Aim To study the impact of specific neuropsychological measures on academic attainment in very preterm (VPT) children.
Methods VPT children (gestational age <31 weeks, N=48) and matched term controls (N=17) aged 9–10 years were assessed with measures of processing speed, executive function and IQ. Teachers reported on academic achievement in a questionnaire.
Results Group differences in academic attainment were significant for maths (OR 6.5; 95% CI 1.7 to 25.8), English/literacy (OR 3.8; 95% CI 1.1 to 13.5), overall academic attainment (OR 11.9; 95% CI 1.4 to 96.9) and special educational needs provision (OR 7.2; 95% CI 1.5 to 35.0). All significant group differences in attainment could be accounted for by processing speed. Birth group, processing speed and working memory were significant predictors of overall attainment (R2=0.57; p<0.001).
Conclusions Processing speed and working memory are important factors underlying academic attainment in VPT children. Specific tests of processing speed and working memory, which together take approximately only 10 min to administer, could potentially be used as efficient screening instruments to assess which children are at risk of educational problems and should be referred for a full neuropsychological assessment.
Statistics from Altmetric.com
Children born very preterm (VPT) or with very low birth weight (VLBW) are known to be at risk of a wide range of developmental problems, including impairments in the cognitive and motor domain and poor educational outcome. VPT and VLBW children are more frequently in special education than term/normal birth weight children, and more often receive special educational support in mainstream schools.1,–,4 Both psycho-educational assessments and teacher ratings of VPT/VLBW children's performance have established depressed achievement across academic subjects compared with term/normal birth weight controls.1,–,3 5 6 Recent studies have shown that the negative impact of VPT birth on academic attainment persists into adolescence and adulthood.7,–,12
Poor academic achievement has been linked to low IQ in VPT/VLBW children,6 13,–,15 and extremely low birth weight/extremely preterm children.1 16,–,18 However, IQ is a general construct that might mask the more specific nature of difficulties underlying poor scholastic performance,19 and full assessment of IQ takes up considerable time of educational psychologists. A set of specific neuropsychological functions linked to cognitive control, referred to as executive functions, has been shown to be affected by preterm birth.20 A range of executive functions has been linked to academic attainment in term children,21,–,24 but the impact on academic attainment in preterm children is currently unknown. Moreover, it has recently been shown that slow processing speed is a key factor underlying executive function and attention problems in VPT children (H Mulder, NJ Pitchford, N Marlow, in preparation). Slow processing speed may thus underlie poor academic attainment in VPT children, but this has yet to be established.
What is already known on this topic
▶. VPT children are at increased risk of poor academic attainment compared with their term peers.
▶. VPT children are at increased risk of slow processing speed and poor executive function compared with their term peers.
What this study adds
▶. Slow processing speed explains significant group differences in academic attainment between VPT and term children in middle childhood.
▶. Processing speed and working memory might be quick and useful screening measures for the risk of poor academic attainment and areas for intervention.
The present study aimed to investigate: (1) group differences in academic attainment between VPT and term children; (2) whether processing speed can account for group differences in attainment; (3) which, if any, executive function measures are predictive of attainment over and above processing speed; and (4) the relative role of executive function and attention measures compared with a standard IQ assessment as predictors of academic attainment.
The sample comprised VPT children (gestational age <31 weeks) born in either Queen's Medical Centre or City Hospital in Nottingham, UK, between 1997 and 1999. At age 9–10 years, VPT children were invited to take part in the study and asked to invite an age (±3 months) and sex-matched classmate to act as a term control. In total, 135 VPT children were eligible to enter the study, of whom 132 were traced. Four of these had significant impairments, leaving 56/128 who took part in the study (44% positive response rate). All VPT children included in the analyses attended mainstream school. A total of 22 control children took part in the study. The mean age of the VPT and term group was 117 months (SD 4) and 117 months (SD 5), respectively. The mean gestational age of the VPT group was 27.6 weeks (SD 1.8, range 25.0–30.9). There were no significant differences between groups in age, sex, maternal education or socioeconomic status. Ethical approval for the study was granted by the Leicestershire, Northamptonshire and Rutland Research Ethics Committee 1. Written consent was obtained from parents and children before the assessments.
Children underwent a neuropsychological evaluation including an assessment of IQ, executive function, attention and processing speed (see appendix for further details).
A questionnaire was returned by teachers for 86% (48/56) of VPT children and 77% (17/22) of term children. Teachers rated academic attainment on English/literacy, maths, science, technology and design, geography, information technology and history on a five-point scale (1, very below average; 3, average; 5, very above average). A total academic achievement score (TAAS) was computed,25 by averaging across the seven subject areas. We determined the number of children performing below average on the continuous variable TAAS (score <2.5), as well as the number of children scoring below average on the core subjects maths and English/literacy (score <3; reflecting performance scored by teachers as ‘very below average’ or ‘below average’). Teachers also rated whether children received special educational needs (SEN) support in school (ie, one-to-one or small group special needs provision, outreach teacher, educational or clinical psychologist, physiotherapist and/or speech therapist).
Data were double entered and analysed using SPSS 15. To address the four research questions outlined above we conducted the following analyses:
We established whether group differences occurred between VPT and term children.
We investigated which combination of neuropsychological factors was most predictive of attainment. As we considered the measures of processing speed to be less likely to draw on many additional processes than the other more complex measures in our neuropsychological test battery, these factors were studied first in relation to attainment.
We investigated which of the measures of executive function were related to attainment over and above processing speed.
Selected measures of processing speed and executive function were compared with full scale IQ as predictors of attainment.
Two sets of analyses were conducted to answer each of the four research questions. First, to investigate the impact of birth group and neuropsychological variables on attainment, categorical measures of attainment (scoring below average or not on maths, English/literacy, or TAAS and receiving SEN provision or not) were analysed using logistic regression. Receiver operating characteristics curves were used to assess the predictive value for significant associates of attainment. The area under the receiver operating characteristics curve (AUC) was used to discriminate between the various measures as predictors of attainment. Second, to investigate relationships between neuropsychological variables and attainment in more detail, TAAS was analysed as a continuous variable using correlations (Pearson's r) and linear regression.
Are there significant group differences in academic attainment between VPT and term children in the current sample?
Categorical measures of attainment
The frequency distribution of scoring below average versus average/above average on teacher-rated academic attainment in VPT and term children is shown in table 1. VPT children scored below average significantly more often than term children on all subjects and TAAS, and more often received SEN provision (χ2 (1)=5.8; p=0.016).
Continuous measure of attainment
Next, academic attainment was studied in more detail, by investigating TAAS as a continuous variable. There was a significant group difference in TAAS favouring term children (VPT M=2.6; SD 0.7; range 1.0–4.0; term M=3.3; SD 0.6; range 2.4–4.7; ΔM=0.8; 95% CI −1.1 to −0.4; p<0.001).
Can processing speed account for group differences in attainment?
Categorical measures of attainment
Group differences in academic attainment were studied in relation to verbal and motor processing speed for the key variables maths, English/literacy, TAAS and SEN provision. Verbal processing speed accounted for the significant group differences in maths, English/literacy, TAAS and SEN provision (table 2). Motor processing speed accounted for the group difference only in English/literacy. When verbal and motor processing speed were entered in logistic regression simultaneously, only verbal processing speed was a significant predictor of maths, English/literacy, TAAS and SEN provision.
Continuous measure of attainment
Correlations between TAAS and verbal and motor processing speed were studied within and across groups (table 3). When verbal and motor processing speed were simultaneously entered in linear regression predicting TAAS, verbal processing speed was a significant predictor (p<0.001), in contrast to motor processing speed (p=0.706). Across groups, increased reaction time (RT) was related to decreased academic attainment, as shown for verbal processing speed in figure 1. In linear regression, birth group explained 21% of variance in TAAS; verbal processing speed accounted for an additional 27% of variance (p<0.001). When verbal processing speed was controlled for, the group difference between VPT and term children was reduced but remained significant (B −0.47; SE 0.16; β=−29; p=0.005).
Are executive function measures predictive of attainment over and above processing speed?
Categorical measures of attainment
Logistic regression models predicting maths, TAAS and SEN provision by group and verbal processing speed were significantly improved by including working memory (letter–number sequencing) as a predictor, but none of the other executive function measures. Odds ratios for group adjusted for both verbal processing speed and working memory (letter–number sequencing) are shown in table 2. None of the executive function measures added significant explained variance to the regression model predicting English/literacy after the inclusion of verbal processing speed.
Continuous measure of attainment
Correlations between executive function and TAAS as a continuous variable are shown within and across groups in table 3. Significant correlations in the total sample occurred for all subtests except selective attention; improved performance on the executive function measures (more items correct or less response time) was related to better attainment. Selective attention data were significantly skewed; non-parametric correlation showed a significant effect of selective attention on academic attainment, with increased time/target related to decreased academic attainment (VPT rs=−0.38; p=0.009; term rs=−0.47; p=0.057; across rs=−0.51; p<0.001). When controlling for verbal processing speed using partial correlations, none of the executive function skills were significantly related to TAAS, except working memory (letter–number sequencing) and shifting accuracy (table 3). After selective attention data were normalised by logarithmic data transformation and excluding two outliers, the partial correlation with TAAS was still not significant (p=0.275).
To explore the amount of variance in TAAS accounted for by the executive function variables working memory (letter–number sequencing) and shifting accuracy, we studied hierarchical linear regression models in which these factors were added as predictors of TAAS to a model including birth group and verbal processing speed. Shifting accuracy added 3% explained variance compared with the model including only birth group and verbal processing speed, a non-significant effect (p=0.090). Working memory added 8% explained variance, and this effect was significant (p=0.002). The relationship between working memory and TAAS is shown in figure 2.
In summary, both verbal processing speed and working memory (letter–number sequencing subtest) were shown to be important predictors of academic attainment. Next, the relative strength of these predictors compared with a standard assessment of IQ as a predictor of attainment was investigated.
What is the relative role of executive function and attention measures compared with a standard IQ assessment as predictors of academic attainment?
Categorical measures of attainment
The ability of verbal processing speed and working memory to discriminate between children at risk of scoring below average or not on maths, English/literacy, TAAS and SEN provision was acceptable (all AUC >0.7 across groups). In addition, verbal processing speed was a good discriminator for scoring below average on maths (AUC 0.83; 95% CI 0.73 to 0.93), and working memory was a good discriminator for being at risk of scoring below average on TAAS (AUC 0.81; 95% CI 0.70 to 0.92). Similarly to processing speed and working memory, IQ also accounted for all significant group differences in academic attainment (table 2). The ability of IQ to discriminate between children at risk of scoring below average on maths, English/literacy and TAAS, or receiving SEN provision was good (all AUC values between 0.8 and 0.9 across groups).
Continuous measure of attainment
The linear regression model predicting TAAS by birth group (B −0.47; SE 0.15; β=−0.29; p=0.002) and the specific predictors verbal processing speed (B −0.09; SE 0.03; β=−33; p=0.005) and working memory (B 0.09; SE 0.03; β=0.36; p=0.002) explained 57% variance (p<0.001). In comparison, the variables birth group and full scale IQ accounted for 65% variance in TAAS. The group difference was no longer significant when IQ was controlled for (B −0.24; SE 0.14; β=−0.15; p=0.079).
As expected, increased academic problems in VPT children compared with term children were found, with VPT children scoring less well in all academic subjects assessed and the composite measure of total academic achievement, and receiving more SEN provision than term children. As executive function has been linked to academic attainment in term children,21,–,24 and VPT children have difficulties with executive function and slower processing speed,20 26 we set out to explore whether measures of processing speed and executive function could account for group differences in attainment.
Group differences in academic attainment between VPT and term children were accounted for by processing speed. Processing speed has previously been linked to reading and maths development in the normal population.27,–,30 Bull and Johnston28 offered two potential explanations for the impact of processing speed on attainment. First, slow processing speed may arise as the result of slow automaticity of basic facts, such as number knowledge. Second, slow processing speed may be due to a more generic processing speed deficit that could impact on cognitive function generally. In the current study, a general underlying processing speed problem seems a more likely explanation, as we have shown the impact of slow processing speed on a wide range of cognitive functions, such as executive function and attention (H Mulder, NJ Pitchford, N Marlow, in preparation). Also, in support of this hypothesis, Schneider et al6 showed general cognitive processing to be a more important predictor of attainment in VPT/VLBW children than specific skills such as number knowledge.
General reductions in processing speed in VPT children may be due to decreased efficiency and/or connectivity of neural networks. Recent MRI studies have shown differences in white matter microstructure between VLBW and normal birth weight children across multiple brain regions.31 32 These alterations may be due to decreased myelination, associated with reduced efficiency of action potential transfer in the brain. In addition, altered patterns of neural activation in preterm compared to term children have been shown with functional imaging, suggesting qualitatively different connectivity.33 Future studies should focus on the association between structural and functional alterations in VPT children in relation to processing speed to investigate these issues further.
Next, we studied whether executive function was predictive of academic attainment, as has previously been shown in studies with term children,21,–,24 and for working memory in VPT, VLBW and extremely low birth weight children.14 34 35 Selective attention, inhibition, working memory, semantic fluency and shifting were related to academic attainment both within the VPT group and across groups, but most of these associations were explained through shared variance with verbal processing speed. However, working memory was predictive of attainment independently of verbal processing speed. The impact of processing speed and working memory on cognition has previously been shown by Fry and Hale,36 37 who identified a developmental cascade model, in which 80% of age-related change in fluid intelligence was explained by the mediating effects of age on processing speed and working memory in their study of 7–19-year-old term children. Similarly, in a study of preterm and term infants, birth status was shown to impact on cognitive development at 2 and 3 years of age through measures of processing speed and memory at 7 months.38 These studies, in conjunction with the current study, highlight the importance of both processing speed and working memory as factors underlying cognitive development.
In the present study, verbal processing speed was a better predictor of academic attainment than motor processing speed. This discrepancy seems surprising, as age-related changes in processing speed have been shown to be domain general in a study of a wide range of measures of response time across 72 studies.39 Moreover, Bull and Johnston28 found motor processing speed to be a better predictor of children's arithmetical problems than the speed of number and letter identification (ie, verbal processing speed). This discrepancy in findings could potentially be explained by the high frequency of minor motor impairments reported in VPT children, which could confound an accurate reflection of general processing speed. An assessment of fine motor skill, which was not used in the current study, would be necessary to test this hypothesis.
The current study has clear clinical implications. First, the impact of slow processing speed on academic attainment may be decreased by developing appropriate educational interventions, in which information is presented to the child slowly and in sequential, rather than simultaneous fashion. Moreover, we showed that measures of verbal processing speed and working memory might be useful to assess which VPT children are most at risk of developing learning problems at school. We found that a regression model including group, verbal processing speed and working memory explained 57% variance in attainment, whereas a model including group and full scale IQ explained 65% variance in academic attainment, a difference of only 8%. As verbal processing speed and working memory measures take approximately 10 min to administer in total, these measures might be used as a general screener for which VPT children are at risk of developing scholastic problems. Children considered at risk might then be given a full assessment of IQ, which can take up to 1.5 h to administer, to establish the level of problems in more detail and the particular pattern of strengths and weaknesses.
A limitation of the current study was the relatively small control group; frequencies of below average performance on academic subjects may have been susceptible to chance and impacted on the odds ratios found for birth group. However, much larger studies have shown a similar percentage of term children to score below average on TAAS (5%) as in the current study (6%),16 18 thus supporting our findings. Also teacher questionnaires were used to assess academic attainment rather than standardised psycho-educational assessments, which might provide a more accurate reflection of academic attainment.
In conclusion, the impact of verbal processing speed and working memory on academic attainment has been demonstrated in VPT children. As measures of verbal processing speed and working memory are very brief to administer, they may be promising as screening tools for assessing VPT children's risk of developing problems with academic attainment. In future studies, the predictive validity of verbal processing speed and working memory measures for later development needs to be studied longitudinally in VPT children from preschool age.
The authors would like to thank all the children, parents and teachers who have been involved in this study. They are grateful to Sarah Beaven for help with IQ testing.
Neuropsychological test battery
▶. IQ was assessed with the Wechsler Intelligence Scale for Children, 4th UK edition (WISC-IV).40 A scaled score full scale IQ was computed, which has a mean of 100 and SD of 15.
▶. Executive function and attention skills were assessed using subtests from the test of everyday attention for children (TEA-Ch)41 and a neuropsychological battery (NEPSY).42. For all executive function and attention tasks, raw scores were used in the analyses. Here we focus only on the subtests that we have previously shown VPT children to perform significantly less well on than the term control group (H Mulder, NJ Pitchford, N Marlow, in preparation), which are:
▶. Sky search: a selective attention subtest from the TEA-Ch, in which children have to find targets in a large display of distracters; efficiency was scored.
▶. Creature counting: a shifting subtest from the TEA-Ch, in which children have to switch in a flexible manner between counting upwards and downwards; efficiency and accuracy were scored. As reported previously (H Mulder, NJ Pitchford, N Marlow, in preparation), 29% of VPT and 5% of term children did not reach the criterion described in the test manual to be given a creature counting efficiency score, thus the number of cases is reduced for analyses including this measure.
▶. Walk don't walk: a response inhibition subtest from the TEA-Ch, in which children have to respond continuously to an auditory stimulus and all of a sudden withhold that response; the number of commission errors was scored.
▶. Opposite worlds: an interference suppression subtest from the TEA-Ch, in which children have to read out a string of ones and twos and say ‘one’ when they see ‘two’ and the other way round; reaction time was scored.
▶. Semantic fluency: a semantic verbal fluency subtest from the NEPSY, in which children have to think of as many words as possible in a certain category within 1 min; the number of words correct was scored.
▶. Digit span backward: a WISC-IV subtest, in which children are given a string of numbers and have to repeat them in backward order; the number of items correct was scored.
▶. Letter–number sequencing: a subtest from the WISC-IV, in which children are given a string of letters and numbers and have to repeat them in numerical and alphabetical order; the number of items correct was scored.
▶. Processing speed was assessed by two measures from the TEA-Ch, as described in a previous report (H Mulder, NJ Pitchford, N Marlow, in preparation). Raw scores were used in the analyses.
▶. Motor processing speed was scored from the sky search motor control task, in which children are asked to circle targets in a large display without distracters present; time per target was scored.
▶. Verbal processing speed was scored from the first item of the same worlds task, in which children have to read out a string of ones and twos; reaction time was scored.
Funding HM is supported by the Medical Research Council.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Leicestershire, Northamptonshire and Rutland Research Ethics Committee 1.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.