Neurocognitive functioning of children with mild to borderline intellectual disabilities and psychiatric disorders: profile characteristics and predictors of behavioural problems.

BACKGROUND: The aim of the current study was twofold: first, to uncover a neurocognitive profile of normative and relative strengths and weaknesses that characterises an extremely vulnerable group of children with mild to borderline intellectual disabilities (MBID) and co-morbid psychiatric disorders, and second, to investigate the relevance of these neurocognitive functions explaining internalising and externalising symptoms.
METHOD: We recruited 45 children (Mage  = 9.5, SDage  = 1.7; range 6-13 years) with MBID (Full-Scale IQ 50-85) and at least one psychiatric disorder. Neurocognitive functioning was examined utilising the Wechsler Intelligence Scale for Children - Fifth Edition (WISC-V) indices and the Cognitive Task Application (COTAPP), a comprehensive computerised self-paced task designed in such a manner that 'g' (an overall tendency of children with MBID to execute tasks with a slower reaction time and a higher error rate) has been corrected for in the administration of the task (i.e. completely self-paced) and in the operationalisation of outcome measures. Behavioural problems were measured using the CBCL and TRF. One-sample t-tests and binomial tests were carried out to compare performance with normative data. Regression analyses were used to examine the relationship between neurocognitive parameters and mental health.
RESULTS: Compared with normative data, very small to very large effect sizes were found, indicating clear heterogeneity amongst neurocognitive domains relevant for children with MBID. Two prominent neurocognitive weaknesses emerged: processing speed - characterised by slowness and unstableness combined with a high drift rate and delayed processing of the previous trial, particularly under higher cognitive demands - and working memory - in terms of a weaker central executive and 'slave' systems to temporarily store information. Both domains were not clearly predictive of internalising or externalising problems.
CONCLUSION: Children with MBID and psychiatric disorders are hampered by a strongly diminished processing speed and working memory capacity, together resulting in an overall limited processing capacity that may underlie the general developmental delays on domains that depend on fast and parallel processing of information (i.e. language, reading, mathematics and more complex forms of social cognition). Neurocognitive vulnerabilities are neither necessary nor sufficient to explain internalising and externalising problems; rather, a mismatch between the support needs and adaptations these children need, arising from their diminished processing capacity, and the inadequacy of the environment to compensate for this vulnerability may be of relevance.

Background

In recent years, the concept of intellectual disability (ID) has been evolving at a rapid pace. A major turning point in this process was the publication of the DSM‐5, which represented a shift from a number‐based ‘disability’ approach to a more neurobiological ‘disorder’ emphasis: the former relying on an IQ score obtained from an intelligence test and the latter relying to a greater extent on measures of neuropsychological and adaptive functioning (Greenspan & Woods 2014; Cervantes et al. 2019). The DSM‐5 even explicitly states that ‘[individual] cognitive profiles based on neuropsychological testing are more useful for understanding intellectual disabilities than a single IQ score’ (American Psychiatric Association 2013). Although there is accumulating scientific evidence in favour of this statement, many questions remain about the cognitive characteristics of individuals with ID and the value they hold in understanding their behaviour. This is especially the case for children with mild to borderline intellectual disabilities (MBIDs; i.e. an IQ between 50 and 85) with co‐morbid psychiatric disorders (we use this designation over the DSM‐5 term of ‘mental disorder’ as we like to emphasise that these conditions are not purely ‘mental’, which is consistent with modern philosophical and neuroscientific views; see Stein et al. 2010 for a comprehensive plea on terminology). In clinical practice, these children often present themselves with a wide range of behavioural, social and practical difficulties. However, the lion's share of studies mainly set the presence of a psychiatric condition as an exclusion criterion (McCarthy & Barbot 2016); although in more recent studies, the merits of including this most vulnerable group of children are recognised (Palmqvist et al. 2020). This is especially remarkable considering the fact that psychiatric co‐morbidity in children with MBID is highly frequent with consistently reported prevalence rates between 30% and 50% (Einfeld et al. 2011; Melby et al. 2020). The results found in studies on children with MBID thus far may therefore have substantial limitations in the generalisability to clinical populations. Hence, our aims are to examine neurocognitive profiles of children with MBID and co‐morbid psychiatric disorders (e.g. attention deficit hyperactivity disorder and autism spectrum disorder) and the relevance thereof explaining internalising and externalising symptoms.

When investigating neurocognitive profiles of children with MBID, the results hitherto suggest a particularly impaired performance on tasks aiming to measure executive functions (EFs) (Alloway 2010; Danielsson et al. 2012). Working memory (WM) has herein been described as a major deficit, with most pronounced problems on verbal WM tasks (Roording‐Ragetlie et al. 2018). Weaker performances on inhibitory and interference control have also been found (Van der Meer & Van der Meere 2004; Van der Meere et al. 2008; Bexkens et al. 2014). Furthermore, several studies have demonstrated an overall slow processing speed (PS) in children with MBID, although this might only be related to situations where quick and/or complex responding is required (Un & Erbahçeci 2001; Schuiringa et al. 2017). It has been suggested that this slower time to perceive, process and respond to a stimulus is a significant influential factor on performances that is not always taken into account in tasks aiming to measure higher‐order cognitive processes (Rommelse et al. 2015; Biesmans et al. 2019). This may be of particular relevance when studying neurocognitive functioning of children with MBID, specifically, and for children with neurodevelopmental problems in general (Toffalini et al. 2017; Alloway 2018; Braaten et al. 2020; Rommelse et al. 2020). In reverse, several specific attentional processes (i.e. arousal regulation and sustained attention) have been found relatively spared in youth with MBID in recent studies (Puga et al. 2019; Zagaria et al. 2021). Nevertheless, PS and attention, together with other domains – like decision‐making, learning efficiency and self‐regulation – so far have been scarcely studied in research on this group due to a strong emphasis on aforementioned EFs or have only been investigated in specific groups like children with Down syndrome (Lanfranchi et al. 2010) or Williams syndrome (Menghini et al. 2010).

In youth with ID, several studies have found a relationship between the presence of neurocognitive deficits, predominantly impaired inhibition, and the severity of their intellectual functioning, highlighting the importance of neurocognitive functioning in the ability to show adaptive and social behaviour in daily life (Van Nieuwenhuijzen et al. 2009; Gligorovic & Buha Ethurovic 2014; Van Nieuwenhuijzen et al. 2017). Strong associations between EF decrements and psychiatric problems in children have also repeatedly been corroborated in research in clinical – yet average to above average IQ – populations (Schoemaker et al. 2013; Braaten et al. 2020), essentially acting as an overall transdiagnostic mechanism (Snyder et al. 2015; Bloemen et al. 2018). Especially in preschoolers, EFs have been found to predict externalising problems, with disinhibition as the most considerable feature (Hughes & Ensor 2008). Concerning children with MBID, the research to date on this topic is scant. One study on children with developmental delays found broad cognitive skills (although combined with adaptive skills) to predict internalising but not externalising problems (Mitchell & Hauser‐Cram 2009), presumably confounded by the presence of speech problems. However, the accumulating research on predictors of internalising and externalising problems in children with MBID so far has predominantly affirmed the importance of adverse environmental and socio‐demographic determinants as conditions for developing psychopathology. Family dysfunctioning, chronic parental stress and adverse parental behaviour can herein be considered as most influential risk factors next to the presence of co‐morbid physical symptoms (Dekker & Koot 2003; Wallander et al. 2006; Tomic et al. 2012). Particularly for the MBID population, also a low socio‐economic status is associated with a higher vulnerability for mental health problems to emerge in children (Essex et al. 2006; Tomic et al. 2012). Given the limited amount of studies on this topic, more research is needed to substantiate these tentative results and to elucidate the importance of neurocognitive dysfunctioning as a potential risk factor.

Hence, the aim of our study is twofold: first, to uncover a neurocognitive profile of normative and relative strengths and weaknesses that characterises an extremely vulnerable group of children (6–13 years) with MBID and co‐morbid psychiatric disorders, and second, to examine which of these neurocognitive parameters contribute to internalising and externalising psychopathology. To our knowledge, this has not been studied before in this specific group of children, thereby making our study rather exploratory in nature. We hypothesised that children with MBID and psychopathology experience most pronounced weaknesses in EF and PS compared with typically developing (TD) children. We further anticipated that particularly these domains relate to behavioural problems. In addition to the Wechsler Intelligence Scale for Children – Fifth Edition (WISC‐V) index scores, parameters of the Cognitive Task Application (COTAPP) – a comprehensive computerised test for children – were used. The COTAPP has been designed in such a manner that ‘g’ [an overall tendency of children with MBID to execute tasks with a slower reaction time (RT) and a higher error rate] has been corrected for in the administration of the task (i.e. completely self‐paced) and in the operationalisation of outcome measures by correction for general measures [i.e. (variability in) RT] when calculating specific measures (i.e. reduction in RT after a prolonged period of performing the task, indicative of the ability to sustain attention).

Methods

Participants

In total, 51 children (6–13 years old) with MBID [Full‐Scale IQ (FSIQ) 50–85] and at least one psychiatric disorder were recruited from a clinical facility for pediatric psychiatry (Karakter) between September 2018 and November 2019 (see Fig. 1). FSIQ was based on the Dutch version of the Wechsler Intelligence Scale for Children – Fifth Edition (WISC‐VNL) (Wechsler 2018). Six enrolled children were excluded for analysis based on an FSIQ > 85. Learning disorders (e.g. dyslexia and dyscalculia) or specific language impairments were considered exclusion criteria, just as children with acquired brain injury and severe sensorimotor problems. We included children with both non‐syndromic and syndromic ID, considering the accumulating evidence that non‐syndromic ID, although perhaps more subtle and less apparent than syndromic ID, is also governed by complex neurodevelopmental processes causing a high degree of heterogeneity of non‐syndromic ID phenotypes (Kaufman et al. 2010; Lee et al. 2018). The participants' demographics are shown in Table 1.

Figure 1

Flow chart of study enrolment process. FSIQ, Full‐Scale IQ. [Colour figure can be viewed at wileyonlinelibrary.com]

Table 1

Demographics of the sample (n = 45)

	n	%
Gender
Male	26	58%
Female	19	42%
DSM‐5 classification
Neurodevelopmental disorders
ADHD	28	62%
ASD	19	42%
Tourette's disorder	1	2.2%
Provisional tic disorder	1	2.2%
Developmental coordination disorder	1	2.2%
Speech sound disorder	1	2.2%
Unspecified communication disorder	1	2.2%
Global developmental delay	1	2.2%
Unspecified neurodevelopmental disorder	3	6.7%
Other specified neurodevelopmental disorder	1	2.2%
Disruptive, impulse control and conduct disorders
Oppositional defiant disorder	6	13.3%
Intermittent explosive disorder	1	2.2%
Trauma‐related and stressor‐related disorders
Post‐traumatic stress disorder	3	6.7%
Reactive attachment disorder	2	4.4%
Disinhibited social engagement disorder	2	4.4%
Other specified trauma‐related and stressor‐related disorder	1	2.2%
Anxiety disorders
Generalised anxiety disorder	2	4.4%
Separation anxiety disorder	1	2.2%
Unspecified anxiety disorder	1	2.2%
Other specified anxiety disorder	3	6.7%
Depressive disorders
Unspecified depressive disorder	1	2.2%
Elimination disorder
Enuresis	1	2.2%
Encopresis	2	4.4%
Somatic symptom and related disorders
Somatic symptom disorder	1	2.2%
Other
Unspecified psychiatric disorder	1	2.2%
	M	SD	Min.	Max.
Age (years)	9.5	1.7	6.4	13.0
WISC‐V Full‐Scale IQ (FSIQ)	73.1	7.7	50	85
CBCL/6‐18 (in T‐scores)
Internalising	62.4	11.2	34	84
Externalising	61.7	11.8	13	81
TRF/6‐18 (in T‐scores)
Internalising	63.8	8.9	47	86
Externalising	59.6	9.6	41	90

ADHD, attention deficit hyperactivity disorder; ASD, autism spectrum disorder; WISC‐V, Wechsler Intelligence Scale for Children – Fifth Edition; CBCL/6‐18, Child Behavior Checklist; TRF/6‐18, Teacher Report Form; M, mean; SD, standard deviation.

Measures

Neuropsychological measures

Neurocognitive functioning was examined using the COTAPP. The COTAPP (Rommelse et al. 2018) is a computerised self‐paced task with a block‐wise structure. The test consists of seven blocks containing specific manipulations that require the use of different aspects of cognitive functioning (for a detailed description, see Tables S1 and S2). The COTAPP has been structured in such a manner that RT, variability in RT and errors – amongst others – have been aggregated across blocks that consist of a two‐choice RT paradigm. Difference scores are generated that contrast RT speed in two blocks that differ in terms of a specific manipulation such as adding distraction or a dynamic tracking algorithm that rewards faster responses than the child's own median RT. As such, both generic and specific neurocognitive processes can be examined. In addition, observations are coded regarding motor and verbal behaviour and required coaching to complete the tasks. A total of 22 parameters are generated being subdivided into six domains: PS, attention, executive control, (working) memory, learning speed and behaviour during the task. Task instructions are embedded in the task. Two response buttons are needed to complete the task (right CTRL = ‘yes’/in green with a tick of approval; left CTRL = ‘no’/in red with a cross). However, all button presses are recorded.

The interpretation of COTAPP parameters is norm based. Norms were constructed using continuous norming based on an age‐balanced and gender‐balanced group of N = 1032 Dutch children (5–13 years) representative for IQ, migration background, geography and parental education. High internal reliability was found for RT scores on all blocks (Cronbach's α between 0.80 and 0.95). The split‐half reliability (based on intraclass correlation coefficients on even and uneven numbered trials) of 14 out of 20 parameters were above 0.70, but mostly >0.80 up to 0.99. For five of the six remaining parameters, split‐half coefficients were around 0.60. For one parameter (errors WM very complex condition), assumptions for parallelism of test halves were not met. Test–retest reliability was 0.75. For the current study, 18 out of 22 parameters were included for analysis. Four parameters consisted of complex and very complex WM tasks only administered when the child's performance level clearly exceeded probability level (≥85% correct). Because the children in our sample hardly entered this block, these parameters were excluded.

Additionally, the five indices of the WISC‐VNL (Wechsler 2018) are incorporated as parameters. These parameters may also reveal relative strengths and weaknesses relevant in understanding MBID, because very different composition of index scores can lead to a similar FSIQ.

Behavioural problems

The internalising and externalising composite scores of the CBCL, filled in by parents/guardians, and the TRF, filled in by teachers, were used (Verhulst et al. 1996; Verhulst et al. 1997; Achenbach & Rescorla 2001). Both are informant‐rated questionnaires from the Achenbach System of Empirically Based Assessment (ASEBA) for measuring emotional and behavioural problems in children in the preceding 6 months. Raw scores were converted into standardised T‐scores (M = 50; SD = 10), with a higher score indicating more severe problems. Applicability of both questionnaires to the ID population has been supported by data on scale reliability and stability (Dekker et al. 2002).

Procedure

The study was approved by the research ethics committee of the Radboud University Medical Center Nijmegen (no. 2018‐4441) and is in accordance with the Declaration of Helsinki. Procedures were in line with the EU guidelines on the General Data Protection Regulation. Care providers were asked to inform children and their legal representatives about the study and for written consent on sharing contact details with the research team. A member of the research team then contacted the legal representatives providing them with more information about the study and answering questions. By agreement to participate, representatives were asked for written informed consent, and children provided oral consent (or written consent ≥12 years). Every child was seen a maximum of two times: one for administering the WISC‐VNL intelligence test (±1.5 h) and another for administering the COTAPP (±35–40 m). In most cases, both assessments were within a 2‐week time period. If a child was on medication with a short duration of effect (e.g. methylphenidate), the use was discontinued on the morning prior to test administration. Youngsters taking psychoactive medication that required a gradual (>24 h) build‐up phase were excluded. If one of the tests was already administered within 1 year of research participation, the data were requested. If, within 1 year prior to participation, one of the tests was administered while being on medication or if any other intelligence test was administered, children were unable to participate. After completion, children received a printed certificate and a small gift as a reward. The legal representatives received a summarised report of the test results.

Data analysis

All analyses were performed using IBM SPSS Statistics 24.0 (IBM Corp. 2016). There were no missing values and no restriction of range. Both independent and dependent variables were continuous ratio variables and corrected for age and sex, because all scores were based on norm‐referenced tests. These were then converted into standardised z‐scores (Van der Waerden method). Effect sizes were calculated utilising Psychometrica's online tool (Lenhard & Lenhard 2016), interpreted by the rules of thumb as defined by Cohen (1988) and expanded by Sawilowsky (2009). The significance level was set at P ≤ 0.05.

To examine neurocognitive profiles of the clinical group compared with norm values, one‐sample t‐tests (normally distributed neurocognitive variables of COTAPP and WISC‐V) or binomial tests (binary variables) were used to compare our children with MBID with the mean/median values derived from the norm sample (Rommelse et al. 2018; Wechsler 2018). To examine the relation between these neurocognitive variables and internalising and externalising problems, first, only predictors were selected that correlated 0.20 or higher with the outcome based on individual correlations between the predictors and internalising and externalising outcomes (see Table S3). The selection was performed independently for internalising and externalising problems on an ‘either/or’ basis for parent and teacher report. Second, separately for internalising and externalising problems, exploratory models were run on parent‐reported symptoms applying the forward selection method in which selected variables were sequentially entered to determine the most saturated model, using a probability of 0.05 and 0.10 as thresholds for entry and removal of predictors at each iteration. Selected variables were then used in confirmatory models run on teacher‐reported symptoms. The same procedure was repeated, but with teacher‐reported symptoms as starting point and confirmed against parent‐reported symptoms. This resulted in the selection of cognitive predictors that were relevant to mental health problems in our group of children with MBID at both home and school. Similarities and discrepancies between both parallel routes were interpreted.

Results

Descriptive data

The sample consisted of 26 boys (58%) and 19 girls (42%). The mean age was 9.5 years (SD = 1.8). The mean FSIQ was 73.1 (SD = 7.7). The average number of psychiatric classifications per child was 1.76 with a median of 2. Most children were diagnosed with attention deficit hyperactivity disorder (62%) and/or autism spectrum disorder (42%). Four children had syndromes (1q [del(1)(q24q25.3)], 18p11.32 duplication, Noonan syndrome and TRIO syndrome) (see Table 1 for an overview of the descriptive statistics). Of interest is that despite the uniform inclusion criterion of FSIQ ≤ 85, a high degree of heterogeneity emerged in intelligence profiles (e.g. the scores for the Processing Speed Index ranging from 45 to 111). This is visualised in Figure S1.

Neurocognitive profile of children with mild to borderline intellectual disabilities and psychiatric disorders

Assumptions on independence of observation, normality and unusual points were checked, and no violations were found. As expected, for all five WISC‐V parameters, children with MBID showed weaker performances compared with the standardised values with very large to huge effect sizes. For the majority of COTAPP parameters (12 out of 18), on average, children with MBID showed weaker performances compared with age‐based and sex‐based norms. Very small to very large effect sizes were found, indicating clear heterogeneity amongst predictors in relevance to children with MBID. Using an ordering based on Cohen's d boundaries, very large effect sizes were found for sloppiness and learning speed. Large effect sizes were found for verbalisation, RT variability, RT, RT under cognitive load and WM RT. Medium effect sizes were found for attentional lapses, required coaching, variability in decision time, delay discounting and WM errors. Small effect sizes were found for inaccuracy, interference control and sustained attention. Very small effect sizes were found for arousal regulation, capacity to improve RT and body movements. When statistically comparing the neurocognitive parameters against each other by examining (non‐)overlapping 95% confidence intervals (CIs), the CIs for the predictor with a very small and negative effect size (body movements) did not overlap with the 12 predictors with medium to huge effect sizes. The CIs for the three predictors with very small (positive) effect sizes (capacity to improve RT, arousal regulation and body movements) did not overlap with the seven predictors with large to huge effect sizes. The CIs for predictors with medium effect sizes did not overlap with predictors with huge effect sizes. These findings suggest not all predictors to be equally relevant in describing the most prominent neurocognitive vulnerabilities. The results are shown in Table 2 and are visualised in Figure 2. On two neurocognitive domains, prominent weaknesses emerged: PS [WISC‐V: Processing Speed Index; COTAPP: RT, RT variability, RT under cognitive load, variability in decision time, attentional lapses (i.e. high drift rate) and sloppiness (i.e. more ‘premature’ responses and ‘extra responses’ that are most likely indicative of prolonged processing of the previous trial)] and WM (WISC‐V: specifically WMI, but also Fluid Reasoning Index and Visual Spatial Index require holding and manipulating information in WM; COTAPP: learning speed, WM RT and errors). Increased verbalisation and required coaching indicated a higher dependence on both self‐speech and the examiner to complete the task.

Table 2

Results of the comparisons of neurocognitive variables between the MBID group and mean/median values derived from the norm samples of COTAPP and WISC‐V

Outcome measure	MBID group (n = 45)		COTAPP norms (n = 1032)		Raw difference					Standardised effect size
					Mean difference	95% CI for difference		t‐value	Sig. (2‐tailed) †	Effect size Cohen's d	95% CI for effect size		Interpretation effect size
	Mean	SD	Mean	SD		Lower	Upper				Lower	Upper
Reaction time	0.87	1.17	0.00	1.00	0.87	0.52	1.22	5.00	<0.001	0.86	0.56	1.16	Large
Reaction time variability	0.88	1.16	0.00	1.00	0.88	0.53	1.23	5.07	<0.001	0.87	0.57	1.17	Large
Reaction time under cognitive load	0.90	1.31	0.05	1.00	0.85	0.46	1.25	4.36	<0.001	0.84	0.54	1.14	Large
Capacity to improve reaction time	0.07	1.54	0.02	0.99	0.05	−0.41	0.52	0.24	0.814	0.05	−0.25	0.35	Very small
Variability in decision time	0.64	0.95	0.00	1.00	0.64	0.36	0.93	4.55	<0.001	0.64	0.34	0.94	Medium
Sustained attention	0.44	1.57	0.05	0.98	0.39	−0.08	0.86	1.66	0.104	0.39	0.09	0.68	Small
Arousal regulation	0.15	1.31	0.09	0.97	0.06	−0.33	0.45	0.31	0.759	0.06	−0.24	0.36	Very small
Attentional lapses	0.71	1.23	−0.08	1.01	0.79	0.43	1.16	4.34	<0.001	0.77	0.47	1.07	Medium
Sloppiness	−0.23	1.73	−1.58	1.14	1.35	0.83	1.87	5.21	<0.001	1.15	0.85	1.46	Very large
Interference control	0.44	1.60	0.03	0.99	0.41	−0.07	0.89	1.71	0.093	0.40	0.10	0.70	Small
Delay discounting	0.82	1.43	0.06	1.24	0.76	0.33	1.19	3.56	<0.001	0.61	0.31	0.91	Medium
Working memory – reaction time	0.84	1.09	0.00	1.00	0.84	0.51	1.17	5.17	<0.001	0.84	0.54	1.14	Large
Working memory – errors	0.00	2.00	−0.93	1.71	0.93	0.33	1.53	3.13	0.003	0.54	0.24	0.84	Medium
Inaccuracy	−0.58	1.27	−0.69	N/A	0.11	−0.27	0.50	0.60	0.552	0.43	0.13	0.73	Small
Body movements	−0.29	1.26	−0.18	1.06	−0.11	−0.49	0.27	−0.58	0.567	−0.10	−0.40	0.20	Very small
WISC‐V Processing Speed Index	−1.65	1.27	0.00 ‡	1.00 ‡	−1.65	−2.03	−1.27	−8.76	<0.001	1.31	1.70	0.90	Very large
WISC‐V Working Memory Index	−1.64	0.61	0.00 ‡	1.00 ‡	−1.64	−1.82	−1.45	−17.91	<0.001	2.67	3.29	2.04	Huge
WISC‐V Verbal Comprehension Index	−1.28	0.58	0.00 ‡	1.00 ‡	−1.28	−1.46	−1.11	−14.90	<0.001	2.22	2.76	1.67	Huge
WISC‐V Visual Spatial Index	−1.28	0.84	0.00 ‡	1.00 ‡	−1.28	−1.53	−1.02	−10.21	<0.001	1.52	1.95	1.09	Very large
WISC‐V Fluid Reasoning Index	−1.37	0.60	0.00 ‡	1.00 ‡	−1.37	−1.55	−1.19	−15.28	<0.001	2.28	2.83	1.72	Huge

Numbers do not add up due to missing data. Interpretation of effect sizes by the rules of thumb as defined by Cohen (1988) and expanded by Sawilowsky (2009).

One‐sample t‐tests.

For WISC‐V variables, the mean is compared with standardised values (z‐scores). For COTAPP variables, the mean is compared with the norm group corrected for age and/or sex.

Total sample size is 1032.

Chi‐squared test.

MBID, mild to borderline intellectual disabilities; COTAPP, Cognitive Task Application; WISC‐V, Wechsler Intelligence Scale for Children – Fifth Edition; CI, confidence interval; SD, standard deviation; N/A, not available.

Figure 2

Sorted Cohen's d effect sizes for standardised neurocognitive variables with 95% confidence intervals. WISC‐V, Wechsler Intelligence Scale for Children – Fifth Edition. The asterisk (*) indicates the sample was selected based on a Full‐Scale IQ <85, so these effects are expected to be the largest. [Colour figure can be viewed at wileyonlinelibrary.com]

MBID group

(n = 45)

COTAPP norms

(n = 1032)

Low/Yes

n (%)

High/No

n (%)

Low/Yes

n (%)

High/No

n (%)

Neurocognitive predictors of internalising and externalising problems in children with mild to borderline intellectual disabilities and psychiatric disorders

All referenced tables are included in the supplement. Based on individual correlations (≥0.20) between the dependent and independent variables, nine predictors (seven COTAPP and two WISC‐V) were selected for internalising problems and 11 (nine COTAPP and two WISC‐V) for externalising problems (Table S3). Assumptions on linearity, homoscedasticity and normality were checked, and no violations were found based on fit indices, visual inspection of the partial plots and plots of studentised residuals versus unstandardised predicted values. Furthermore, assumptions on multicollinearity and independence of observation were met. Inspection of unusual points revealed a total of 20 cases with elevated leverage values, with one case exhibiting a very high leverage (0.61). Because this case overall had the most significant influence on several models, it was excluded from the analyses, which resulted in N = 44. The forward selection model predicting internalising problems according to parents included two predictors: more verbalisation (COTAPP) and better visual–spatial information processing (WISC‐V) were associated with higher internalising symptoms [verbalisation (β = −0.45) and Visual Spatial Index (β = 0.28)]. A confirmatory regression using this model predicting teacher‐reported internalising problems fitted significantly better than the default model (ΔF = 4.87, ΔR 2 = 0.20, ΔP = 0.01) including sex and age alone and explained an additional 20% of variance (see Table S4). Sensitivity analyses using an exploratory regression on teacher‐rated internalising problems followed by a confirmatory regression analysis on parent‐rated internalising problems resulted in nearly similar outcomes (Table S5). Only arousal regulation (β = 0.33) (COTAPP) was additionally selected, but exclusively for teacher‐reported symptoms. Utilising the same procedure, the forward selection model predicting externalising problems generated no consistent predictive pattern across raters (Tables S6 and S7). Based on significant, mostly small, correlations (Table S3), externalising symptoms related to a general poorer performance.

We computed an observed power for all exploratory multiple regression analyses (Tables S4–S7) using the G*Power 3.1 software package (Faul et al. 2009), given an observed probability level of 0.05, the number of predictors in the model and the observed R 2. The effect sizes used for this assessment were as follows: small (f 2 ≥ 0.02), medium (f 2 ≥ 0.15) and large (f2 ≥ 0.35) (see Cohen 1988). The post hoc analyses revealed that the statistical power for this study ranged between 0.781 (TRFEXT) and 0.872 (TRFINT) to detect medium effect sizes. Power ranged between 0.935 (CBCLEXT) and 0.995 (CBCLINT) for the detection of large effect sizes. Thus, there was adequate power at the medium to large effect size level.

Discussion

The current study set out to uncover a neurocognitive profile of normative and relative strengths and weaknesses that characterises an extremely vulnerable group of children with MBID and co‐morbid psychiatric disorders, and the relevance thereof explaining internalising and externalising symptoms. We hypothesised that children between the age of 6 and 13 with MBID and psychopathology experience most pronounced weaknesses in EFs and PS relative to other cognitive domains compared with TD children. On average, the neurocognitive profile of children with MBID was characterised by prominent weaknesses in combination with more mildly affected or completely spared abilities, indicating neurocognitive profiling is also informative in children with a ‘weak g factor’. Conforming expectations, one of the most prominent weaknesses in children with MBID was PS, typified by slowness and unstableness combined with a high drift rate and delayed processing of the foregoing trial(s), particularly under higher cognitive demands. In contrast to expectations, no pronounced general weaknesses in EF or attentional control were found: when controlling for PS, children with MBID did not substantially deviate from normative development regarding interference control, (impulsive) errors, sustained attention or arousal regulation. However, WM capacity in terms of the central executive and ‘slave’ systems to temporarily store information was clearly impaired. Unexpectedly, these major deficits in their neurocognitive profile did not relate to both internalising and externalising symptoms. Rather, a better visual–spatial ability and higher tendency to verbalise during task performance predicted both parent‐rated and teacher‐rated internalising symptoms. Yet no consistent predictive pattern was found for externalising problems, although poorer performance across several domains related to more externalising symptoms.

Our findings are supported by past research as they fit within neuropsychological studies of childhood behavioural and psychiatric disorders mentioning poor PS and WM problems as the most ubiquitous results (Brunnekreef et al. 2007; Willcutt et al. 2008; Nigg et al. 2017; Alloway 2018; Braaten et al. 2020). Regarding WM, the results also broadly correspond with the existing body of knowledge regarding individuals with MBID, stating that this domain is crucial in understanding their functioning (Carretti et al. 2010; Danielsson et al. 2012; Schuiringa et al. 2017; Roording‐Ragetlie et al. 2018). The fact that the children in our sample made greater use of talking, humming and/or singing might be interpreted as ways of enhancing task engagement due to a delay in their inner speech development, related to structural deficits in verbal WM that children with MBID tend to show (Alderson‐Day & Fernyhough 2015; Roording‐Ragetlie et al. 2018). This is consistent with findings regarding delayed use of verbal strategies and subvocal repetition in children with MBID (Poloczek et al. 2014; Poloczek et al. 2016) and may also reflect a more anxious coping as we have found verbalisation to predict internalising problems. Furthermore, the profound PS difficulties that our sample showed confirm the relevance of a slow and unstable PS for children in the MBID population (Un & Erbahçeci 2001; Schuiringa et al. 2017) and are consistent with recent views that a pervasive slow/unstable PS is a characteristic feature of a broad spectrum of child neuropsychiatric disorders (Braaten et al. 2020; Rommelse et al. 2020). PS and WM capacity are often found to be interrelated (as was the case in our study), and together, they are considered the core of ‘processing capacity’: the efficiency and adaptive flexibility to use available neurocognitive capacity (Rommelse et al. 2020). Processing capacity typically develops with increasing age: children are increasingly able to process large amounts of information in parallel when task demands are complex (Miller 1956). This, in turn, is crucially important for the mental representation of concepts of higher relational complexity essential to abstract thinking/intelligence (Halford et al. 1998). For this reason, it has been hypothesised that intelligent individuals have a very specific advantage in the speed of higher‐order information processing, explained by a more efficient transmission of information from frontal attentional and WM processes to temporal–parietal processes of memory storage (Schubert et al. 2017). Following this line of reasoning, it may be concluded that children with MBID are hampered by an inefficient transmission of information in these brain networks, resulting in broad developmental delays on domains depending on the fast transfer of large amounts of information (i.e. language, reading, mathematics and more complex forms of social cognition).

Our findings further suggest that some neurocognitive domains are relatively spared in children with MBID. In contrast to some previous studies, we found no convincing evidence that general difficulties in attentional control or executive functioning can be regarded as key problems in individuals with MBID (Van der Meer & Van der Meere 2004; Van der Meere et al. 2008; Danielsson et al. 2012; Bexkens et al. 2014). We propose that this dissimilarity can be best explained by the fact that previous studies did not correct for PS in the operationalisation of EF measures as was the case in our study, as EF measures are often confounded by PS, also called task impurity (Snyder et al. 2015; Nigg et al. 2017). Additionally, poorer performance on EF tasks may also be the result of not correcting for PS during task administration. Our results point out that children with MBID are not necessarily more error prone than their TD peers but well capable of performing accurately when given sufficient time to respond, time to learn the task and (if needed) external coaching. This matches earlier findings about the information processing profiles of children with internalising and externalising problems (Brunnekreef et al. 2007; Schuiringa et al. 2017). Currently, researchers already made attempts in investigating the role of coaching as a possible beneficial component in the treatment of children with MBID (Roording‐Ragetlie et al. 2017; Favre et al. 2018). Taken together, it is too imprecise to link a general weaker EF to MBID; poorer performance on tasks aimed to measure EF may be explained by a range of non‐EF factors that are not taken into account.

Unexpectedly, processing capacity (PS and WM) did not clearly relate to internalising or externalising symptoms. Rather, parent‐rated and teacher‐rated internalising symptoms related to better visual–spatial abilities and higher tendency to verbalise during task performance. This is partly consistent with Mitchell & Hauser‐Cram (2009) who found a small effect for cognitive/adaptive skills in explaining internalising problems in young children with MBID (possibly confounded by speech problems). No consistent predictive pattern for externalising problems was found, although correlations were all in the direction of poorer neurocognitive abilities relating to somewhat higher parent‐rated and teacher‐rated externalising problems. While one explanation may be that parents and teachers both might have different frames of reference for assessing children's behaviour, another possibility may be due to differences in demands on attentional and self‐regulatory capacities that various settings put on these children (Tassé & Lecavalier 2000; Dekker et al. 2002; Jacobson et al. 2011). In the latter case (for which there is greater scientific support), internalising and externalising problems in this particular group of children may emerge due to a mismatch between the support needs and adaptations these children need arising from their diminished processing capacity, and the inadequacy of the environment to compensate for this vulnerability (Baker et al. 2003; Essex et al. 2006; Tomic et al. 2012; Meppelder et al. 2015).

Strengths and limitations

An important strength of our study was the use of a computerised task adjusting for weaker PS in children with MBID in both the task administration and operationalisation of neurocognitive parameters, by which we were able to circumvent the task impurity problem. Many of these domains have not been investigated in children with MBID before but are of potential relevance in understanding their functioning. The use of such tasks is highly recommended in future research. Also, several limitations of our study are discussed. First, we were confronted with a limited sample size. Even though the research team made great efforts in the recruitment of participants, a substantial amount did not respond after repeated attempts or declined participation. Interpretation was therefore mainly based on effect sizes. A second limitation was the exclusion of youngsters taking psychoactive medication that required a gradual (>24 hour) build‐up phase, possibly causing a selection bias as treatment of children with the most severe behavioural problems usually involves pharmacological interventions. However, given the very significant findings for neurocognitive problems in the included group, the overall weak association between neurocognitive problems and internalising and externalising symptoms is unlikely to be related to the exclusion of these specific children. We also excluded children with learning disorders or specific language impairments. In retrospect, we do not have a clear rationale for this. However, we do not think this impacted the findings, as only two children were excluded based on these criteria.

Conclusion and implications

Concerning children with MBID and psychiatric disorders, we have provided evidence for a combination of prominent weaknesses in conjunction with mildy affected and spared abilities in comparison with TD peers, demonstrating the relevance of neurocognitive profiling in children with MBID. Two prominent weaknesses are PS and WM, together accounting for a strongly diminished processing capacity. The inefficiency of fast and parallel processing of information may explain the general developmental delays on domains that depend on this processing capacity: that is, language, reading, mathematics and more complex forms of social cognition. However, the current results suggest that it is inaccurate to interpret that as a ramification; the development of all neurocognitive skills will be limited. When accommodating for extra needs by offering extra practice opportunities, coaching and avoiding time pressure, performance on several EF and attentional control measures is comparable with TD children. The fact that neither PS nor WM was clearly predictive of internalising or externalising problems suggests that not neurocognitive but environmental determinants may be key risk factors for the emergence of these problems. Children with MBID and psychiatric disorders do not by definition develop these problems but are possibly more prone to encounter a mismatch between the support needs arising from their exceedingly vulnerable neurocognitive profiles and an environment that is insufficiently meeting those needs.

Source of Funding

This research was funded by Karakter Child and Adolescent Psychiatry. No external funding was received.

Conflict of Interest

The authors report no conflicts of interest.

Table S1. The Cognitive Task Application (COTAPP) – Task description

Table S2. The Cognitive Task Application (COTAPP) – Output parameters

Table S3. Pearson correlation matrix including the 18 COTAPP parameters, 5 WISC‐V indices, and internalising and externalising subscales of the CBCL and TRF

Table S4. Multiple regression analyses with selected neurocognitive predictors of internalising problems on exploratory models as rated by parents (CBCL) via enter method (age, sex) and then forward selection method; subsequently on confirmatory models as rated by teachers (TRF) via enter method (n = 44)

Table S5. Multiple regression analyses with selected neurocognitive predictors of internalising problems on exploratory models as rated by teachers (TRF) via enter method (age, sex) and then forward selection method; subsequently on confirmatory models as rated by parents (CBCL) via enter method (n = 44)

Table S6. Multiple regression analyses with selected neurocognitive predictors of externalising problems on exploratory models as rated by parents (CBCL) via enter method (age, sex) and then forward selection method; subsequently on confirmatory models as rated by teachers (TRF) via enter method (n = 44)

Table S7. Multiple regression analyses with selected neurocognitive predictors of externalising problems on both exploratory models as rated by teachers (TRF) via enter method (age, sex) and then forward selection method; subsequently on confirmatory models as rated by parents (CBCL) via enter method (n = 44)

Figure S1. Individual profile plot of the standard scores of WISC‐VNL indices for all 45 cases

Click here for additional data file.