The effect of special educational assistance in early childhood education and care on psycho-social difficulties in elementary school children

Between three and seven percent of pre-schoolers have developmental problems or child psychiatric disorders [1, 2], which are an important risk factor for mental disorders in adulthood [3]. Efforts to promote healthy growth and development in children who struggle in the early years can accordingly improve children’s long-term life opportunities [4]. Indeed, a recent review reported overwhelmingly positive effects of non-cognitive skills on academic, psycho-social, cognitive and health outcomes [5], though effect sizes are typically not large. It has been hypothesized that the effect of interventions decreases as children grow older and therefore, investing resources later, at the age of school entry or beyond, may show less of an effect [6, but also see 7].

Interventions in early childhood are often described as an effective method to improve the long-term outcomes of children from disadvantaged backgrounds [8], or those with specific developmental or behavioural problems like attention deficit hyperactivity disorder, autism, or behaviour or language problems [9]. Interventions in early childhood education and care (ECEC) can be especially effective because, in contrast to parental training programs, their implementation relies less on parents’ abilities or motivation, and on average 93% of three to five year old children in Organisation for Economic Co-operation and Development (OECD) countries are enrolled in ECEC [10, more than 95% or 5 year old children in Norway are in ECEC]. Randomized controlled trials (RCTs) reported clear effects of early interventions in ECEC for a horizon of up to 9 months, for instance for language problems [11], children with ADHD or autism [12], and for teacher classroom management programs [13].

However, the effect sizes of such interventions are not generally large, and less is known about their effect when interventions are provided outside the well-structured context of RCTs. Even though RCTs are, due to their interval validity, the gold standard for estimating treatment effects, differences between study sample and target population and differences in treatment-implementation between study and regular care contexts make a generalization of findings from RCT samples to populations of interest difficult [14‐17]. Since RCTs often take place in a controlled setting, it may be difficult to replicate the results in other, less rigid settings. For instance, field professionals in ECEC institutions will draw on a much wider range of sources than formal experimental evidence in order to inform their actions. Thus, while evidence from RCTs is encouraging, it remains unclear how it generalizes to interventions in ECEC provided in regular care.

Only a handful of studies examined the effects of special educational assistance (SEA) interventions in ECEC when they are implemented outside of RCTs. These studies used propensity scores to deal with the problematic internal validity in observational studies—due to treatment by indication—and found that children who received SEA in ECEC showed the same or worse outcomes compared children who did not receive SEA [18, 19].

The Norwegian ECEC-system facilitates the investigation of SEA, because children who cannot fully benefit from standard education and care have the right to receive free SEA. Similar to other OECD countries [2], around 4.5% of pre-schoolers in Norwegian ECEC have impaired functioning. The most common impairment being language and communication difficulties, followed by psycho-social difficulties [20], i.e., the inability to partake in daily activities in a manner that is beneficial to oneself and others due to impaired social or psychological functions. Around 2.6% of pre-schoolers in Norwegian ECEC receive SEA, which is provided for several hours per week and targeted at individual children. Children with language problems typically receive one to three hours SEA a week, and children with combined or more severe developmental problems typically receive more hours SEA (see Additional file 1: Fig. S4). After stimulation of language development, social- and behaviour-training and training of independence are the most frequent types of SEA provided.

In Norwegian ECEC, SEA is provided by individuals with varying qualifications, including personnel with a special education degree, kindergarten teachers, or assistants without specialized training [20]. Most parents (75%) of children with SEA report that ECEC institutions implement SEA with training/pedagogy and social inclusion as equally important goals. Consistently, SEA is typically provided in the context of joint activities of all children and less frequently in one-on-one sessions where child and teacher are isolated from the other children. 80% of parents report that there exists an individualized learning plan for children with SEA, though only 60% report that SEA is implemented according to the plan [20]. To date, no study has–to the best of our knowledge–examined the effect of SEA in ECEC on children’s psycho-social difficulties. Related studies on SEA in Norwegian schools reported that students who received SEA have similar or slightly worse scholastic outcomes compared to those who did not receive it [21, 22, see also 23].

In sum, the few studies examining effects of SEA in ECEC outside the context of RCTs reported small negative, to no effects of SEA. Moreover, most studies focused on educational outcomes, such that the effect of SEA on the development of psycho-social difficulties remains largely unclear. Hence, this large-scale prospective cohort study adds to the existing literature by investigating how SEA in ECEC provided outside RCTs affects the psycho-social development of children with developmental or behavioural problems.

The study sample includes 2499 participants (c.f., Fig. 1). Thirty-three percent of the children in the sample received SEA. Table 1 describes the study sample. Additional file 1: Figs. S4 and S5 show that children with more severe problems (e.g. ASD) were more likely to receive SEA and also received SEA from better educated personnel.

Inverse probability weights reduced the differences in mean values for covariates between participants followed up and those lost to follow up to less than 0.1 SMD (c.f. Additional file 1: Fig. S1; [47]). Cumulative distribution plots showed that weighting balanced the entire distributions of covariates (Additional file 1: Figs. S7 and S8).

This research used observational data from a longitudinal population based prospective cohort study to investigate the effect of special educational assistance (SEA) in ECEC on psycho-social difficulties of children with developmental or behaviour problems. We found that, after adjustment for treatment indicators, mothers of children who received SEA in kindergarten reported fewer psycho-social difficulties three years later, compared to mothers whose children did not receive SEA.

While there was some variation in the extent of the positive effect of SEA between groups and different psycho-social difficulties, these differences were not reliably different from zero (c.f. Additional file 1: Figs. S14 and S15). Because the credible intervals for these differences are large compared to the magnitude of the estimated overall effect and the random effects standard deviations are clearly non-zero (S4), these results do not exclude the possibility of group differences. Instead, they might reflect difficulties in reliably measuring exposure, covariates, and outcomes based on parent reports only. Still, the available data were sufficient to reveal an overall positive effect of SEA.

While the positive effect reported in this study is consistent with the results of randomized controlled trials [12, 13] and with reports of the positive effects of preschool child care quality [48], it also stands in contrast to previous observational studies, which estimated no or a small negative “effects” of special education. This apparent contradiction can be due to a number of differences between the current and previous studies. We had estimates of pre-treatment difficulties, and could estimate effects of special education on the change of psycho-social difficulties. Moreover, we used adjustment for treatment predictors instead of propensity score weighting. Adjustment is the preferable approach if treatment-predictors are not colliders on a backdoor path from the outcome to the treatment and if the sample size is large enough to allow for inclusion of many adjustment variables. Another important difference is that whereas previous studies focused on scholastic outcomes, we focused on the effect on psycho-social difficulties. This is a to date little examined but important outcome of SEA, because early psycho-social difficulties are associated with impaired functioning in adulthood [3]. Interestingly, the clear results of SEA on externalizing behaviour suggests that, in addition to helping children with DBP, it can also benefit their families by reducing disruptive behaviour.

The estimated effect size for the reduction of psycho-social difficulties is with on average 0.10 standardized mean difference small. In comparison, previous meta analysis about school– or ECEC–based interventions found effect sizes of between -0.3 and 1.3 SMD for children with or at risk for ADHD [49, 50] or SMDs between 0.3 and 1.1 for children with autism [51]. Randomized trials of classroom management training for kindergarten teachers showed effect sizes similar to our results [Cohen’s d around 0.3 for high risk children at the nine-months follow up, 52]. A recent meta-analysis of reported effect sizes around 0.2–0.3 SMD from experimental manipulations of non-cognitive skill on psycho-social outcomes [5], and smaller effects around 0.1 SMD from non-experimental longitudinal studies. It is possible that the small effect sizes we estimated are, in addition to above mentioned measurement problems, due to the fact the SEA was often provided by personnel with limited training, especially for children with typically less severe problems (c.f. Additional file 1: Fig. S5).

More generally, the decentralized organization of the Educational and Psychological Counselling Service is likely to lead to a large variation in the implementation of SEA [53]. MoBa did not collect more detailed data about SEA, which could help to elucidate when it is most effective. Another possible explanation is that the composition of the study sample, which over-represents well-educated families compared to the population [39], leads to an underestimation of the true effect size, because well-educated parents could reduce children’s psycho-social difficulties even without SEA [54].

While the current study showed that mothers reported fewer psycho-social difficulties in elementary school when their children received SEA in ECEC, a causal interpretation of this result as reflecting an effect of SEA rests on a number of assumptions encoded in Fig. 2. One un-testable assumption is that there are no unmeasured confounders that predict both which children receive SEA and their developmental pathway. Even though the reported analysis includes obvious confounders, other unobserved confounders like e.g. parental engagement could still account for some of the positive association of SEA and psycho-social development. However, because RCTs of SEA and similar interventions typically report positive effects, and thus confirm a causal role of SEA, it appears unlikely that the effects estimated in this study are primarily due to confounding.

The current study has a number of limitations that should be addressed in future studies. Outcomes should be assessed through blinded raters or objective instruments and the quality and quantity of the treatments need to be assessed in greater detail. Moreover, it is important that study samples include participants with a higher a priori prevalence of mental health problems (i.e. no over-representation of highly educated parents which characterizes MoBa) and that care is taken to avoid loss to follow up. Better measurements and more representative samples will be useful to investigate reasons for the relatively small effects observed in the current study, and to identify criteria for effective interventions in ECEC.

Previous RCTs about special educational assistance and teacher management programs showed that interventions in ECEC have a positive immediate impact for children with developmental or behavioural problems, but provide little guidance on long-term effects. The current study has, due to its observational character, a lower internal validity than RCTs, but complements them in terms of external validity and by examining long-term effects. It thus strengthens the view that interventions in ECEC are a useful approach to support pre-schoolers with developmental or behavioural problems.

In sum, the current study suggests that the psycho-social development of children with developmental or behaviour problems may be modified in a positive way through interventions in ECEC, also when provided outside the structured context of randomized controlled trials. Future research with better measurements and more representative samples should investigate under which conditions such interventions are most effective.