DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY

ORIGINAL ARTICLE

Stability of cognitive performance in children with mild intellectual disability OSKAR G JENNI 1,2

| SYLVIA FINTELMANN 1 | JON CAFLISCH 1 | BEATRICE LATAL 1,2 | VALENTIN ROUSSON 3 |

AZIZ CHAOUCH 3 1 Child Development Center, University Children’s Hospital Z€urich, Z€urich; 2 Children’s Research Center, University Children’s Hospital Z€urich, Z€urich; 3 Centre Hospitalier Universitaire Vaudois and University of Lausanne, Statistical Unit, Institute of Social and Preventive Medicine, Lausanne, Switzerland. Correspondence to Oskar G Jenni at Department of Pediatrics, Child Development Center, University Children’s Hospital, Steinwiesstrasse 75, CH-8032 Z€urich, Switzerland. E-mail: [email protected]

PUBLICATION DATA

Accepted for publication 19th September 2014. Published online 2nd November 2014. ABBREVIATIONS

FSIQ SES

Full-scale IQ Socio-economic status

AIM Longitudinal studies that have examined cognitive performance in children with intellectual disability more than twice over the course of their development are scarce. We assessed population and individual stability of cognitive performance in a clinical sample of children with borderline to mild non-syndromic intellectual disability. METHOD Thirty-six children (28 males, eight females; age range 3–19y) with borderline to mild intellectual disability (Full-scale IQ [FSIQ] 50–85) of unknown origin were examined in a retrospective clinical case series using linear mixed models including at least three assessments with standardized intelligence tests. RESULTS Average cognitive performance remained remarkably stable over time (high population stability, drop of only 0.38 IQ points per year, standard error=0.39, p=0.325) whereas individual stability was at best moderate (intraclass correlation of 0.58), indicating that about 60% of the residual variation in FSIQ scores can be attributed to between-child variability. Neither sex nor socio-economic status had a statistically significant impact on FSIQ. INTERPRETATION Although intellectual disability during childhood is a relatively stable phenomenon, individual stability of IQ is only moderate, likely to be caused by test-to-test reliability (e.g. level of child’s cooperation, motivation, and attention). Therefore, clinical decisions and predictions should not rely on single IQ assessments, but should also consider adaptive functioning and previous developmental history.

Cognitive performance tests are widely used in the clinical evaluation of children with intellectual disabilities. Medical diagnostic and therapeutic procedures as well as the need for special educational services are often decided on the magnitude of intellectual impairment of these individuals. For example, special needs services are offered depending on individual IQ test scores (e.g. children with an IQ10 points). Whitaker stated, however, that only few studies are available for individuals with three or more tests and an initial age of 7y) were tested with five different tests. Information about parental occupation was gathered at study entry and converted into the International SocioEconomic Index.18 The higher International Socio-Economic Index score of the father or mother was used as the socio-economic status (SES) of the child. The SES ranged from 21 to 88 with a median of 40 points on the International Socio-Economic Index scale. We note that the International Socio-Economic Index score in our study sample was lower than in the representative Swiss PISA study (n=1190, mean 52.9 [SD 16.4], unpublished data from Urs Moser, PhD, 2012). The institutional review board confirmed that the study was performed according to the Declaration of Helsinki, and conformed to legal and ethical norms.

Statistical analysis Our modeling approach addressed both definitions of stability, namely the evolution of the population mean over time in a group of children with mild intellectual disability (population stability) and how these individuals track their own centile given that average evolution in the population (individual stability). Stability of longitudinal profiles of FSIQ among children with mild intellectual disability was analyzed using a linear mixed model (see Appendix S1, supporting information published online for more details of the statistical analysis). In this model the FSIQ of an individual is adjusted for the test, sex, and SES at study entry. Cohort and practice effects have also been included in the model as potential confounders (Appendix S1). The mean evolution (population stability) is summarized by the coefficient b5 whereas the intraclass correlation defined on the interval [0,1] quantifies how children globally track their own centile (individual stability or reproducibility), with low values (close to 0) referring to poor tracking or low stability and large values (close to 1) indicating a strong tracking or high stability. Confidence intervals (CIs) for model parameters were calculated using 2000 bootstrap-t samples (see Davison and Hinkley19 [p. 194] and Appendix S1). RESULTS Figure 1 illustrates individual longitudinal profiles of all 36 individuals included in the study. Substantial changes in

individual FSIQ scores occurred between tests, although most individuals remained in the low IQ range (50–85). Table II presents estimates and 95% CIs for model parameters. We note that the test means at study entry (a values) vary. However, these coefficients should be interpreted with caution as they may encompass a possible confounded effect of age at entry. The cohort effect was not significant (p=0.721) which is not surprising since all children were born within a period of only 13 years. On the other hand, for children tested with the same assessment test within 1 year, a practice effect (i.e. IQ gain caused by taking the same test at two consecutive assessments) was observed, with children scoring on average 5.26 points higher (p=0.001) at the second occasion. Neither sex nor SES had a statistically significant impact on FSIQ, although the eight females in our sample had a slightly lower average FSIQ compared to the 28 males (b1=2.31). The evolution of FSIQ in the population was found to be barely negative with an estimated drop of 0.38 points of IQ per year (standard error=0.39), but this result was again not statistically significant in our limited sample (p=0.325). The intraclass correlation was estimated at 0.58 (95% CI 0.46–0.74) indicating that about 60% of the residual variation in FSIQ scores of the participants can be attributed to between-child variability (individual stability).

DISCUSSION Clinicians are frequently asked by parents and other professionals to make predictions for the developmental course of children with intellectual disability. While a prediction can be made or is more robust in children with moderate to severe intellectual disability that is often associated with a genetic syndrome, the prediction for children with borderline to mild intellectual disability is far more difficult. Our knowledge about the longitudinal course of these children is based on studies which have included only two data points. Although the exisiting literature proposes a ‘high’ correlation between initial and subsequent IQ testing in children with non-syndromic intellectual disability,10 we are relatively often faced with the clinical situation where children have made improvements or have declined between assessments. Thus, more data on the longitudinal course of these children are needed. In this study including a true clinical sample, we have used a linear mixed model to study the stability of individual profiles of FSIQ among a group of children with borderline to mild non-syndromic intellectual disability. We used the time since the first evaluation as the metric for recording time while adjusting for the IQ test for two reasons. First, in IQ testing current age and tests are partly collinear because tests change with increasing age. Second, children referred to our center for suspected intellectual disability and with a number of impairments in adaptive functioning had to score between 50 and 85 at their first FSIQ assessment to enter the study. Thus, it was not appropriate to use age for recording time because new participants entering the study would have constantly pulled Cognition in Children with Intellectual Disability Oskar G Jenni et al.

465

IQ Test HAWIK-III HAWIVA HAWIK-R 5

10

15

K-ABC SON-R SON-Y

20

5

10

15

20

5

10

15

20

100 85 70 55 40 100 85 70 55 40

Full-scale IQ

100 85 70 55 40 100 85 70 55 40 100 85 70 55 40 100 85 70 55 40 5

10

15

20

5

10

15

20

5

10

15

20

Age (years)

Figure 1: Illustration of the individual longitudinal profiles of all 36 individuals. K-ABC, German version of the Kaufman-Assessment Battery for Children; HAWIK-R or HAWIK-III, German version of the Wechsler Intelligence Test for Children; HAWIVA, German version of the Wechsler Preschool and Primary Scale of Intelligence; SON-R or SON-Y, Snijders-Oomen Nonverbal Intelligence test. 466 Developmental Medicine & Child Neurology 2015, 57: 463–469

Table II: Parameter estimates with 95% confidence intervals p

Coefficient

Description

Estimate

Standard error

a1 a2 a3 a4 a5 a6 b1 b2

SON-Y test mean SON-R test mean HAWIVA test mean K-ABC test mean HAWIK-R test mean HAWIK-III test mean Cohort effect Practice effect (IQ gain if two consecutive assessments separated by 1y use the same test) FSIQ difference for females (compared to males) FSIQ change associated with an increase of 1 pt SES FSIQ difference per year of follow-up Intraclass correlation

73.77 69.48 67.48 67.81 63.35 72.73 0.19 5.26

2.58 3.37 2.53 2.93 3.47 3.94 0.52 1.61

– – – – – – 0.721 0.001

69.88; 62.22; 62.34; 62.03; 57.38; 66.09; 0.85; 2.80;

77.91 75.84 70.54 73.08 69.45 78.94 1.21 7.66

2.31 0.07 0.38 0.58

4.11 0.09 0.39 –

0.577 0.445 0.325 –

11.95; 0.16; 1.18; 0.46;

9.43 0.29 0.40 0.74

b3 b4 b5 q

95% Bootstrap CI

Bootstrap-t using 2000 replicates (resampling of individuals) were used to construct 95% CI. A variance stabilizing transformation was used to compute the CI of the intraclass correlation. Note that a 95% CI that does not include zero indicates a statistically significant result (p

Stability of cognitive performance in children with mild intellectual disability.

Longitudinal studies that have examined cognitive performance in children with intellectual disability more than twice over the course of their develo...
319KB Sizes 0 Downloads 8 Views