Article Text
PDF
Statistics from Altmetric.com
Introduction
Background
Early childcare has a positive impact on children’s academic readiness, as well as cognitive, language and preacademic skills.1 2 However, its impact on behaviour is debated.3 4
While some studies reported that childcare prior to school entry was associated with elevated levels of behavioural difficulties,5 6 particularly externalising problems,5 others found no such effects.7 To the contrary, other researchers found no negative effect of childcare on children’s behaviour,8 or only in the very short term.4 Finally, some studies suggested that the protective effects of early childcare are strongest among children growing up in high-risk families.9
Existing discrepancies in prior studies may result from several factors: varying childcare policies, childcare quality, age(s) at which children’s behaviour was ascertained and the use of parental versus teacher reports of children’s behaviour. Studies tend to compare either centre-based childcare with informal care, that is, childcare by non-professional caregivers such as grandparents, relatives, friends or neighbours, or formal care, that is, childcare by a professional, with informal care. Few studies were conducted in countries with universal childcare policy. Policies concerning early childcare differ from country to country, and therefore not only may the age at which children enter preschool differ but also the time spent in childcare, its quality and availability. Furthermore, with notable exceptions,6 prior studies only focused on short-term behavioural, and sometimes emotional, difficulties, and little is known about the potential long-term consequences of early-life childcare.
France provides an interesting setting to examine this question: (1) the number of places in formal childcare for children under 3 years of age is high (approximately 52% of children can attend compared with 33% on average in Europe)10; (2) childcare is of high quality (Unicef Report Card)11; (3) nearly all children enter the formal school system at 3 years of age (97% start by age 3 and 99% by age 4),12 making it possible to isolate the role of childcare prior to age 3; and (4) early childcare is based on the principle of universality, that is, all children irrespective of their background can access.
French families tend to use two different types of formal childcare: centre-based care (approximately 27% of children) and childminders, professional caregivers with a state degree, authorised to take care of two to six children in their own home (approximately 49% of children).10 13 Childcare is subsidised, and the two types have similar cost but vary in terms of educational content, with centre-based care being generally more structured and standardised than care provided by a childminder. The number of available places is under the supervision of municipalities and varies across place and time.13 Centre-based childcare, highly regarded by families, is in high demand. However, the number of places is limited and the hours are not always flexible, leading some families to turn to childminders.14
Taking advantage of this unique, quasi-experimental, situation, where families choose whether or not their child attends childcare but where the type of childcare is determined by availability, we test the associations between childcare type and children’s trajectories of behavioural and emotional symptoms between ages 3 and 8, using data from the mother–child EDEN (Etude des Déterminants du développement et de la santé de l’ENfant) cohort study. Contrary to other studies where childcare was defined as care received up to the age of 5–7 years, we aimed to study whether childcare in the first 3 years of life is linked to behavioural and emotional levels between ages 3 and 8. We further investigated whether time spent in childcare (<1 year vs ≥1 year) plays a role.
Finally, based on studies showing that childcare use and influence vary depending on maternal education and psychological status, as well as the child’s gender,15–17 we test whether the effects on behaviour vary across these three characteristics.
Methods
Study design, setting and participants
The cohort included 2002 women recruited at 24 weeks of amenorrhoea in two maternity wards in two cities in France: Nancy and Poitiers (2003–2006). All women (n=3758) who visited the prenatal clinic of the local university hospitals were invited to take part in the study. Exclusion criteria included multiple pregnancies, known diabetes, French illiteracy and mothers planning to leave the area in the following 3 years.18 There were 1907 women with a live birth who agreed (95 refusals) to take part in a regular follow-up of their child. Participating children underwent four clinical examinations: at birth, 1 year, 3 years and during the 6th year (average of 5 years and 8 months). Self-reported questionnaires were administered during these examinations, as well as in-between (at 4 months, 8 months, 2 years and 8 years of age). Compared with the general population, mothers participating in the EDEN cohort were slightly more educated, from urban areas and from higher income households than non-participants. Children’s characteristics such as preterm births or admission to neonatal care unit were, however, of the same level.18
Childcare use depends on measurable factors, such as the number of childcare places available, and parents’ financial and employment status, and on unmeasurable factors, such as personal preference. We excluded children with no information on childcare and behaviour between ages 3 and 8 years, resulting in a final analytical sample of 1428 children.
Compared with families included in our analysis, those who did not participate were characterised by lower maternal (12.86, SD=2.67 vs 13.92, SD=2.59 years; p<0.0001) and paternal (12.46, SD=2.54 vs 13.28, SD=2.63 years; p<0.0001) educational level, higher maternal (31.6% vs 18.4%; p<0.0001) and paternal (13.6% vs 5.9%; p<0.0001) levels of unemployment, higher family financial difficulties (61.0% vs 50.7%; p<0.0001), and lower maternal social support (64% vs 71.5%; p=0.002); however, there were no significant differences in terms of children’s sex, birth weight, prematurity or parental history of childhood behavioural difficulties.
Variables
Childcare
Data about childcare arrangements in the first 3 years of life were reported by mothers at 4, 8 and 12 months, and at 2 and 3 years of age. Based on this information, we created three mutually exclusive groups:
Childminder (636, 44.5%).
Centre-based childcare (367, 25.7%): day-care centre, day nursery or crèche staffed with professionals.
Informal childcare (425, 29.8%): primarily parents exclusively (71.3%), sometimes complemented by grandparents or other non-professional caregivers (eg, other relatives, friends, neighbours, occasional baby-sitters).
If the mother reported that the child was in centre-based care on at least one assessment, children were included in the ‘centre-based childcare’ group. If not, and if children were looked after by a childminder on at least one assessment, they were included in the ‘childminder’ group. Children who were looked after exclusively by their parents (n=303, 21.2%) or by their parents along with other non-professional caregivers (n=122, 8.5%) had similar characteristics and outcomes and were grouped in ‘informal care’. Transitions from one type of childcare to another (mainly from informal care to either type of formal care) primarily occurred during the first year of life (n=823).
Depending on the type of childcare reported at each time point, we found that children either spent a few months in childcare (<1 year) or several months (≥1 year). To take into account time spent in childcare from birth to age 3, in additional analyses, children were classified into five groups, using 1 year in childcare as cut-off (childminder <1 year; childminder ≥1 year; centre-based childcare <1 year; centre-based childcare ≥1 year; informal childcare).
Childhood behavioural and emotional development: SDQ score
To ascertain children’s emotional and behavioural patterns, we used the French version of the Strengths and Difficulties Questionnaire (SDQ) reported by the children’s mother at ages 3, 5.5 and 8 years.19
The SDQ includes 25 items which make up 5 scales of 5 items each: one positive (prosocial behaviours) and four negative (emotional symptoms, conduct problems, symptoms of hyperactivity/inattention and peer relationship problems). Each scale ranges from 0 to 10 points, with higher scores representing more problematic behaviours/symptoms except in the case of prosocial behaviours.20
Covariates
The characteristics potentially associated with childcare type or children’s psychological development and ascertained prior to age 3 years were studied as covariates.
These include the following:
Study centre (Nancy or Poitiers): there were no differences in the methods of assessment between the two centres, but Nancy, an urban environment, has a higher density of childcare centres than Poitiers; therefore, the probability of being in centre-based childcare differs between these two settings.
Child characteristics: sex, birth weight (<2500 vs ≥2500 g), premature birth (<37 weeks vs ≥37 weeks of amenorrhoea), birth order (first-born vs not first-born), maternal age at delivery, duration of breast feeding (in months) and age at preschool entry.
Parents’ sociodemographic characteristics measured longitudinally between pregnancy and age 3: marital status (single vs married/cohabiting), partner support (the mother reported whether she felt her partner was supportive, at inclusion: yes vs no), parental educational level (lower education vs higher education), parental employment status (unemployed vs employed), family financial difficulties (to pay for medical care, food or electricity at least once during the first 3 years of life, yes vs no), family income (<€ 1500 vs ≥€1500/month) and parents’ activities with the child (eg, playing, reading stories and so on between 4 months and 2 years: almost daily, often or rarely).21 22
Parents’ mental health: maternal depression during pregnancy, identified using the 20-item version of the Center for Epidemiological Studies Depression Scale (Cronbach’s alpha=0.88) and dichotomised using a cut-off of 2323; maternal postnatal depression ascertained using the Edinburgh Postnatal Depression Scale (Cronbach’s alpha=0.85) at 4, 8 and 12 months post partum and dichotomised using a threshold of 1324; maternal treatment of mental health difficulties (eg, use of tranquillisers, antidepressants, sleeping aids, counselling) during the follow-up; maternal history of psychiatric problems prior to pregnancy; parental history of childhood behavioural difficulties; and maternal use of psychoactive substances in pregnancy and afterwards (cigarettes, cannabis, alcohol).
Statistical methods
All analyses were performed with SAS V.9.4.
Children may have fluctuating levels of symptoms at the three times SDQ was reported, that is, some children may have high levels at 3 years but then decrease over time, while for others it may be the opposite or stable. To identify groups of children with similar patterns of SDQ symptoms over time, we used group-based trajectory modelling.25 The best-fitting model for each SDQ subscale was defined based on a maximised Bayesian information criterion and group parsimony.25
On average, 6.4% of data on study covariates were missing (with a maximum of 22% for maternal depression at 12 months). Missing data were handled using the multiple imputation technique with the fully conditional specification method26 (10 imputations). The imputation model included all variables included in our substantive analysis or predictive of missingness.27
To render the three different childcare groups as similar as possible, we used propensity scores, calculated based on all observed variables associated with childcare or behavioural difficulties.28 All variables predictive of children’s behavioural difficulties and/or childcare attendance were used to calculate propensity scores using multinomial logistic regression.
Propensity scores were included in our statistical models via inverse probability weights (IPWs), which are the reciprocal of receiving the predicted exposure.28 Extreme weights were curtailed at the 99th percentile, and the stabilised mean weight was approximately 1. Finally, we verified that confounding factors were more evenly distributed across the three childcare groups after applying the IPWs.28 29
To test whether childcare type predicts children’s trajectories of emotional and behavioural symptoms, we carried out multinomial logistic regression analyses adjusted for IPWs.
Additionally, we also tested whether children’s behaviour varied with the time spent in formal childcare (<1 year vs ≥1 year). Finally, we tested for statistical interactions between childcare type and (1) child sex, (2) maternal educational level and (3) maternal depression, and stratified on these factors.
Results
Descriptive analysis
In our study population (table 1), 48.0% of children were female, and the average age at preschool entry was 3.45 years (SD=0.6). The average SDQ scores were 9.94 (SD=4.8) at 3 years, 8.75 (SD=5.2) at 5.5 years and 8.62 (SD=5.2) at 8 years. Overall, the characteristics of children in centre-based care or cared for by a childminder were similar, while children in informal care had systematically less favourable profiles.
Characteristics of the study population (n=1428)
Childcare type predicts children’s emotional and behavioural trajectories
Figure 1 shows children’s trajectories of emotional and behavioural symptoms between the ages of 3 and 8 years. For each SDQ subscale, the best model predicts three distinct trajectories representing three levels of symptoms (low, intermediate and high level symptoms). Children belonging to the ‘low level symptoms’ group served as the reference category, except for prosocial behaviours where the ‘high level symptoms’ group was the reference.
Children’s emotional and behavioural symptoms measured by the Strengths and Difficulty Questionnaires from ages 3 to 8 years (EDEN cohort study, n=1428, 2003–2011, France).
The majority of children had low-level or intermediate-level symptom trajectories (except for prosocial behaviours). Overall 15.5% had persistently high levels of conduct problems, 15.1% high symptoms of hyperactivity/inattention, 16.0% of emotional symptoms, 6.8% of peer relationship problems and 13.1% had persistently low levels of prosocial behaviours.
In bivariate analyses (table 2), compared with children in informal childcare, those attending centre-based childcare were less likely to have high levels of emotional symptoms, peer relationship problems, hyperactivity/inattention and conduct problems, while those cared for by a childminder were less likely to have high levels of peer relationship problems. After controlling for child and family characteristics via IPW (table 2), compared with children in informal childcare, those who attended centre-based childcare were less likely to have high levels of emotional symptoms (ORIPW-adjusted=0.35, 95% CI 0.17 to 0.71) and peer relationship problems (ORIPW-adjusted=0.31, 95% CI 0.15 to 0.67). Children looked after by a childminder were somewhat more likely to follow a high trajectory of conduct problems (ORIPW-adjusted=1.72, 95% CI 1.05 to 2.81).
Early childcare (CM, CBC or IC) from 0 to 3 years and children’s trajectories of behavioural and emotional difficulties (low, intermediate or high) from ages 3 to 8 years: EDEN cohort study (n=1428)
Among children who were in formal childcare, those who attended centre-based childcare had the lowest levels of emotional symptoms and peer relationship problems irrespective of duration (table 3) (respectively, <1 year: high emotional symptoms: ORIPW-adjusted=0.23, 95% CI 0.11 to 0.50, and high peer relationship problems: ORIPW-adjusted=0.14, 95% CI 0.05 to 0.39; ≥1 year: high emotional symptoms: ORIPW-adjusted=0.27, 95% CI 0.12 to 0.58, and high peer relationship problems: ORIPW-adjusted=0.43, 95% CI 0.20 to 0.91). This was also the case for children who spent 1 year or more with a childminder (high emotional symptoms: ORIPW-adjusted=0.38, 95% CI 0.18 to 0.80), and high peer relationship problems: ORIPW-adjusted=0.32, 95% CI 0.13 to 0.76). However, only children who spent at least 1 year in centre-based childcare had a decreased likelihood of having high levels of hyperactivity/inattention (ORIPW-adjusted=0.42, 95% CI 0.22 to 0.82) and low levels of prosocial behaviours (ORIPW-adjusted=0.50, 95% CI 0.28 to 0.90).
Time in early childcare (CM, CBC or IC) from 0 to 3 years and children’s behavioural and emotional difficulties from ages 3 to 8 years: EDEN cohort study (n=1428)
Subgroup analyses
Further analyses (figure 2) showed that girls, children whose mother had higher education and those whose mother was not depressed benefited particularly from formal childcare. Compared with girls who were in informal childcare, those who were in centre-based childcare or with a childminder had a lower likelihood of a high trajectory of emotional symptoms and peer relationship problems. Compared with boys in informal care, those who were in centre-based childcare seemed to have fewer emotional symptoms, but those in a childminder’s care were more likely to have conduct problems. Only children whose mother had higher education were less likely to have high levels of emotional symptoms and peer relationship problems in case of centre-based or childminder’s care, as compared with informal care. Children whose mother experienced depression were less likely to have high levels of peer relationship or emotional problems if they attended centre-based childcare.
Early childcare (0–3 years) and children’s behavioural and emotional difficulties (ages 3–8 years), stratifying by child sex, mother’s education and depression. EDEN cohort study (n=1428), inverse probability weight-adjusted multinomial regression models (95% CI).
Discussion
Main results
Our study, conducted among a community sample of children in France, shows three types of trajectories in children’s behavioural and emotional symptoms between ages 3 and 8: persistently high, intermediate and low levels of symptoms. Formal childcare prior to the age of 3 years predicts low levels of emotional symptoms, peer relationship problems and high levels of prosocial behaviours into middle childhood, particularly if children attend for at least 1 year. Childcare also appears to confer benefits in terms of prosocial behaviours. The effects of centre-based, collective, childcare are stronger than those conferred by care provided by a childminder. Our data suggest that opportunities for young children’s socialisation and stimulation, such as those offered by quality centre-based childcare, can serve to prevent children’s emotional difficulties and help develop their psychosocial skills over the long term.
Limitations and strengths
Prior to interpreting our data, we need to acknowledge several limitations. First, the EDEN study was conducted in two distinct cities in France and is not nationally representative. The study population is overall more educated and has higher household income than families in the general population of France. Nevertheless, our sample is socioeconomically heterogeneous and the distribution of childcare types is comparable with the general population.10 Second, children’s psychological and behavioural symptoms were reported by the mother and may have been influenced by maternal reporting style. Although research has shown that mothers tend to report somewhat higher levels of children’s problems than teachers,30 both types of informants are generally consistent and valid, particularly in early and middle childhood.31 Third, the time per day spent in childcare and childcare quality are not directly measured in our study. However, centre-based childcare is generally of high quality in France.32 Time spent in formal childcare is 37–38 hours per week.10
The following are our study’s main strengths: (1) a ‘quasi-experimental’ setting where children with similar characteristics attend either centre-based childcare or a childminder, depending on availability rather than family choice13; (2) repeated measures of children’s and families’ characteristics; (3) longitudinal assessments of children’s psychological and behavioural characteristics from early to middle childhood33; and (4) use of propensity scores and IPWs to render exposure groups strictly comparable34 to account for selection and confounding factors. This does not entirely rule out the possibility of unmeasured confounders (eg, contextual factors such as the local availability of childcare), but it is unlikely that these would explain the associations observed.
Findings’ interpretation
Ours is one of few studies to examine the association between childcare type and children’s psychological development over several years in a country with a policy of universal access to early childcare, directly comparing two different childcare types.
Our data contrast with studies reporting that non-maternal childcare is associated with children’s elevated levels of behavioural problems.8 35 However, one was based in the USA, where childcare is unsubsidised—that is, very expensive—or aimed at children from disadvantaged families and of relatively low quality (the USA ranks in the bottom half of industrial countries in this area),11 and both measured children’s behaviour at one time point, which can introduce bias.
However, our results are consistent with research from Norway, where childcare access and quality are similar to France.36 Nevertheless, in contrast to Norway, in France the number of places available in centre-based childcare centres does not entirely cover demand, and approximately one-third of children 0–3 years are cared for by a childminder. Childminders are trained and qualified; however, their activity is less regulated than that of centre-based childcare centres and the quality of care they offer is heterogeneous.37 This might explain why in our study childminder care appears less beneficial than centre-based care. Enhancing childcare quality might have a positive impact on social and emotional development from early childhood onwards.38
Low levels of emotional symptoms, peer relationship problems and hyperactivity/inattention observed among children attending centre-based childcare may reflect the benefits of cognitive stimulation through play, praise and reading,39 opportunities for socialisation and acquisition of rules which can contribute to children’s self-esteem,38 as well as quality caregiver–child interactions (ie, positive emotionality, sensitivity and responsiveness, the avoidance of harsh physical punishment).40 The benefits of centre-based childcare in terms of children’s development may result from the combination of these different positive elements.
Interaction between maternal and child characteristics and childcare benefits
In our study, girls appear to reap more benefits from formal childcare than boys, particularly in terms of peer relations, emotional symptoms and prosocial behaviours. This is in line with studies showing that centre-based childcare mainly reduces internalising problems, which are most frequent in girls.41 However, in contrast to other studies, we found no extra benefit associated with childcare for children from high-risk families.15 To the contrary, low-risk children appeared to have especially low levels of psychological difficulties when attending childcare. It may be that the universal curriculum that is proposed to all children does not suffice to compensate the difficulties resulting from family socioeconomic or mental health difficulties.
Conclusion
Access to high-quality childcare in the first years of life may improve children’s emotional and cognitive development, prevent later emotional difficulties and promote prosocial behaviours.42 Future research taking into account the daily time spent in childcare in nationally representative samples of children is needed to confirm these benefits on psychological development and whether they translate to a reduced likelihood of psychiatric disorders later in life.
What is already known on this subject
The scientific literature on the relationship between childcare and behavioural and emotional difficulties shows contrasting results.
More recent studies suggest that if childcare is of high quality, it is associated with lower levels of symptoms, especially in children growing up in socioeconomically disadvantaged families.
What this study adds
We studied this topic in France, where early childcare is subsidised and universally accessible.
Children’s behavioural and emotional symptoms in middle childhood (3–8 years) were studied in relation to childcare during the first 3 years of life (centre-based, childminder, informal care).
Formal childcare in the first 3 years of life predicts a lower likelihood of high levels of behavioural and emotional symptoms in middle childhood, especially in children who were in centre-based childcare; this was particularly true for emotional symptoms and peer relationship problems.
Those who were in centre-based childcare for at least one year benefited the most, as did girls and children from socioeconomically more advantaged backgrounds.
Overall, high-quality centre-based childcare specifically predicts lower levels of emotional difficulties, peer problems and low prosocial behaviours into middle childhood.
Acknowledgments
The EDEN Mother–Child Cohort Study Group includes I Annesi-Maesano, JY Bernard, J Botton, MA Charles, P Dargent-Molina, B de Lauzon-Guillain, P Ducimetière, M de Agostini, B Foliguet, A Forhan, X Fritel, A Germa, V Goua, R Hankard, B Heude, M Kaminski, B Larroque, N Lelong, J Lepeule, G Magnin, L Marchand, C Nabet, F Pierre, R Slama, MJ Saurel-Cubizolles, M Schweitzer and O Thiebaugeorges.
References
Supplementary materials
Press release
Press release
Footnotes
Contributors MM and JEBvdW were responsible for data acquisition. MM, JEBvdW, FE-K, LP, SC and RG contributed to the design of the study. RG carried out the analyses and drafted the work. All authors contributed to critically interpreting the results and revising the draft before the approval of the final version of the manuscript. MM coordinated the progress of the study.
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.
Patient consent Not required.
Ethics approval The EDEN mother–child cohort received approval from the ethics committee (CCPPRB) of Kremlin Bicêtre on 12 December 2002 and from CNIL (Commission Nationale de l’Informatique et des Libertés), the French data privacy institution.
Provenance and peer review Not commissioned; externally peer reviewed.