Transcript Document

Brenda M. Stoesz1, Janine M. Montgomery1, Kevin P. Stoddart2,3, Lillian Burke3, & Mark Stokes4
1Social
Cognition Lab, Department of Psychology, University of Manitoba; 2University of Toronto; 3The Redpath Centre; 4Deakin University
Background
Preliminary Results
Knowledge of Asperger Syndrome (AS) in children has increased substantially in
recent years; yet, information on diagnosis in adults is limited. Adult assessment
may be difficult for various reasons: early developmental history may be
unobtainable; comorbid conditions may be present; and/or experienced and
knowledgeable clinicians may be unavailable. Tests to assist in diagnosing adult AS
exist, but they have limitations:
Because sample sizes were small for the HFA and ADHD groups, we collapsed the AS
and HFA groups and the TD and ADHD groups. The two groups were comparable in
terms of age and IQ.
1) many tests were created for use with children;
2) information on reliability and validity of test scores are often unavailable; and
3) test developers do not always obtain independent confirmation that their
participants actually have AS.
Thus, the usefulness of tests for identifying AS in adults remains questionable and
may result in inaccurate diagnoses1.
Objectives
In this preliminary report, we describe the results of the KADI, GADS, AQ, EQ, ASDI,
and the RAADS-R. Due to time restraints the AAA interview was not administered. In
Table 1, we provide evidence for the validity of each measure in the form of
Sensitivity (Se), Specificity (Sp), and Overall Correct Classification.
Table 1. Sensitivity, Specificity, and Overall Correct Classification.
Threshold
Se %
Sp %
Overall correct
classification
KADI (n = 22)
80
56
100
82
GADS (n = 16)
80
83
100
94
80; 30
21; 67
97; 82
74;78
ASDI (n = 47)
5
36
100
81
RAADS-R (n = 46)
65
77
76
76
AQ; EQ (n = 46; 47)
Given the above, the goals for our current study were to augment the existing
information for several tests of adult AS to determine:
1) how well these instruments discriminate between groups of people with ASD
(discriminant validity);
2) if the instruments are strongly related to other tests for AS/ASDs (convergent
validity); and
3) if the instruments differ from tests of other conditions (divergent validity).
Table 2. Means (SD) of test scores, comparisons between AS/HFA group scores and
TD/ADHD group scores using independent samples t-tests (2-tailed).
Further, we examined the accuracy of AS diagnoses when instruments are used
individually or in combination.
Method
In a pilot study, we recruited adults (aged 18+ years; n = 12) with AS and controls in
each of three groups:
1) high functioning autism (HFA; n = 2),
2) attention deficit hyperactivity disorder (ADHD; n = 5), and
3) typically developing (TD; n = 228).
These groups were chosen to evaluate the sensitivity of existing measures for
discriminating between similar (HFA), overlapping (ADHD), and distinct (typically
developing) groups. All clinical participants were diagnosed by a clinician prior to
their participation and had a verbal IQ ≥ 85.
Participants completed various AS assessment instruments:
1) Krug Asperger Disorder Index (KADI)2;
2) Gilliam Asperger’s Disorder Scale (GADS)3;
3) The Autism Quotient (AQ)4 and the Empathy Quotient (EQ)5 from the Adult
Asperger Assessment (AAA)6;
4) Asperger Syndrome Diagnostic Interview (ASDI)7; and
5) Ritvo Autism Asperger Diagnostic Scale- Revised (RAADS-R)8]; and
6) other tests included to measure convergent and divergent validity.
References:
1
Stoesz et al. (2011). The Clinical Neuropsychologist, 25, 376-401. 2 Krug & Arick
(2003). Krug’s Asperger’s Disorder Index: Examiner’s manual. Austin, TX: Pro-Ed. 3 Gilliam (2003). Gilliam
Asperger’s Disorder Scale: Examiner’s manual. Austin, TX: Pro-Ed. 4 Baron-Cohen et al. (2001). . JADD, 31, 517. 5 Baron-Cohen & Wheelwright (2004). JADD, 34, 163-175. 6 Baron-Cohen et al. (2005). JADD, 35, 807819. 7 Gillberg et al (2001). Autism, 5, 57-66. 8 Ritvo et al. (2011). JADD, 41, 1076-1089.
AS/HFA
TD/ADHD
KADI, raw scores
n=9
48.7 (31.1)
n = 13
2.5** (3.6)
GADS, raw scores
n=6
62.5 (13.1)
n = 10
7.1*** (9.7)
AQ
EQ
n = 14
n = 12
25.7 (9.4)
27.8 (8.9)
n = 33
15.6* (7.0)
39.5** (10.8)
ASDI
n = 14
3.6 (1.5)
n = 33
.12*** (.3)
RAADS-R
n = 13
114.1 (41.8)
n =33
46.4*** (28.0)
Canonical discriminant function coefficients reveal that ASDI, followed by GADS,
and then by RAADS-R contributed most to the model (see Table 3). This model
resulted in a classification table with perfect agreement between predicted and
observed. When the model was replicated using a within groups covariance
matrix, the cross validation (leave-one-out method) table was obtained, also
resulting in 100% agreement between predicted and observed cases.
Table 3. Canonical discriminant function coefficients, Structure coefficients,
Tolerance at step 3, and Unstandardized function coefficients for each group for
each of the three retained variables.
Standardized
discriminant
function
coefficients
Unstandardized
discriminant
function
coefficients
ASDI
1.63
GADS
RAADS
TD Function
Coefficient
AS Function
Coefficient
Structure
matrix
Tolerance
step 3
1.49
0.33
0.28
-1.38
13.05
1.47
.13
0.50
0.41
-0.04
1.21
-1.09
-.03
0.25
0.32
0.07
-0.26
-1.52
-46.67
Constant
-3.66
Discussion
Findings are preliminary and must be interpreted with caution given the small
sample sizes for some of the groups and measures, however, they do suggest that
some instruments (i.e., GADS, RAADS-R) are better at correctly identifying and
discriminating amongst subgroups and that combining selected instruments (i.e.,
GADS, ASDI, and RAADS-R) improves classification rates. As such, data collection in
our laboratory is ongoing to determine whether these findings will be observed in a
larger sample.
*p < .05, **p < .01 , ***p < .001
Conclusions
An exploratory hierarchical discriminant function analysis was undertaken on
diagnosis (TD vs. AS) over the predictors of AQ, EQ, ASDI, GADS, KADI, and RAADS-R.
Statistical significance criteria for F were set to allow variables to enter (p < 0.05) or
be removed (p > 0.10) from the model. Prior probabilities were computed from
group sizes, and the covariance matrix was based upon separate groups. Covariance
matrices were assessed for equality by Box’s M (1.82), which was not significant
(F(1,438.724) = 1.68, p = 0.19).
The model resolved at three steps, with the canonical covariate significantly
accounting for 96.2% of variance (λ = 25.91; Wilk’s λ = 0.04, 2(3) = 37.87, p < 0.001;
functions at group centroids: TD = 3.87, AS = 5.80).
Families and individuals affected by a late diagnosis of AS experience frustrations
and difficulties. Many clinicians feel inadequately prepared to assess adults on the
spectrum, as resources for late diagnosis are limited. Thus, it is essential that
clinicians feel competent to address the unique diagnostic needs of adults
suspected of having AS.
The results from our study provide a step towards improving this situation at a
clinical level, and we anticipate that this will likewise positively impact individuals
and families by enabling accurate diagnosis and, thus, access to appropriate
supports and treatment. Our findings are essential to informing understanding of
AS in adults and improving clinical assessment and diagnosis of AS.
Acknowledgments:
This research was supported by two University of Manitoba grants: (1) Kick-Start: A Program to Support Research
and Collaboration in Psychology; and (2) the University Research Grants Program. We thank the individuals and
their families for their participation in this research. We thank the members of the Social Cognition Laboratory –
Kelly Carpick, Katelin Neufeld, Stephany Huynh, Brigitte Jeanson, Daniel Buchman, Melanie Caister, Haidee Baker,
Deepak Singh, Leah Funk, and Erica Holiday -- for their help with data collection.
Correspondence: [email protected] or [email protected]