Another Look at the Social Communication Questionnaire and its Relationship to the Autism Diagnostic

1
Another Look at the Social Communication
Questionnaire and its Relationship to the Autism
Diagnostic InterviewC. Corsello, D. Anderson, S.
Qui, S. Risi, C. Lord University of Michigan
Autism and Communication Disorders Center
In order to examine a possible interaction
between age and IQ, a one way ANOVA, with age and
IQ group as the independent variable (no mental
retardation and younger than 7, mental
retardation and younger than 7, no mental
retardation and older than 7, no mental
retardation and older than 7) and SCQ scores were
run for each diagnostic group (nonspectrum,
pdd.nos, autism). Tukey Post Hoc comparisons were
used to analyze specific group differences. In
the non-spectrum group, the SCQ scores of the
younger group without mental retardation ( M
10.16, SD 5.84) were significantly lower than
both older groups, with mental retardation (M
18.08, SD 6.78) and without mental retardation (M
14.30, SD 7.84), F (3, 65) 3.63, p lt .02,
with. There was also a significant difference for
the autistic group, F (3, 168) 14.46, p lt
.001, with the younger group without mental
retardation (M 18.58, SD 6.23) again scoring
lower than both the older group with mental
retardation (M 25.09, SD 5.83) and the older
group without mental retardation (M 22.56, SD
56). There was not a significant difference for
age and IQ groups for the pdd.nos diagnostic
group, F (3, 80) 2.35, p .079 (See Figure 1).
Abstract
Participants
Method
The study included 327 participants, 174 had
a diagnosis of autism, 84 had a diagnosis of
pervasive developmental disorder, not otherwise
specified, and 69 had non-spectrum diagnoses
(e.g. receptive - expressive language disorder,
obsessive compulsive disorder, mental
retardation, depression). Of this group, 292 were
included in a previous study presented at SRCD.
Demographic information was available for each of
the participants (see Table 1). There was no
significant difference in SCQ scores between the
SCQ first and ADI first group for Race, F (5,
321) .937, p .457 or Maternal Education
(college, some college, college degree or more),
F (2, 310) .099, p .906. Participants who
completed the ADI-R prior to the SCQ were part of
a longitudinal study and were originally referred
for possible autism at the age of 2 years, or in
a control group of young children with other
developmental disabilities. These children were
included in the initial validation sample.
Participants who completed the SCQ prior to the
ADI R were consecutive referrals to diagnostic
clinics at the Universities of Chicago and
Michigan. All participants received a best
estimate diagnosis, usually by a child
psychiatrist and a clinical psychologist who
administered the ADI-R and the ADOS as part of
the evaluation (See Table 2).
In the clinic - referred sample, parents
completed the SCQ for their child prior to the
diagnostic evaluation as part of a pre-evaluation
packet. For the longitudinal sample, the parents
completed the SCQ following their childs
diagnostic evaluation, in a packet of information
sent to them as part of the initial validation
study of the SCQ. Only SCQs resulting in a
diagnostic classification were included in the
analyses. SCQs with incomplete information (items
that were not answered) were omitted if including
the missing items would have changed the
diagnostic classification. An ASD classification
on the SCQ was based on the recommended cut-off
of 15. To establish a diagnosis, each
participant underwent an extensive diagnostic
evaluation that included the ADOS, cognitive or
developmental assessment, and the ADI-R.
The Social Communication Questionnaire (SCQ),
formerly the Autism Screening Questionnaire
(ASQ), is a screening instrument for Autistic
Spectrum Disorders (ASDs), completed by a parent.
An initial study reported good sensitivity and
specificity of the SCQ in an older sample (ASD
mean age 23 years), with whom the Autism
Diagnostic Interview Revised (ADI-R) was
administered prior to the SCQ. More recent
studies, with younger children and completion of
the SCQ prior to the ADI-R have been less
encouraging. Some studies have replaced the ADI-R
with the SCQ, for determining eligibility for
research studies, reporting good agreement
between the measures. The present study examined
the effectiveness of the SCQ as a screening
instrument in two groups, one of which had the
ADI-R prior the SCQ (N108) and the other of
which had the SCQ prior to the ADI-R (N 219),
and compared the ADI-R with the SCQ. Findings
indicated that sensitivity was higher and
specificity was lower for the group that had the
ADI-R prior to the SCQ (sensitivity .80
specificity .60) than for the clinic referred
group that had the SCQ prior to the ADI-R
(sensitivity .66 specificity .64). The SCQ
first group had significantly higher IQs and was
significantly younger than the ADI-R first group.
Effects of age and IQ were further examined by
comparing sensitivity and specificity in a group
over 7 years and under 7 years of age and a group
with and without mental retardation. Sensitivity
was lower for a younger age group (.63) than for
an older age group (.79) and specificity was
lower for the older age group (.51) than for the
younger group (.74). Presence of mental
retardation also affected SCQ scores, with better
sensitivity in the group with mental retardation
(.79) than in a group without mental retardation
(.62). Specificity was much lower in the group
with mental retardation (.38) than the group
without mental retardation (.73). Results will be
discussed in terms of theoretical and practical
implications for screening for research and
clinical purposes.
Results
Chi square analyses were run to determine
whether the SCQ differentiated between the ASD
and nonspectrum groups, both when the ADI-R was
administered first and when the SCQ was
administered first. Results indicated that the
SCQ differentiated between the diagnostic groups
in each of the conditions when the ADI-R was
administered first, ?2 (1, N 108) 14.39,
plt.001 and when the SCQ was administered first,
?2 (1, N 219) 13.15, plt.001. Sensitivity
and specificity were calculated in order to
determine how accurately the SCQ classified
individuals with ASDs under both conditions. When
the SCQ was administered after the ADI-R, the
sensitivity was higher than when the SCQ was
administered first. The specificity was slightly
lower when the ADI-R was administered first than
when the SCQ was administered first (See Table
3). Sensitivity and specificity were also
calculated for the ADI-R under both conditions
Again, the sensitivity and specificity were
higher when the ADI-R was administered first than
when the SCQ was administered first (see Table
3). Given that both instruments were more
sensitive in the group for which the ADI-R was
administered first, age and IQ differences
between the groups were explored. A one way ANOVA
with order of measure (ADI-R first, SCQ first) as
the independent variable and age as the dependent
variable, resulted in a significant difference in
age in which the SCQ first group was younger (M
80.27, SD 26.85) than the ADI-R first group (M
125.83, SD 36.78), F (1, 325) 55.43, p
lt.001. One way ANOVAs on Verbal and Nonverbal
IQ scores were also run, with order of measure as
the independent variable. Verbal IQ was
significantly higher for the SCQ first group (M
71.20, SD 31.83) than for the ADI-R first group
(M 55.61, SD 33.08), F (1, 325) 16.89, plt
.001. Nonverbal IQ was also higher for the SCQ
first group (M 80.27, SD 26.85) than for the
ADI-R first group (M 68.25, SD 31.27), F (1,
325) 16.89, p lt.001. In order to determine if
the order of instrument accounted for the
differences in sensitivity associated with age
and IQ, an ANOVA with order of instrument as the
independent variable and SCQ score as the
dependent variable was run, covarying age and
verbal IQ score. The effect of instrument was no
longer significant when age and IQ were covaried,
F (1, 326) .304, p .582. There was a
significant effect of age F (1, 326) 18.38, p lt
.001 and of verbal IQ, F (1, 326) 60.75, p lt
.001. The effects of age and IQ were further
investigated by re-examining sensitivity and
specificity based on age groups and IQ groups.
Since the median age of individuals in this
sample was 82 months, age groups were divided
into younger or older than 7, with 7 year old in
the under 7 group. Sensitivity and specificity
were calculated for the under 7 group and the
older 7 group, and revealed higher sensitivity
but lower specificity in the older age group than
the younger age group (See Table 4).
Demographics by Clinical Diagnosis and Order of
Diagnostic Instrument
To examine the effect of age and IQ on types of
diagnostic errors, a 2 x 2 ANOVA, with status
(false positive, false negative) and presence of
mental retardation (no mental retardation, mental
retardation) as the independent variables and age
as the dependent variable was run. There was a
significant effect of status, with false
negatives (M 87.16, SD 50.21) significantly
younger than false positives (M 123.38, SD
49.13), F (1, 97) 10.26, plt.01 (See Figure 2).
IQ, SCQ score, and ADI Score by Clinical
Diagnosis and Order of Diagnostic Instrument
Introduction
Over the past two decades, there has been
significant progress in the development of
reliable and valid standardized instruments for
diagnosing autism and Autistic Spectrum Disorders
(ASDs) (Lord, Rutter, DiLavore, Risi, 1999
Lord, Rutter, LeCouteur, 1994 LeCouteur, Lord,
Rutter, 2003), which has led to consistency in
classification of ASDs and better identification
of children at the extremes of the spectrum,
including children as young as 24-30 months
(Lord, 1995). As it has become possible to
diagnose very young children with autism, there
has become an increasing need for reliable and
valid screening instruments. Several such
instruments have been developed for children with
autism, but most of them have limitations
including high false negative rates (Baird, et.
al., 2000). The first published study on the
Social Communication Questionnaire (SCQ),
formerly the Autism Screening Questionnaire
(ASQ), reported good sensitivity (.85) and
specificity (.75) in a sample of 200 (160 with
autistic spectrum disorders and 40 language
impaired developmentally delayed controls), using
15 as the cutoff for to differentiate ASDs from
children with nonspectrum diagnoses (Berument,
Rutter, Lord, Pickles, Bailey, 1999). While
these numbers were encouraging, the parents had
been interviewed with the Autism Diagnostic
Interview Revised (ADI-R) and the participants
had been diagnosed prior to the completion of the
SCQ. The sample was also older than a group that
is typically screened, with a mean age of 23
years in the autistic sample, 7 years in the
atypical autistic sample, and 17 years in the
Aspergers Sample (Berument, et. al., 1999).
Bishop Norbury (2002) also found that the SCQ
was able to classify children with an ASD in a
school age sample of children with speech and
language impairments. These children had received
diagnoses through the school system prior to
completion of the SCQ. More recent studies with
younger children have been less encouraging
(Hanson, et.al., 2002 Corsello, Leventhal, Cook,
2003). In a recent pilot study (Hanson, et. al.,
2002) the SCQ was administered to 42 children
between the ages of 19 months and 4.5 years of
age (M 3.2 years) prior to diagnosis and the
administration of the ADI R, using the cut-off
of 15 for ASD. In this study, the sensitivity
(.68) and specificity (.41) of the instrument
were not as high as originally reported. The
sensitivity was increased somewhat when a cut off
of 14 was used (.79), with specificity remaining
the same. It is unclear whether the younger age
of the children in this sample accounted for the
lower sensitivity and specificity. However,
another recent study of a larger clinic referred
sample also reported low sensitivity and high
specificity for both children gt 5 years
(sensitivity .61, specificity .81) and 5
years (sensitivity .62, specificity .89)
(Corsello, et. al., 2003). These results seem to
suggest that age differences within childhood may
not be the only factor leading to the lower
sensitivity of the SCQ. Another possible reason
for the difference in sensitivity in more recent
studies, is that the SCQ was administered prior
to the ADI-R. In the initial study, the ADI-R was
administered prior to the SCQ and could have led
to higher sensitivity in that sample. The current
study examines whether or not the order in which
these parent report instruments are administered,
as well as age and IQ, affects the sensitivity
and specificity of the SCQ.
Conclusions
The sensitivity of the SCQ was lower in this
sample than the initial study, and mean scores
continue to be lower than in the initial
study. The sensitivity and was better for a
group of children who had been referred for
autism at the age of 2 years than for a more
heterogeneous clinical sample, most of whom were
referred at later ages. In the longitudinal
sample, the ADI-R was administered prior to the
SCQ, and the children had a lower IQ scores and
were older than the clinic referred sample. The
effect of order of administration was not
significant, when controlling for age and IQ.
There was an effect of age on SCQ scores, in
which children who were younger had lower scores
on the SCQ, resulting in lower sensitivity in
younger age groups. Children who were older and
had a lower IQ score were more likely to be
incorrectly classified as ASD on the SCQ. The SCQ
had higher sensitivity in older age groups (above
the age of 7 years) but lower specificity. Item
analysis is needed to determine if certain items
may be better at accurately classifying young
children with an ASD at younger ages. Different
cut off scores should also be explored for
different age groups, as some studies have found
that lowering the cut off scores resulted in
greater sensitivity.
Table 3. Sensitivity and Specificity of ADI First
and SCQ First
Measures
The SCQ is a screening instrument based on
questions included in the ADI-R, a parent
interview with demonstrated good reliability and
validity (Rutter, et. al., 2003). It covers the
areas of communication, reciprocal social
interactions, and restricted and repetitive
behaviors and interests. The SCQ is a checklist
consisting of 40 questions, covering the areas of
communication, reciprocal social interactions,
and restricted and repetitive behaviors and
interests. It is designed to be completed by a
parent or someone who knows the child well. Each
item is checked as yes or no, and assigned a
point rating of 1 (presence of abnormal
behavior) or 0 (absence of abnormal behavior).
The first item is not included in the scoring, as
its purpose is to determine if the child has
sufficient language to score items concerning
abnormalities in language. If the child is not
verbal, the next six items are skipped. The
points are summed and result in a total possible
score of 0 39 for verbal children and 0 33
for nonverbal children. The totals are compared
to a cut off of 15 for ASD and 22 for
autism. A variety of cognitive and developmental
measures were used depending on the age and
developmental level of the child or adult. These
most often included the Mullen Scales of Early
Learning (Mullen, 1995), the Differential Ability
Scales, preschool and school age edition (Elliot,
1990), and less frequently the Wechsler Scales of
Intelligence (Wechsler 1984, 1989, 1991),
Ravens Colored Progressive Matrices Form Board
(Raven, 1986), the Peabody Picture Vocabulary
Test (Dunn, 1997), the Vineland Adaptive Behavior
Scales (Sparrow, Balla, Cichetti, 1984), or the
Reynell Developmental Language Scales (Edwards,
et. al., 1997).
Table 4 Sensitivity and Specificity in Under 7,
Over 7, MR, and no MR Groups
Contact Christina Corsello, Ph.D.,
ccorsell_at_umich.edu University of Michigan Autism
and Communication Disorders Center