional components that are centre-based (for example, educational programmes) also need to be analysed. Finally, possible social and economic confounding variables must be controlled for, as specified in the content of this protocol. A new review will guide policymakers and parents in decision-making about centre-based day care for children under five years old.
To assess the effects of centre-based day care on the development and well-being of children and their families in high-income countries (as defined by the World Bank (World Bank 2011)).
Criteria for considering studies for this review
Types of studies
Randomised controlled trials and quasi-randomised controlled trials.
Types of participants
Children under five years (at the time of enrollment) and their families in high-income countries.
Types of interventions
We will include centre-based day care, defined as supervision of children in a publicly accessible location, with or without snack and meal provision and/or a child education component.
We will exclude studies that include co-interventions not directed toward children (for example, parent programmes, home visits, and teacher training); the presence of such programmes would weaken the extent to which findings can be attributed to centre-based care alone. We will also exclude studies in which enrollment was limited to children with physical or intellectual disabilities (for example, autism or IQ under 80), orphans, children living in hospital, or children with HIV/AIDS.
Types of outcome measures
We will assess the effects of centre-based day care on child and family well-being outcomes by extracting data on the following outcomes. In studies reporting more than one measure of an outcome, we will extract data for meta-analysis using methods described below (see Measures of treatment effect). The primary outcomes and the first secondary outcome will be included in 'Summary of Findings' tables.
1. Child intellectual development
1.1. Cognitive ability (IQ or developmental quotient)
1.2. Educational attainment (for example, measures of reading, writing, or mathematics, and retention in grade)
2. Child psychosocial development
2.1. Any behavioural measure (that is, self, parents’ and teachers’ reports of externalising behaviour/aggression, prosocial or antisocial behavior)
2.2. Disrupted child attachment to mother
3. Maternal and family outcomes
3.1. Paid parental employment
3.2. Household income
4. Child long-term outcomes
4.1. High school completion
Search methods for identification of studies
Studies will be considered regardless of publication status or language, though all searches and author communications will be conducted in English. Foreign language abstracts associated with titles of interest will be translated by fluent speakers of the relevant foreign language. Study reports will be discussed with a fluent speaker and translated if the study may meet the inclusion criteria.
We will search the following databases.
- Cochrane Register of Controlled Trials (CENTRAL), part of The Cochrane Library
- Ovid MEDLINE(R)
- Social Sciences Citation Index (SSCI)
- Conference Proceedings Citations Index - Social Sciences & Humanities (CPCI-SSH)
- Global Health Library
- WorldCat (limited to theses and dissertations)
- International Clinical Trials Registry Platform (ICTRP)
We will use the following search strategy in Ovid MEDLINE and adapt it for other databases using appropriate controlled vocabulary and syntax. This strategy includes the Cochrane recommended filter for identifying randomised trials (Lefebvre 2011).
1. child day care centers/
2. Schools, Nursery/
3. "Early Intervention (Education)"/
4. ((early adj2 education$) or ECCE).tw.
5. (creche$ or nurser$ or kindergarten$ or kinder-garten$ or preschool$ or pre-primary or preprimary or playgroup$ or play-group$ or pre-school$ or (child$ adj3 centre$) or (child$ adj3 center$)).tw. (32921)
7. child care/ or child care.tw.
8. (centre$ or center$ or facilit$ or "out of home" or polic$ or program$ or scheme$).tw.
9. 7 and 8
10. exp child/
11. exp Infant/
12. (infant$ or baby or babies or toddler$ or child$ or boy$ or girl$ or kid$ or pre-kindergarten$ or prekindergarten$ or preschool$ or pre-school$).tw.
14. Day Care/
15. daycare$ or day-care$ or daycentre$ or daycenter$ or (centre-based adj3 care$) or (center-based adj3 care$) or (day$ adj3 (centre$ or center$))).tw.
16. 14 or 15
17. 13 and 16
18. 6 or 9 or 17
19. randomized controlled trial.pt.
20. controlled clinical trial.pt.
27. exp animals/ not humans.sh.
28. 26 not 27
29. 18 and 28
Searching other resources
We will examine reference lists from previous studies. We will contact the authors of all included studies to request details of ongoing and unpublished studies.
Data collection and analysis
Selection of studies
Two review authors will independently review all titles and abstracts. Relevant articles will be collected and independently screened to determine which studies meet the inclusion criteria. We will contact study authors if further information is required. Disagreements will be resolved through discussion and consultation with other review authors.
Data extraction and management
Two review authors will independently extract the following data from all included studies. Disagreement will be resolved through discussion and consultation with other review authors.
- Year of study
- Study design (that is, case control, cohort)
- Unit of analysis (for example, individual- or cluster-randomised)
- Methods used to control for confounding factors
- Setting (that is, urban or rural, specific region or city if provided)
- Number of study participants and clusters randomised to each included group
- Inclusion and exclusion criteria
- Household income (if reported)
For each intervention or comparison group of interest:
- Dose of centre-based care
- Duration of centre-based care
- Frequency of centre-based care
- Co-interventions (if any)
- Quality of care (if measured)
For each study, we will use the Cochrane recommended lists for identifying study design (Higgins 2011), and report characteristics of study designs in a ‘Characteristics of included studies’ table.
Assessment of risk of bias in included studies
Two review authors (TWB and FVU) will code each included study using the Cochrane tool for assessing risk of bias (Higgins 2011, section 8.5.a), including: sequence generation; allocation concealment; blinding of study participants, personnel, and outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias. In addition to these, we will account for the risk of bias due to confounding and for outcome validity. We will judge the risk of bias in each category by a rating of low, high or unclear. A rating of 'low' shall indicate that evidence was sufficient to judge that study authors used appropriate methods to avoid bias, as determined by the Cochrane Handbook for Systematic Reviews of Interventions criteria for judging risk of bias in the 'Risk of bias' assessment tool (Higgins 2011, section 8.5.c); a rating of 'high' shall indicate that evidence was sufficient to judge that study authors did not use appropriate methods to avoid bias; and a rating of 'unclear' shall indicate that there was insufficient information to judge the extent to which study authors used the appropriate methods to avoid bias.
We will report assessments of confounders using additional tables that will identify which confounding factors were considered and controlled for in each study. Confounding factors that will be explicitly assessed and reported include: age of child, sex of child, household income, quality and fees of the day care centre.
Disagreements will be resolved through discussion with a third author (EMW) and, if necessary, a fourth author (BW).
Measures of treatment effect
Studies often report outcomes using multiple definitions and outcome measures. We will give preference to data that involved the least manipulation by authors or inference by review authors; that is, we will extract raw values (for example, means and standard deviations) rather than calculated effect sizes (for example, Cohen’s d). If outcomes are reported as final values and as changes from baseline, we will extract the final values.
For studies with multiple time points we will include the time point that occurred the greatest number of days after randomisation. If possible, we will also conduct an analysis of prespecified time points: up to 25 months, 25 months or more.
We will calculate relative risks or rate ratios (RR) and 95% confidence intervals for dichotomous outcomes. When risk ratios or rate ratios cannot be calculated (when total sample size is unknown), we will calculate odds ratios (OR). If we cannot calculate RR for all studies included in an analysis, but can calculate OR for all studies, we will report OR for all studies included in that analysis. We will not meta-analyse RR and OR together (see Data synthesis). We will give preference to denominators in the following order: events per person-year, events per person with definite outcome known (or imputed, as described in Dealing with missing data), events per person randomised.
We will use Hedges' (adjusted) g (a standardised mean difference) for each outcome for which there is continuous data.
Unit of analysis issues
Some data in this review may come from cluster-randomised trials, which randomise groups of people rather than individuals. For each cluster-randomised trial, we will first determine whether or not its data incorporate sufficient controls for clustering (such as robust standard errors or hierarchical linear models). If the data do not have proper controls, then we will attempt to obtain an appropriate estimate of the intracluster correlation coefficient (ICC). If we cannot find an estimate in the report of the trial, then we will request an estimate from the trial report authors. We will use the ICC estimate to control for clustering, according to procedures described in Higgins 2011.
Dealing with missing data
For all analyses, we will attempt to include all study participants, and we will contact authors to request data including all participants randomised for all outcomes. When analyses are reported for completers as well as controlling for dropout, we will extract the latter. If participant data are missing in a study, or if reasons for dropout are not included, then we will contact the study authors for additional information. For studies with dichotomous data, if no information can be gathered from authors, we will assume that participants in all groups for whom data are missing experienced negative outcomes. All missing data will be recorded on the data extraction sheet and reported in the 'Risk of bias' tables.
Assessment of heterogeneity
Differences among included studies are discussed in terms of their participants, interventions, outcomes, and methods. For each meta-analysis, we will visually inspect forest plots to see if the confidence intervals of individual studies have poor overlap, conduct a Chi
Due to the likelihood of variability in participants and co-interventions across different sites, there is a chance that this review will include studies that are clinically heterogeneous. If studies are determined to be too clinically heterogeneous, we will not conduct a primary meta-analysis but will discuss results narratively including detailed descriptions of the interventions of all included studies.
Assessment of reporting biases
For each meta-analysis that includes 10 or more studies, we will draw a funnel plot and look for asymmetry to assess the possibility of small study or reporting bias (see Sensitivity analysis).
The primary meta-analysis will include all centre-based day care programmes versus any non centre-based childcare (for example, home care by a parent). Subgroup analysis will be conducted as detailed below in Subgroup analysis and investigation of heterogeneity.
We will use Review Manager (RevMan) Version 5.1 (Review Manager 2011) to conduct all meta-analyses. All meta-analyses will be conducted using the random-effects model. Relative risks or rate ratios and 95% confidence intervals will be calculated for dichotomous outcomes and combined using Mantel-Haenszel methods. When risk ratios or rate ratios are not reported and cannot be calculated (in case the exact sample size is unknown) for all studies included in a meta-analysis, we will calculate and report odds ratios. Studies for which effect sizes are expressed as RR and OR will not be meta-analysed jointly. If studies report dichotomous data in multiple formats that cannot be combined in RevMan, we will use Comprehensive Meta-Analysis Version 2 software (Borenstein 2005) to calculate log risk ratios and standard errors for the data, and enter these log risk ratios and standard errors into RevMan.
Subgroup analysis and investigation of heterogeneity
We will conduct the following subgroup analyses.
- Age: < 3 years versus 3 to 5 years.
- Setting: urban versus rural (as identified by authors).
- Co-intervention: centre-based day care alone versus no-intervention; centre-based day care with a co-intervention (for example, nutritional intervention) versus the same co-intervention without centre-based day care.
- Household income: high-income versus low-income households.
Sensitivity analysis will be used as follows.
- We will repeat the primary meta-analysis excluding studies that reported outcomes for completers only.
- For cluster-RCTs, if the study authors do not provide an ICC for relevant outcomes, then we will conduct sensitivity analyses assuming high and low effects of clustering to determine if the results are robust.
All authors received internal support from the Centre for Evidence Based Intervention, Department of Social Policy and Intervention, University of Oxford (UK). EMW also received support from the Centre for Outcomes Research and Effectiveness, Research Department of Clinical, Educational & Health Psychology, University College London (UK).
This protocol was produced within the Cochrane Developmental, Psychosocial and Learning Problems Group.
Contributions of authors
All authors contributed to drafting the protocol.
Declarations of interest
Taylor Brown - none known
Evan Mayo-Wilson - none known
Felix Van Urk - none known
Rebecca Waller - none known
Sources of support
- Centre for Outcomes Research and Effectiveness, Research Department of Clinical, Educational & Health Psychology, University College London, UK.Salary
- Centre for Evidence Based Intervention, Department of Social Policy and Intervention, University of Oxford, UK.Salary, Graduate Stipends
- No sources of support supplied
The authors are publishing another related protocol: Van Urk FC, Brown TW, Waller R, Mayo-Wilson E. Centre-based day care for children under five in low- and middle-income countries. Cochrane Database of Systematic Reviews, Issue 5, 2013.