Intellectual disability is defined within the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders, fourth edition, Text Revision (DSM-IV), as a standard score of approximately 70 on an individually administered measure of cognitive ability and ‘concurrent deficits or impairments in adaptive functioning’. Diagnostic criteria for intellectual disability in DSM-5 continue to include requirements for both intellectual and adaptive impairment. In assessing intelligence, clinicians may choose between many standardized measures, some of which estimate overall intellectual ability by measuring two core domains – verbal and non-verbal reasoning skills – while others include additional domains (e.g. efficiency of information processing). One of the most commonly used IQ tests, the Wechsler Intelligence Scale for Children, fourth edition (WISC-IV), offers both a Full Scale IQ (FSIQ) – which is comprised of scores from four domains, including verbal comprehension, perceptual reasoning, processing speed, and working memory – and a General Abilities Index (GAI), which is calculated using only the verbal comprehension and perceptual reasoning scores (see Fig. 1).
Figure 1. Wechsler Intelligence Scales for Children, fourth edition (WISC-IV) Full Scale IQ (FSIQ) and General Abilities Index (GAI) subscale composition.
Download figure to PowerPoint
The use of more comprehensive measures of intellectual ability versus briefer, more targeted measures may have important implications for assessment of ‘intellectual impairment’. Thus, the distinction between GAI and FSIQ is important because the latter broadens its definition of intelligence to include measures of cognitive efficiency, such as working memory and processing speed. Use of the FSIQ potentially increases the occurrence of false positives when identifying intellectual disability, i.e. children may achieve lower FSIQ scores because of the speed and attentional or working memory demands that are part of this summary score, even though their broad reasoning skills may be largely intact. In non-clinical populations, the GAI is considered to provide an estimate of intellectual functioning comparable to that offered by the FSIQ. However, slowed processing speed is a common characteristic of a variety of neurocognitive disorders. For example, processing speed impairments have been found in reading disabilities,[4, 5] traumatic brain injury,[6, 7] and attention-deficit–hyperactivity disorder[4, 5, 8] (ADHD). In addition, working memory weaknesses are commonly found in ADHD,[5, 9] traumatic brain injury,[10, 11] hydrocephalus,[12, 13] myelomenigocele,[14, 15] and reading disabilities,[5, 16] and in some individuals with intellectual disability.
Because working memory and processing speed deficits may lower overall intellectual functioning but are not specific to the intellectual disability diagnostic criteria, the GAI may provide a more accurate estimate of reasoning ability in clinically referred children when low processing speed or working memory scores are suspected of reducing the FSIQ. The GAI may, therefore, better represent intellectual ability in these clinical groups, as their deficits may influence the speed of their output but not their ability to analyze and comprehend input. Harrison et al. compared FSIQ with GAI scores for the Wechsler Adult Intelligence Scale (third edition) among individuals with neurocognitive disorders (e.g. learning disabilities, ADHD), psychiatric diagnoses (e.g. depression, anxiety), and those with no diagnosis. They found a significant discrepancy between FSIQ and GAI scores in the neurocognitive group (with GAI greater than FSIQ), but not among the psychiatric disorders or no diagnosis groups. In addition, Bremmer et al. analyzed WISC-IV scores in a sample of clinically referred Canadian children, finding that in children diagnosed with either ADHD (inattentive subtype) or learning disabilities, GAI scores were significantly higher than a composite made up of the total of processing speed and working memory scores.
The purpose of the present study was to investigate the utility of the GAI, compared with the FSIQ, in identifying children with intellectual disability. Few existing studies compare FSIQ and GAI scores or examine the implication of potential GAI or FSIQ discrepancies in clinical groups. Appropriate identification of intellectual disability is essential for the determination of educational support and services, allocation of social services and funding, and even criminal proceedings. Given previously demonstrated differences between FSIQ and GAI scores for children and adults with various neurocognitive impairments, we hypothesized that GAI scores would be significantly higher than FSIQ scores in our sample. Furthermore, we expected that fewer children would meet criteria for intellectual impairment when using GAI as a measure of intellectual functioning compared with when using FSIQ. Given the positive correlation between adaptive functioning and IQ scores, we further hypothesized that low FSIQ and GAI scores would predict whether individuals would fall below the impairment threshold on adaptive functioning scores with acceptable sensitivity, specificity, and positive or negative predictive values. In addition, we anticipated that individuals in our sample with a GAI score greater than 70 but an FSIQ score of 70 or less would have higher adaptive functioning scores than those individuals with both FSIQ and GAI in the impaired range.
- Top of page
Full details regarding sample demographics and key variables, by group (‘both impaired’ vs ‘discrepant’) as well as for the total sample are shown in Table 1. All participants included in the sample had complete WISC-IV and ABAS-II data. The mean FSIQ score for the total sample was in the low–average range. Within the entire sample, FSIQ and GAI were significantly correlated (r=0.957, p<0.001). ABAS-II GAC scores were also significantly correlated with both FSIQ (r=0.398) and GAI (r=0.367) at the 0.001 level. No significant differences in age or sex between the ‘both impaired’ and ‘discrepant’ groups were observed (Table 1). There were significant differences between FSIQ and GAI (GAI>FSIQ) in the total sample (t=−19.14, p<0.001, Cohen's d=−0.20), and in the ‘discrepant’ group (t=−12.62, p<0.001, Cohen's d=−0.255), but not in the ‘both impaired’ group (t=−1.68, p=0.095, Cohen's d=−0.08).
Table 1. Mean FSIQ, GAI, and GAC scores for the total sample, ‘both impaired’, and ‘discrepant’ groups
| ||Total sample (SD)||‘Both impaired’ group (SD)||‘Discrepant’ group (SD)|| t || p ||Cohen's d|
| n ||833||152||44||–||–||–|
|Age (y)||10.4 (2.74)||11.42 (2.72)||11.17 (2.85)||0.518||0.605||0.09|
|FSIQ||84.08 (18.07)||57.23 (7.59)||67.00 (3.33)||−8.30||<0.001||−1.67|
|GAI||87.77 (19.14)||57.88 (8.97)||75.75 (3.54)||−12.91||<0.001||−2.62|
|ABAS GAC||74.63 (17.17)||63.61 (14.82)||70.32 (15.55)||−2.61||0.01||−0.44|
|ABAS conceptual||77.87 (14.46)||66.97 (13.61)||72.75 (13.54)||−2.49||0.014||−0.43|
|ABAS social||81.52 (17.45)||72.55 (15.04)||80.23 (16.0)||−2.94||0.004||−0.50|
|ABAS practical||75.11 (19.11)||63.95 (18.55)||72.07 (19.36)||−2.53||0.012||−0.43|
Table 2 presents the number of participants meeting the criteria for intellectual disability based on different adaptive and intellectual criteria. More participants met criteria for intellectual impairment using the FSIQ than the GAI. Approximately one-quarter of our sample met the criteria for adaptive impairment with GAC scores of 70 or less. In accordance with DSM-IV criteria for intellectual disability, all children in the present sample with GAC scores of 70 or less exhibited at least two skill areas of adaptive impairment. Of those with adaptive impairment, more participants met criteria for intellectual disability using FSIQ (FSIQ ≤70 and GAC ≤70), compared with those with adaptive impairment who met criteria for intellectual disability using GAI. A total of 118 children (14%) met intellectual disability criteria with both IQ scores (both FSIQ and GAI ≤70, and GAC ≤70), while 26 children (3%) did not meet criteria for intellectual disability diagnosis when assessed with GAI rather than FSIQ.
Table 2. Participants meeting criteria for intellectual and/or adaptive impairment
| ||Total sample (%)||FSIQ ≤70 (%)||GAI ≤70 (%)|
| n ||833||196 (24)||159 (19)|
|GAC ≤70||406 (49)||144 (17)||152 (14)|
|One ABAS domain ≤70||460 (55)||152 (18)||127 (15)|
Descriptive data regarding adaptive functioning for the sample as a whole, the ‘both impaired’ group, and the ‘discrepant’ group are presented in Table 1. An independent samples t-test showed that the GAC scores were significantly higher in the ‘discrepant’ group relative to the ‘both impaired’ group. Similarly, each of the ABAS domain scores was significantly higher in the ‘discrepant’ group than the ‘both impaired’ group. The proposed DSM-5 criterion for adaptive impairment (requiring impairment in at least one domain rather than at least two impaired skill areas) was comparable to using the GAC criterion score of 70 or less (see Table 2). Classification of adaptive impairment using one impaired domain score increased rates only slightly among the entire sample.
When using IQ scores in the impaired range (≤70) to predict impaired adaptive functioning (GAC ≤70), FSIQ and GAI were comparable (p≥0.10 for all comparisons) in terms of sensitivity (GAI=0.30, FSIQ=0.35) and specificity (GAI=0.91, FSIQ=0.88). The same was true for the positive (GAI=0.75, FSIQ=0.73) and negative (GAI=0.58, FSIQ=0.59) predictive values. Using stata 12 to conduct roctab ROC analysis, we found both measures of intellectual functioning demonstrated comparable performance in detecting intellectual disability (Figs 2 and 3). The area under the curve was fairly consistent between FSIQ (0.69) and GAI (0.70) in terms of predicting adaptive impairment. For reference, an area under the curve approaching 1.0 suggests more accurate detection of intellectual disability, whereas an area under the curve closer to 0.5 is analogous to chance.
Figure 2. Receiver operating characteristic (ROC) curve examining Full Scale IQ relative to adaptive impairment, as determined by the Adaptive Behavior Assessment System, second edition, General Adaptive Composite scores.
Download figure to PowerPoint
Figure 3. Receiver operating characteristic (ROC) curve examining General Abilities Index relative to adaptive impairment, as determined by Adaptive Behavior Assessment System, second edition, General Adaptive Composite scores.
Download figure to PowerPoint
- Top of page
This study examined the use of WISC-IV GAI and FSIQ scores in identifying intellectual disability within a clinically referred sample. FSIQ and GAI scores were highly correlated in this clinical sample. It is not surprising that the correlation between FSIQ and GAI is greater than either of the correlations between the IQ estimates and the GAC, given the method variance among the IQ and GAC scores (performance-based testing vs parent report) and the fact that they represent two different constructs (intellectual functioning vs adaptive functioning). In contrast, FSIQ and GAI are representative of the same construct, share the same measurement methodology, and are calculated from many of the same scales.
In spite of accumulating evidence that processing speed and working memory may be differentially affected in clinical groups, little diagnostic value was added by the use of a more reasoning-based measure of intellectual functioning (i.e. the GAI) in quantifying intellectual disability. Moreover, data suggest that this practice may not be in the best interest of the individual with intellectual and adaptive skill concerns.
Use of GAI resulted in fewer children meeting criteria for intellectual disability. Examining the sample as whole, we found that GAI scores were significantly higher than FSIQ scores in this clinically referred sample of school-aged children. This finding is consistent with previous studies demonstrating that working memory and processing speed deficits are more common within clinically referred samples. While the clinical significance of this statistically significant difference in our sample is questionable (i.e. the average difference between the two scores was slightly more than three standard score points), the diagnostic implications of the GAI–FSIQ discrepancy were considerable, as fewer children met criteria using the GAI as opposed to FSIQ. When considering concurrent adaptive impairment (GAC ≤70) with intellectual functioning scores, again, a greater percentage of the sample met diagnostic criteria for intellectual disability when assessed with FSIQ as opposed to GAI. When tracking individuals for consistency of classification across intellectual indices used, we found that 26 children (3%) who would meet diagnostic criteria for intellectual disability using FSIQ and GAC no longer met the intellectual impairment criterion when using GAI and GAC.
Individuals who ‘fell out’ of a diagnosis of intellectual impairment by use of GAI (i.e. the ‘discrepant’ group) had significantly higher adaptive composite scores on average; however, as a group they still fell within the impaired range (mean GAC standard score=70.32). Moreover, the mean GAI score of the ‘discrepant’ group was a standard score of 75.75 (SD 3.54), which is typically considered the upper end of the margin of error for the intellectual impairment criteria of intellectual disability. As such, there is little evidence to suggest that this group's relative strength in reasoning abilities constituted ‘intact’ intelligence, nor did it appear to afford much in the way of support for the application of functional daily skills. Despite their relatively stronger reasoning skills, children in the ‘discrepant’ group are still likely to be vulnerable to functional impairment in adaptive skills without supports and/or accommodations. Furthermore, sensitivity, specificity, and predictive values were comparable for the two methods when examined against adaptive functioning, suggesting that a GAI score of above 70 is not more likely to be associated with higher adaptive scores. However, it is noteworthy that the low sensitivity for both GAI and FSIQ, individually, in predicting adaptive dysfunction is consistent with previous research showing a high incidence of adaptive dysfunction in clinically referred young people without intellectual deficits, potentially owing to dysfunction in a variety of other clinical domains. As such, referred young people with low intelligence are likely to demonstrate adaptive dysfunction, but similar levels of adaptive dysfunction are also common in young people with ADHD, affective disorders, and medically complex conditions.
The evidence here indicates that the use of GAI for intellectual disability diagnostic decision-making disqualifies a subgroup of children with significant intellectual and adaptive deficits. It calls into question the practice of relying on GAI scores in formulating a comprehensive clinical impression of a child's overall functioning, particularly within a clinically referred group. A more accurate and representative picture might be obtained from examining each index score independently to assess a child's particular strengths and weaknesses and integrating this information to guide rehabilitation and/or interventions. Demonstration of a GAI score greater than 70 does not necessarily translate to intact adaptive abilities. Many children with neurocognitive deficits who demonstrate generally intact reasoning abilities still require a great deal of assistance in daily functioning. Thus, it may be the case that working memory and processing speed, rather than being considered separate from ‘intelligence’, are actually vital components of the efficiency of overall intellectual functioning.
One limitation of the present study is that patient diagnosis was not available for analysis. Given preliminary evidence that there are indeed unique GAI–FSIQ relationships in various clinical groups, future research in this area could investigate the association between these indices and adaptive functioning within various diagnostic populations. Furthermore, it is important to note that the use of a standard score cut-off score of 70 or less means that some individuals with low measured intellectual ability (i.e. those with scores in the low 70s) were placed into the ‘not impaired’ group. Considering the standard error of measurement, those with scores in the low 70s may actually be better conceptualized as intellectually impaired.
In conclusion, the use of GAI rather than FSIQ in quantifying intellectual functioning for this clinical population only slightly reduced the incidence of intellectual disability, by placing less weight on deficit skill areas and subsequently raising intellectual functioning scores across the board. Children who demonstrate impaired intellectual functioning based on FSIQ but show relatively more intact GAI scores may not necessarily display intact adaptive abilities. Based on this information, we conclude that the GAI provides little additional value in terms of diagnostic accuracy for intellectual disability, predicting adaptive functioning, or determining more appropriate services for children who are neurocognitively compromised. While this proxy of intellectual ability is thought of as being less stringent, it does not appear to drastically alter the rates of disability against a general composite score in our sample.