Delayed school entry and academic performance: a natural experiment
Abstract
Aim
Recent reports suggest that delayed school entry (DSE) may be beneficial for children with developmental delays. However, studies of the effects of DSE are inconclusive. This study investigated the effects of DSE versus age-appropriate school entry (ASE) on children's academic achievement and attention in middle childhood.
Method
In total, 999 children (492 females, 507 males; 472 born preterm) were studied as part of a prospective population-based longitudinal study in Germany. Using a natural experimental design, propensity score matching was applied to create two matched groups who differed only in terms of DSE versus ASE. Teacher ratings of achievement in mathematics, reading, writing, and attention were obtained in Year 1, and standardized tests were administered at 8 years of age.
Results
There was no evidence of a difference in the odds of DSE versus ASE children being rated as above average by teachers in Year 1. In contrast, the standardized mean test scores for DSE children were lower than ASE children's mean scores in all domains (mathematics: B=−0.28 [−0.51 to −0.06)], reading: B=−0.39 [−0.65 to −0.14], writing: B=−0.90 [−1.07 to −0.74], and attention: B=−0.58 [−0.79 to −0.36]).
Interpretation
DSE did not affect teacher-rated academic performance. However, missing 1 year of learning opportunities was associated with poorer average performance in standardized tests at 8 years of age. Future research is needed to determine the long-term effect of DSE on academic achievement.
What this paper adds
- Delayed school entry (DSE) has no effect on Year 1 teacher ratings of academic performance.
- DSE is associated with poorer performance in standardized tests of reading, writing, mathematics, and attention at age 8 years.
This article is commented on by Fastenau on pages 596–597 of this issue.
Abbreviations
-
- ASE
-
- Age-appropriate school entry
-
- DSE
-
- Delayed school entry
-
- PSM
-
- Propensity score matching
A child's initial entry into formal schooling marks an important developmental transition. Compulsory school entry age is determined according to a child's birthdate relative to a country-specific cut-off date which indicates the start of the academic year. Research has shown that, within the same academic year, the youngest children have lower academic achievement compared with the oldest children in class.1-3 In countries such as Germany, paediatricians assess children's school readiness and may recommend that children who fail such tests should enter school a year later. Preterm children can be born up to 4 months before their due date, and may have to enter school less mature than their peers.4 Thus, the issue of delayed school entry (DSE) is particularly pertinent for these children and their parents often enquire as to whether they can delay entry into school as they believe it would benefit their child to enter school a year later.
Delaying school entry may be beneficial because teachers may not be able to provide differential support for the less mature children in class,5 and thus they may not receive developmentally appropriate teaching if they start school at the age-appropriate time. Delayed school entry may, therefore, prevent psychological pressure and negative feedback for those who delay school entry because of developmental immaturity. Indeed, positive effects of DSE on primary school achievement have been reported.6, 7 Conversely, DSE may disadvantage children as it denies them the opportunity to receive the early intellectual input they may need to catch up with their peers. Accordingly, DSE has also been shown to negatively affect school performance.8
However, previous studies are characterized by inadequate power and selection bias because, compared with age-appropriate school entry (ASE) children, DSE children more often had special educational needs or behavioural problems and were more often summer-born males and from disadvantaged backgrounds.7, 9 Consequently, it remains unclear if DSE has any positive or negative long-term effects on academic achievement.7, 10 Given that randomized controlled trials are not feasible, applying propensity score matching (PSM) to observational data within a natural setting provides a method for reducing selection bias.11 PSM can be used to select two comparable groups of children who are similar in baseline characteristics but differ in their treatment, i.e. whether they are DSE or ASE. Using PSM, the aim of this study was to investigate the effect of DSE on children's academic performance. The specific research questions were (1) does DSE affect children's mathematics, reading, writing, and attention as rated by teachers at the end of Year 1? (2) Does DSE affect children's performance in standardized mathematics, reading, writing, and attention tests at 8 years of age? (3) Are differences between ASE and DSE children's performance explained by the length of formal schooling received at the time of assessment?
In addition, given the study's particular relevance for preterm children, we repeated analyses on a matched subsample of children born preterm.
Method
Participants
Data were collected as part of the prospective Bavarian Longitudinal Study,12 a whole-population study of children born in 1985 and 1986 within a geographically defined area of Southern Bavaria (Germany) who required admission to a children's hospital within the first 10 days of life (n=7505; 10.6% of all live births). Additionally, 916 healthy term-born control infants were identified at birth from the same hospitals in Bavaria. Of the initial 8421 children, 1316 survivors stratified by sex, socio-economic status, and degree of neonatal risk were assessed at 6 and 8 years of age.12, 13 Of these, 118 children were excluded who were either (1) born at more than 41 weeks’ gestation, (2) entered school 1 year early (in Germany, parents can ask to have their child tested for early school entry), or (3) were enrolled directly in a school for children with special needs (i.e. teacher ratings would not be comparable). In addition, 199 children had incomplete information on baseline characteristics and could not be included in PSM.11 Descriptive characteristics of the 999 children included and 317 excluded in the present study are shown in Table SI (online supporting information).
Procedure
Parents were approached within 48 hours of the infant's hospital admission and were included in the study once they had given written consent for their child to participate. Ethical approval was obtained from the University of Munich Children's Hospital ethics committee. Study follow-up assessments were scheduled at 5, 20, 56 months, and 6 years of age to assess children's pre-school neurological and cognitive abilities. For the present report, two sets of formal assessments were used as dependent variables. (1) At 8 years of age (irrespective of school year), children's mathematics, reading, writing, and attention regulation abilities were assessed using standardized tests by psychologists who were blind to group membership. (2) At the end of school Year 1, when ASE children were aged 6 to 7 years and DSE children aged 7 to 8 years, teacher ratings of mathematics, reading, writing, and attention were obtained.
Measures
Delayed school entry
In 1990 to 1993, Bavarian policy required that all children were assessed by a community paediatrician 3 to 12 months before their chronological-age school entry date to determine their school readiness. All children reaching 6 years of age before 30 June would ordinarily start school the following September. Assessments lasted 30 to 60 minutes and covered key developmental domains. Paediatric recommendation for DSE was based on the results of these assessments. Schools, in conjunction with parents, made the final decision regarding DSE. Parents could not request DSE if their child passed the test.
Baseline characteristics
Twelve variables were included in PSM. These were measured between birth and 56 months of age (before the DSE decision was made) and were selected as potential confounders of the association between DSE and academic achievement (see Table SII, online supporting information, for details).
Teacher ratings of mathematics, reading, writing, and attention at the end of Year 1
Teacher ratings of children's performance in mathematics, reading, writing, and attention were obtained from written reports of primary school teachers’ assessments when children had completed their first year of schooling, and coded into a binary variable (0=average or below average, 1=above average). These reports describe children's achievement during Year 1 in comparison with what is expected from an average child in their age group. Psychologists achieved excellent interrater reliability on these codings (100% agreement in a subsample of 50 reports).
Standardized assessment of mathematics, reading, writing, and attention regulation at 8 years of age
To assess intuitive mathematical competencies, individual children were given a mathematics test.13, 14 Tasks were presented to children in book form with 29 items assessing estimation, reasoning, and visual-spatial problem solving. Item responses were scored for accuracy and summarized into a total score (range 5–24, mean 13.25, SD 3.45). Reading was assessed using the validated Zürich Reading Test15 and a pseudoword reading test (Leon-Villagra J, Wolke D. Pseudoword reading test. Munich: unpublished manuscript, 1993).13 Both scores were highly correlated (r=0.69) and combined into one Reading Test Score (range 1–235, mean 33.49, SD 29.62). Orthographic abilities (writing and spelling) were assessed with a structured diagnostic test (DRT-2; range 0–24, mean 11.19, SD 5.64).16 Children's attention regulation during a standardized test situation was evaluated with the task orientation sub-scale (range 11–60, mean 44.72, SD 5.36) of the Tester's Rating of Child Behaviour.17 Assessments were scheduled to take place when children were 8 years, 5 months; however, DSE children on average had their assessments slightly later than ASE children (mean difference 0.77mo; 95% CI 0.21–1.33) and had completed fewer months of schooling when assessments were administered (mean difference 7.65mo; 95% CI 6.84–8.46). Test scores were z-standardized according to the total sample included in this study (see Table SIII, online supporting information, for more details on dependent variables).
Preschool numeric and phonetic knowledge
Assessments were made to objectively measure children's knowledge before they started elementary school at 6 years of age. Standardized numeric and phonetic tests comprised 10 and 34 items respectively,12, 13 and two total accuracy scores were computed.
Length of schooling
The number of completed months of schooling when standardized tests were administered was calculated.
Statistical analysis
Propensity score matching
Propensity scores were estimated with a logistic regression of DSE when compared with ASE on baseline characteristics. When cases were matched, the difference in baseline characteristics between DSE and ASE children disappeared.11 Thus, PSM resulted in an unbiased estimate of the association of DSE/ASE with academic achievement.18 We used the radius algorithm whereby each DSE child was ‘paired’ with one or more ASE children who had similar propensity scores.18 Children were excluded from the matched sample if no participants shared similar values. Analysis used the matched sample incorporated weights that reflected the ratio of DSE and ASE children sharing similar PSM values.18 The weighting on the matched sample accounts for the fact that an individual DSE child may have been matched to more than one ASE child in the matching process, i.e. matching with replacement. For example, if one DSE child was matched to only one ASE child, this ASE child received a weight of 1. If one DSE child was matched to four ASE children, each ASE child received a weight of one quarter in the analysis. The overall matching balance was measured using pseudo R2, where a value close to 0 indicates the probability that DSE/ASE would be independent of all baseline variables. Standardized bias was used to evaluate the balance in individual variables (>5%=meaningful imbalance). PSM was implemented using the psmatch2 and pstest stata statistical packages (StataCorp, College Station, TX, USA).
Ordinary least squares linear/logistic regressions
Analyses predicting academic achievement were conducted using (1) the final PSM sample, (2) the whole sample, and (3) a subgroup of the PSM sample comprising only preterm children. The effect of DSE on academic achievement was estimated with logistic (above average Year 1 teacher ratings) and linear (standardized test scores at age 8y) regressions. The effects of preschool knowledge were then taken into account in adjusted models. Bootstrapping (1000 replications) and matching weights were used to account for changes in the sample distribution introduced by PSM.18
It was not possible to control for length of schooling in these regression models as it reflects the decision for DSE/ASE. Children who delayed school entry had completed, on average, 8 fewer months of schooling when standardized assessments were administered. We calculated the effect of months of schooling on each standardized assessment within the ASE group only and used this predictive function to predict the performance of DSE children if they had received the same amount of schooling (i.e. an additional 8mo) as ASE children at the time of assessment (proxy estimates).
Results
Propensity score matching
The matched total PSM sample included 959 children (DSE [n=99], ASE [n=860]) with similar propensity scores. Table 1 shows baseline characteristics for the unmatched and matched PSM samples. Before matching, there were moderate to large differences on all 12 baseline covariates between DSE and ASE children. After matching, none of these variables remained statistically different, and bias for all 12 covariates was less than 5% in the matched sample, indicating that a good balance was obtained for all variables. A good overall matching quality was also indicated by Pseudo R2=0.12.
| DSE (unmatched=104; matched=99) | ASE (unmatched=895; matched=860) | Standardized bias (%) | |
|---|---|---|---|
| Child sex, n (%) female | |||
| Unmatched | 40 (38.5) | 452 (50.5) | −24.3 |
| Matched | 38 (38.3) | 315 (36.6) | 3.6 |
| SES at birth, n (%) | |||
| Low | |||
| Unmatched | 30 (28.8) | 295 (33.0) | –a |
| Matched | 28 (28.2) | 267 (31.1) | –a |
| Middle | |||
| Unmatched | 48 (46.2) | 319 (35.6) | 21.7 |
| Matched | 46 (46.5) | 382 (44.5) | 4.0 |
| High | |||
| Unmatched | 26 (25.0) | 281 (31.4) | −14.2 |
| Matched | 25 (25.3) | 210 (24.4) | 1.9 |
| Neonatal index, n (%)27 | |||
| 1. Quartile | |||
| Unmatched | 26 (25.0) | 294 (32.9) | –a |
| Matched | 26 (26.3) | 221 (25.7) | –a |
| 2. Quartile | |||
| Unmatched | 21 (20.2) | 259 (28.9) | −20.4 |
| Matched | 21 (21.2) | 187 (21.7) | −1.1 |
| 3. Quartile | |||
| Unmatched | 14 (13.5) | 179 (20.0) | −17.5 |
| Matched | 13 (13.1) | 120 (13.9) | −2.0 |
| 4. Quartile | |||
| Unmatched | 42 (41.3) | 163 (18.2) | 52.1 |
| Matched | 39 (39.4) | 333 (38.7) | 1.6 |
| Birthweight, g | |||
| Unmatched | 2158.2 | 2738.5 | −62.4 |
| Matched | 2182.1 | 2204.1 | −2.4 |
| SGA, n (%) | |||
| Unmatched | 35 (33.7) | 212 (23.7) | 22.1 |
| Matched | 34 (34.3) | 288 (33.5) | 1.8 |
| GA, wk | |||
| Unmatched | 34.4 | 37.0 | −62.8 |
| Matched | 34.6 | 34.6 | −1.5 |
| Parent–infant relationship problems, n (%) | |||
| Unmatched | 55 (52.9) | 284 (31.7) | 43.7 |
| Matched | 50 (50.5) | 442 (51.4) | −1.9 |
| IQ, 20mo | |||
| Unmatched | 98.3 | 105.3 | −68.5 |
| Matched | 99.7 | 99.8 | −0.8 |
| AWST-language test, 56mo | |||
| Unmatched | 89.0 | 98.6 | −58.1 |
| Matched | 89.9 | 89.8 | 0.6 |
| Attention span, 56mo | |||
| Unmatched | 2.7 | 3.1 | −58.4 |
| Matched | 2.7 | 2.7 | 2.5 |
| Visual–Motor Integration, 56mo | |||
| Unmatched | 5.8 | 6.9 | −47.5 |
| Matched | 5.9 | 5.9 | 0.4 |
| Behaviour regulation, 56mo, n (%)27 | |||
| 1. Quartile | |||
| Unmatched | 44 (42.4) | 251 (28.0) | –a |
| Matched | 41 (41.4) | 373 (43.4) | –a |
| 2. Quartile | |||
| Unmatched | 20 (19.2) | 169 (18.9) | 0.9 |
| Matched | 19 (19.2) | 157 (18.3) | 2.4 |
| 3. Quartile | |||
| Unmatched | 30 (28.8) | 325 (36.3) | −15.9 |
| Matched | 29 (29.3) | 243 (28.3) | 2.1 |
| 4. Quartile | |||
| Unmatched | 10 (9.6) | 150 (16.8) | −21.2 |
| Matched | 10 (10.1) | 86 (10.0) | 0.9 |
| Outcome variables | |||
| Teacher ratings at the end of Year 1 (% above average) | |||
| Mathematics | 42 (62.7) | 487 (58.9) | |
| Reading | 42 (61.8) | 483 (58.4) | |
| Writing | 34 (60.7) | 434 (62.5) | |
| Attention | 33 (58.9) | 404 (60.6) | |
| Standardized mean test scores at age 8y (SD) | |||
| Mathematics | −0.6 (1.2) | 0.1 (1.0) | |
| Reading | −0.6 (1.4) | 0.1 (0.8) | |
| Writing | −1.1 (0.8) | 0.1 (0.9) | |
| Attention | −0.8 (1.2) | 0.1 (0.9) | |
- a STATA default-standardized bias was not computed on the reference group of nominal variables. Numbers are reported as means if not stated otherwise. ASE, age-appropriate school entry; AWST, Aktiver Wortschatz Test (German standard vocabulary test); DSE, delayed school entry; GA, gestational age at birth (wk); SES, socio-economic status; SGA, small for gestational age.
The matched PSM subsample included 472 preterm children (see Table SIV, online supporting information).
Does DSE affect teacher ratings of children's mathematics, reading, writing, and attention at the end of Year 1?
Table 2 shows that there was no evidence of a difference in the odds of DSE compared with ASE children being rated as performing above average in mathematics, reading, writing, and attention by teachers in Year 1. Regression results were robust across the final overall PSM sample, the logistic regression sample, and the PSM subsample of preterm children. Controlling for children's preschool numeric and phonetic knowledge did not change these findings. Better preschool numeric knowledge was consistently associated with higher odds of being rated as performing above average in mathematics, reading, and writing, when compared with lower preschool numeric knowledge.
| Unadjusted PSM models (n=959) | Fully adjusted PSM models (n=959) | Logistic regression models (n=1198)a | Fully adjusted PSM models only including preterm children (n=343)a,b | |
|---|---|---|---|---|
| Mathematics | ||||
| Delayed entryc | 1.59 (0.92–2.73) | 1.61 (0.90–2.86) | 1.49 (0.77–2.89) | 1.62 (0.60–4.39) |
| Preschool numeric knowledge | 1.28 (1.19–1.37) | 1.19 (1.11–1.28) | 1.19 (1.07–1.31) | |
| Preschool phonetic knowledge | 0.94 (0.89–1.00) | 0.92 (0.85–1.00) | 0.89 (0.80–1.00) | |
| Reading | ||||
| Delayed entryc | 1.59 (0.97–2.62) | 1.48 (0.88–2.51) | 1.34 (0.73–2.47) | 1.40 (0.53–3.68) |
| Preschool numeric knowledge | 1.24 (1.16–1.31) | 1.14 (1.06–1.21) | 1.17 (1.06–1.29) | |
| Preschool phonetic knowledge | 1.06 (1.01–1.12) | 0.99 (0.93–1.05) | 1.00 (0.90–1.10) | |
| Writing | ||||
| Delayed entryc | 1.32 (0.75–2.34) | 1.30 (0.73–2.33) | 1.12 (0.58–2.16) | 2.28 (0.69–7.51) |
| Preschool numeric knowledge | 1.18 (1.11–1.26) | 1.08 (1.00–1.16) | 1.15 (1.03–1.27) | |
| Preschool phonetic knowledge | 0.98 (0.92–1.04) | 1.00 (0.93–1.06) | 0.92 (0.82–1.03) | |
| Attention | ||||
| Delayed entryc | 1.20 (0.66–2.16) | 1.04 (0.57–1.88) | 1.06 (0.56–2.01) | 0.48 (0.16–1.43) |
| Preschool numeric knowledge | 1.10 (1.03–1.17) | 1.01 (0.95–1.07) | 1.00 (0.90–1.10) | |
| Preschool phonetic knowledge | 1.06 (1.00–1.13) | 1.01 (0.95–1.08) | 0.96 (0.86–1.07) | |
- aAdjusted for all baseline characteristics used in selecting PSM sample. bAdditionally adjusted for child sex. cOdds ratios (OR) indicate the predicted likelihood of DSE compared with ASE children being rated as performing above average. All estimates were run with 1000 bootstrap replications. DSE, delayed school entry; PSM, propensity score matching.
Does DSE affect children's performance in standardized mathematics, reading, writing, and attention tests at 8 years of age?
In contrast to the outcomes for Year 1, there was evidence that DSE predicted significantly lower mean scores in all four standardized tests (mathematics, reading, writing, and attention regulation) and across all PSM and ordinary least squares models at age 8 years of age (Table 3, also see Fig. S1, online supporting information) with one exception: preterm children's intuitive mathematics mean scores were not affected by DSE after controlling for children's preschool knowledge and sex. Preschool numeric knowledge consistently predicted better average performance across all four domains.
| Unadjusted PSM models (n=959) | Fully adjusted PSM models (n=959) | OLS models (n=1198)a | Fully adjusted PSM models only including preterm children (n=343)b | |
|---|---|---|---|---|
| Mathematics | ||||
| Delayed entryc | −0.31 (−0.55 to −0.07) | −0.28 (−0.51 to −0.06) | −0.27 (−0.48 to −0.07) | −0.30 (−0.65 to 0.04) |
| Preschool numeric knowledge | 0.12 (0.10 to 0.14) | 0.04 (0.02 to 0.06) | 0.14 (0.10 to 0.17) | |
| Preschool phonetic knowledge | −0.02 (−0.04 to 0.00) | 0.00 (−0.02 to 0.01) | −0.04 (−0.07 to 0) | |
| Reading | ||||
| Delayed entryc | −0.48 (−0.76 to −0.19) | −0.39 (−0.65 to −0.14) | −0.42 (−0.67 to −0.18) | −0.62 (−1.01 to −0.24) |
| Preschool numeric knowledge | 0.13 (0.10 to 0.16) | 0.06 (0.03 to 0.09) | 0.12 (0.07 to 0.17) | |
| Preschool phonetic knowledge | 0.03 (0.01 to 0.05) | 0.01 (0.00 to 0.03) | 0.02 (−0.01 to 0.06) | |
| Writing | ||||
| Delayed entryc | −0.93 (−1.11 to −0.74) | −0.90 (−1.07 to −0.74) | −0.94 (−1.10 to −0.77) | −0.98 (−1.17 to −0.78) |
| Preschool numeric knowledge | 0.12 (0.09 to 0.14) | 0.08 (0.06 to 0.1) | 0.11 (0.07 to 0.14) | |
| Preschool phonetic knowledge | 0.03 (0.00 to 0.05) | 0.01 (−0.01 to 0.04) | 0.02 (−0.01 to 0.05) | |
| Attention | ||||
| Delayed entryc | −0.64 (−0.89 to −0.38) | −0.58 (−0.79 to 0.36) | −0.57 (−0.77 to −0.36) | −0.74 (−1.05 to −0.43) |
| Preschool numeric knowledge | 0.11 (0.09 to 0.14) | 0.03 (0.01 to 0.06) | 0.09 (0.05 to 0.12) | |
| Preschool phonetic knowledge | 0.01 (−0.01 to 0.03) | 0.00 (−0.02 to 0.02) | 0.00 (−0.03 to 0.03) | |
- aAdjusted for all baseline characteristics used in selecting PSM sample. bAdditionally adjusted for child sex. cCoefficients indicate DSE children's mean scores compared with ASE children's mean scores (baseline). The B coefficients express the effect size. All regression coefficients B where the 95% CI does not include 0 are statistically significant at the p<0.05 level. All estimates were run with 1000 bootstrap replications. DSE, delayed school entry; OLS, ordinary least squares regression; PSM, propensity score matching.
Are differences in performance between children with age-appropriate and delayed school entry explained by the length of schooling?
We calculated the effect of 8 additional months of schooling within the ASE group and used this predictive function to test if DSE children would have reached a similar level of performance if they had received the same amount of schooling as ASE children at the time of assessment. Figure 1 shows DSE children's observed standardized test results at 8 years of age (dots) and predicted mean test results (lines) if DSE children had received, on average, 8 additional months of schooling. Although not shown in the figure, predictions were based on the observed effect of 8 months of schooling on ASE children's scores. These results suggest that even if DSE children had received the same amount of schooling as ASE children, most would show worse mean performance than ASE children (i.e. dots are below the line). Similar findings were obtained on the PSM subsample of preterm children (Fig. S2, online supporting information).
Discussion
This study found that there was no significant difference in the odds of ASE versus DSE children being rated as performing above average in mathematics, reading, writing, and attention by their teachers at the end of Year 1. In contrast, DSE children's standardized mean test scores of mathematics, reading, writing, and attention at 8 years of age were lower than ASE children's mean scores. To our knowledge, this is the first study to have assessed the effect of DSE in a large sample after minimizing selection bias and accounting for confounding effects of preschool knowledge.
The issue of DSE has received considerable interest.7-9, 19 Parents who wish to delay their child's school entry may do so because their child has developmental problems. If these are because of developmental immaturity, then starting school a year later may give their child additional time to mature.9 However, these problems may often indicate developmental impairments or special educational needs that could be better addressed with early intervention rather than DSE. Our results suggest that DSE has no effect on academic achievement but may delay formal instruction and the provision of special educational support during a key developmental period.
The findings presented here are based on both assessments according to age (standardized tests) and according to grade level (teacher ratings). In primary school, age has a large impact on performance as older children outperform younger children within a class.20 Thus, neither age-level nor grade-level assessments alone provide unbiased outcome measures. Our results indicate that despite DSE children being older, they did not outperform the younger ASE children on teacher ratings. Similarly, DSE did not lead to any performance advantage in standardized tests. DSE children had completed, on average, 8 months of schooling less than ASE children when standardized assessments were administered at age 8 years, and it was not possible to control for length of schooling in regression models as it reflects the decision for DSE/ASE. However, we estimated the effect of 8 months of schooling on mean test scores within the ASE group and used the regression line to estimate the impact of longer schooling within the DSE group. Still, more DSE children would have lower mean test scores than ASE children. Thus, overall, delayed school entry may not provide any advantage for achievement at school.21 However, this analysis needs to be interpreted cautiously.
These findings are particularly applicable to preterm children who are born up to 4 months before their due date and may enter school less mature compared with their peers.4 Preterm children were over-represented in the DSE group, and our analyses on a matched subsample of preterm children confirmed the total sample results. Although there was only one significant DSE by preterm birth interaction on attention regulation but not on academic performance, this analysis is important for a couple of reasons: prematurity is associated with increased need for special educational support22, 23 and specific impairments may only become apparent during primary school because of the larger demands on cognitive and socioemotional abilities.24, 25 Parents of preterm children often ask if they can delay school entry as they believe it would benefit their child. This may arise from an expectation that preterm children will developmentally catch-up with their peers over time. Accordingly, it has recently been suggested that DSE may particularly benefit preterm children.26 However, parents and professionals should be aware that DSE may not promote preterm children's academic performance. Teachers are essential resources in encouraging learning and providing formal instruction. Raising education professionals’ awareness of the potential problems faced by preterm children may aid in ensuring they receive the specific help they need to thrive in school. In some situations, delaying a child's school entry may seem an appealing solution, but our results suggest that DSE children increase heterogeneity in class,8 making it more difficult for teachers to address individual needs. Thus, the decision to delay school entry should be taken with due caution as there may be disadvantages arising from missing a year of learning opportunities or not receiving special educational support during the critical primary years.8, 9
Strengths and limitations
The longitudinal data reported here were collected in a large whole-population sample of children born across the full spectrum of gestational ages. Assessors were blind to DSE versus ASE group membership, but teachers were not and thus their ratings could have been biased. Although this is a general problem of all studies that include teacher ratings, these assessments are highly reliable as teachers based their judgement on knowing the child for a whole school year. The application of PSM provided precise effect estimates of DSE on academic performance while taking into account a comprehensive number of confounding variables. In this way, we carefully matched both groups on variables previously shown to be relevant for predicting school progress and controlled for preschool numeric and phonetic knowledge. Most birth cohort studies (e.g. all British cohort studies) schedule assessments according to age. Thus, both DSE and ASE children were assessed with standardized tests at the same age (8y), rather than according to length of school experience. However we also included an assessment according to grade level using teacher reports at the end of Year 1, the first year of schooling for all children in the study. This assessment is based on the length of academic experience and not age. Thus, we were able to assess the effects of DSE according to both age (standardized tests) and grade level (teacher assessments). We believe that this combination of grade-level and age-level assessments allows us to be confident in our conclusion that DSE did not enhance school achievement compared with ASE. Although there may be additional confounding factors not assessed, there was no indication that DSE could be beneficial for academic progress and regression models confirmed these findings. In Germany, decisions to delay children's school entry are based on community paediatric assessments and not on parents’ requests; thus our data are comparable with international studies on retaining (the school's decision to delay entry) but less with studies on ‘redshirting’ (the parents’ decision to delay school entry).
Conclusion
Delaying school entry has no advantage for the likelihood of positive teacher ratings of academic achievement taken at the same point in Year 1 or on children's mean scores in standardized tests of mathematics, reading, writing, and attention at 8 years of age. Effectively, DSE may mean that children miss out on learning opportunities during the critical early years. Future research is needed to determine the long-term impact of DSE on academic performance and attainment at the end of formal schooling.
Acknowledgements
This study was supported by grant EDU/40442 from the Nuffield Foundation. The Nuffield Foundation was not involved at any stage in the drafting of the manuscript. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the Nuffield Foundation. There are no financial relationships with any organizations that might have an interest in the submitted work in the previous 3 years. The authors have stated that they had no interests that might be perceived as posing a conflict or bias.
