816065

research-article2018
AUT0010.1177/1362361318816065AutismAldosari et al.

Original Article

Autism

Validation of the Arabic version of the 1­–8

© The Author(s) 2019
Article reuse guidelines:
Social Communication Questionnaire sagepub.com/journals-permissions
DOI: 10.1177/1362361318816065
https://doi.org/10.1177/1362361318816065
journals.sagepub.com/home/aut

Mohammed Aldosari1, Eric Fombonne2, Hesham Aldhalaan3,

Mohammed Ouda4, Saba Elhag5, Hawraa Alshammari5,
Iman Ghazal5, Asma Alsaleh3, Tala Alqadoumi6, Richard Thomson5,
Mohanad Al Khasawneh7, Mohamed Tolefat8 and Fouad Alshaban5

Abstract
Validated screening and diagnostic tools for autism spectrum disorder for use in Arabic-speaking individuals are scarce.
This study validated the Arabic version of the Social Communication Questionnaire. The total study sample included
206 children with autism spectrum disorder and 206 typically developing children (73.8% male; mean age: 8.5 (standard
deviation = 2.6) years). The mean Social Communication Questionnaire total score was significantly higher in autism
spectrum disorder children than in typically developing children (p < 0.0001). Scores on the three Social Communication
Questionnaire subscales also differed significantly between the groups (p < 0.001). Of the 39 items, 37 were endorsed
significantly more often in the autism spectrum disorder group. The total Social Communication Questionnaire score
did not vary by age or gender. Internal consistency was excellent (alpha = 0.92). In the receiver operating characteristic
analysis, the area under the curve for the total score showed excellent discrimination between autism spectrum disorder
and typically developing children (area under the curve = 0.95; 95% confidence interval: 0.93–0.97). The areas under the
curve for the scale subscores were 0.923 (95% confidence interval: 0.898–0.949) for the social interaction score, 0.872
(95% confidence interval: 0.838–0.905) for the communication score, and 0.856 (95% confidence interval: 0.819–0.893)
for the repetitive behaviors score. The findings support the use of the Arabic Social Communication Questionnaire to
successfully differentiate children with clinically diagnosed autism spectrum disorder using the established cutoff value
for the English version.

Keywords
Arabic, autism spectrum disorder, cutoff values, early intervention, epidemiology, screening, Social Communication
Questionnaire, validity

Introduction the use of validated screening tools (Allen et al., 2007).

However, limited access to cross-culturally validated
Autism spectrum disorder (ASD) is a neurodevelopmen-
tal disorder that is characterized by impairments in com-
1Cleveland Clinic Foundation, USA
munication and social interaction, repetitive behaviors, 2Oregon Health & Science University, USA
and limited areas of interest that manifest in the first 3Center for Autism Research, Saudi Arabia
3 years of life (American Psychiatric Association (APA), 4National Guard Health Affairs, Saudi Arabia

2013). Early diagnosis and intervention have been shown 5Neurological Disorders Research Center, Qatar Biomedical Research

to significantly improve cognitive and adaptive behavior Institute, Hamad Bin Khalifa University, Qatar
6King Saud University, Saudi Arabia
and reduce the severity of ASD (Dawson et al., 2010). In 7Qatar University, Qatar
addition, early clinical intervention remarkably decreases 8Shafallah Center for Children with Disabilities, Qatar
the financial burden of ASD, and the estimated cost sav-
ings have been shown to outweigh the costs of early Corresponding author:
Fouad A Alshaban, Neurological Disorders Research Center, Qatar
intensive behavioral intervention programs (Peters- Biomedical Research Institute, Hamad Bin Khalifa University, P.O. Box
Scheffer et al., 2012). Research has shown that early 34110 Qatar.
detection and ensuing intervention can be achieved with Email: [email protected]
2 Autism 00(0)

screening and diagnostic tools for ASD poses major chal- instructions for the parents to complete it. No individual
lenges to clinicians and researchers worldwide. Moreover, data were collected on caregiver socioeconomic status
there are few validated screening tools for Arabic- (SES) or education level but recruitment sites provide ser-
speaking individuals (Mohamed et al., 2016; Seif Eldin vices to a diverse population in each country.
et al., 2008).
Before initiating a country-wide autism epidemiological
Data collection tools
study, we needed to translate a screening tool and evaluate
the overall screening properties of that tool within the local SCQ. The SCQ is a parent-report questionnaire that evalu-
population. We also wanted to determine which cutoff val- ates three major aspects of ASDs: communication, social
ues were associated with optimal values for specificity and interaction, and repetitive behaviors. The SCQ aids in
sensitivity in order to conduct a large, population-based identifying patients who require further evaluation for
epidemiological study of ASD. ASD. The development of the SCQ was modeled after the
Autism Diagnostic Interview to generate a brief, parent-
completed, screening tool (Berument et al., 1999). The
Materials and methods questionnaire exists in two forms: lifetime and current.
The “lifetime” form evaluates the patient’s developmental
Participants history as well as current behaviors, whereas the “current”
Children were recruited from two neighboring Arabian form assesses the child’s behavior during the past 3-month
Gulf countries, Qatar and Saudi Arabia. Expatriate resi- period only. It is conveniently brief and relatively inexpen-
dents were excluded when their primary language was not sive with 40 questions per form with “yes” or “no”
Arabic. The total study sample comprised 412 children: responses that can be answered in less than 10 min. Each
206 children with an established clinical diagnosis of ASD item is scored as 0 or 1, and the sum of 39 items yields a
and 206 gender-matched typically developing (TD) chil- total SCQ score ranging from 0 to 39. (Question no. 1 doc-
dren attending mainstream primary schools. uments whether or not the child has phrase speech and
The sample with ASD was recruited from 10 autism does not have any scoring value.) On the basis of the origi-
centers and special education schools in Saudi Arabia nal validation study based on a clinical sample, cutoffs of
(N = 93) and from the main autism center in Qatar (Shafallah 15 and 22 have been proposed to select children likely to
Center for Children with Special Needs; N = 113). To be have a broader or narrower form of ASD (Berument et al.,
included in the ASD sample, children had to be aged 1999). In subsequent epidemiologic studies, a cutoff of 12
between 5 and 12 years and have a clinically confirmed has been proposed to optimize SCQ performance in popu-
diagnosis of ASD obtained through a multidisciplinary lation-based samples (Eaves et al., 2006).
evaluation. All subjects met the full criteria for ASD as
defined by the Diagnostic and Statistical Manual of Mental Arabic version of the SCQ. For the purpose of this study, the
Disorders (5th ed.; DSM-5; APA, 2013). Diagnosis was English version of the SCQ was translated into Arabic by
established by experienced clinicians by combining all the study authors (M.S.A. and F.A.A.). Following the initial
developmental history, clinical observations, and examina- translation, each item was reviewed and culturally adapted
tions and using established diagnostic tools, including the to minimize barriers to comprehension and improve rates of
Childhood Autism Rating Scale, the Autism Diagnostic completion by the study participants. One of the few changes
Interview—Revised (ADI-R), and the Autism Diagnostic made to the original instrument was to remove references to
Observation Schedule. Children with incomplete clinical the British rhymes “The Mulberry Bush” and “London
information were excluded. The ASD sample was recruited Bridge is Falling Down” as the examples used for item 34.
by distributing the Social Communication Questionnaire The two authors worked with the publisher (Western Psy-
(SCQ—Lifetime version) forms to the centers directly by chological Services) to have an SCQ author-assigned
the study’s researchers. reviewer revise the back-translation. Following multiple
The TD group was recruited from a total of 20 primary revision cycles, the final version of the Arabic SCQ was
schools in Qatar (N = 120) and Saudi Arabia (N = 86). To be approved by the study’s authors and the publisher.
included, children were aged 5–12 years, were enrolled in
regular classrooms, and had no developmental, behavioral,
Statistical methods
or academic concerns by parental and teacher report. We
excluded children with learning disabilities. TD children SCQ data were entered into an Excel (Microsoft) spread-
were matched by gender with the ASD children on a 1:1 sheet as raw scores. The statistical analysis was done in SPSS
ratio. (IBM Corp.) by one of the authors (E.F.). Conventional sta-
In Qatar, the comparison sample was recruited by mail- tistical tests (Student’s t-tests, analyses of variance
ing the SCQ forms to the families. The Saudi sample was (ANOVAs), chi-square test, and Fisher’s exact test) were
recruited by distributing forms to the schools. In both performed to compare continuous and categorical variables.
instances, the SCQ lifetime version was sent alongside Internal consistency was measured with Cronbach’s alpha
Aldosari et al. 3

Table 1. Sample characteristics (n = 412).

ASD children TD children p-Valuesa

(n = 206) (n = 206)
Mean age, years (SD) 8.43 (2.6) 8.5 (2.7) NS
Age group, years (n)
5–6 66 63
7–8 40 43
9–10 59 57 NS
⩾11 41 43
Site (n)
Qatar 113 120
Saudi Arabia 93 86 NS
Gender male, n (%) 152 (73.8) 152 (73.8) NS
SCQ score, mean (SD)
Total score 20.2 (6.7) 6.4 (4.1) <0.0001
Social interaction subscore 8.4 (3.9) 1.9 (2.0) <0.0001
Communication subscore 6.5 (2.2) 3.0 (1.9) <0.0001
Repetitive behaviors subscore 4.1 (2.0) 1.2 (1.7) <0.0001

ASD: autism spectrum disorder; TD: typically developing; SD: standard deviation; SCQ: Social Communication Questionnaire.
aChi-square test or Fisher’s exact test for categorical variables; t-tests for continuous variables.

coefficient. Receiver operating characteristic (ROC) analysis total SCQ scores was highly significant (p < 0.001), with
was performed to examine the overall performance of the a mean difference of 13.8 points between the two groups
SCQ and to estimate sensitivity and specificity for different (Table 1). The three subscales of SCQ scores also dif-
cut points (Fombonne, 1991). Throughout, 0.05 was retained fered significantly between the two groups (all p-values
as the level for statistical significance. < 0.0001). The corresponding effect sizes for these dif-
ferences as measured by Cohen’s d were all very large:
for the SCQ total score, 2.5; for the social interaction
Ethical approval
score, 2.1; for the communication score, 1.6; and for the
The research design and methods were approved by both repetitive behaviors score, 1.5.
the Qatar Biomedical Research Institute and the Cleveland
Clinic institutional review boards. Age and gender effects
Separate ANOVAs were performed on the four SCQ scores
Results as dependent variables with gender and age as two- and
four-level independent factors. For the total SCQ score,
Participants
there was no effect of gender (p = 0.18) or age group
Sample characteristics of the study participants are sum- (p = 0.70) and no significant interaction between gender
marized in Table 1. The study sample included 412 chil- and age group (p = 0.67). For the social interaction sub-
dren (206 ASD and 206 TD). The majority of participants score, there was no significant effect of gender (p = 0.59)
(56.6%) were from Qatar. There was an overall boy-to-girl or age group (p = 0.95) or their interaction (p = 0.87). For
ratio of 2.8:1 (73.8% male). Male over-representation was the communication score, there was no interaction effect
similar in the groups because of matching on gender when (p = 0.10) and no significant effect of age group (p = 0.09);
selecting the two groups. The mean age of the sample was however, gender was significant (p = 0.008), with girls
8.46 (standard deviation (SD) = 2.65) years, with no sig- showing fewer impairments than boys (4.13 vs 4.96). The
nificant difference between the ASD group and the com- difference was more pronounced for girls aged 7–8 years
parison group when age was treated either as a continuous and 9–10 years. Finally, for the repetitive behaviors score,
or as a categorical variable (see Table 1). no significant effect for gender (p = 0.21), age group
(p = 0.70), or their interaction (p = 0.90) was detected.
SCQ scores in ASD and TD children
Internal consistency
The distribution of total SCQ scores in the ASD and TD
groups is shown in Figure 1. As expected, variability was We measured the reliability of the SCQ using Cronbach’s
somewhat larger in the children with ASD than in the TD alpha coefficient. All measures of internal consistency
children as illustrated by the SDs. The difference between were high to satisfactory. For the total SCQ score, alpha
4 Autism 00(0)

Figure 1. Total score on the Social Communication Questionnaire (SCQ) in the children with autism spectrum disorder (ASD)
and the typically developing comparison group.

was 0.916 when we used all 39 items of the SCQ (after higher frequencies among children with ASD than in the
exclusion of item 1). Because 6 items (items 2–7) are comparison group. The two items that did not discrimi-
not applicable to nonverbal subjects, we re-estimated nate between the two groups were item 4 (communica-
the reliability coefficient for the 33 items (items 8–40) tion, ever coding; inappropriate questions or statements;
that applied to all subjects. The corresponding value for OR = 1.3: NS) and item 13 (repetitive behaviors, ever
alpha was 0.929. Internal consistency estimates were as coding; unusually intense special interests; OR = 1.2:
follows for the three SCQ subscales: 0.901 for the social NS). For the remaining 37 items, OR point estimates
interaction subscale (15 items; n = 401), 0.708 for the ranged from a low of 0.6 (item 23; communication, age:
communication subscale (13 items; n = 293), and 0.818 4–5 years; use of gestures) to a high of 83.5 (item 40;
for the repetitive behaviors subscale (8 items; n = 290). social interaction, age: 4–5 years; group or cooperative
Fewer subjects were available for the latter two analy- play with peers). Abnormal scores for item 23 were
ses as values were missing for nonverbal subjects for endorsed significantly more often by TD children than
six items (one in the repetitive behaviors subscales by children with ASD (56.6% vs 45.6%, respectively;
(item 7) and five in the communication subscale (items χ2 = 4.9, df = 1; p < 0.05).
2–6)). For the 13 items on the communication subscale (1
for the current period, 4 for the ever period, and 8 for
the age 4–5 years period), ORs ranged from 0.6 (item
Item discriminant ability 23; use of gestures; age: 4–5 years) to 23.2 (item 2; cur-
We evaluated the discriminant ability of each item by rent period; able to have to and fro “conversation”),
comparing their frequency in ASD and TD children with a median OR of 6.4 (item 5; ever period; pronouns
(Table 2). We calculated odds ratios (ORs) to estimate mixed up). For the 15 items on the social interaction
the magnitude of the association with case-control sta- subscale (1 current, 2 ever, and 12 age 4–5 years), ORs
tus. Of 39 comparisons, all but 2 were significant, indi- ranged from 1.8 (item 32; coordinated requesting; age:
cating that 37 of 39 items of the SCQ had significantly 4–5 years) to 83.5 (item 40; age: 4–5 years; group play
Aldosari et al. 5

Table 2. Discriminant validity of the SCQ items (n = 412).

Item no. Period Item label Domain OR p-Values, Frequency (%)

chi-square
TD ASD
Children Children
Items 1–7 only applicable to verbal childrena
2b Current Conversation C 23.2 82.9*** 3.7 47.3
3b Ever Stereotyped utterances C 8.3 64.3*** 30.2 78.2
4b Ever Inappropriate questions C 1.3 0.55 NS 14.2 17.4
5b Ever Pronoun reversal C 6.4 51.9*** 18.4 59.1
6b Ever Neologisms C 2.4 11.22*** 22.6 40.9
7b Ever Verbal rituals R 11.5 80.2**** 12.8 63.0
Items 8–40 applicable to all children
8 Ever Compulsions and rituals R 7.3 79.8**** 12.7 51.7
9 Ever Facial expression S 7.1 49.6*** 7.3 35.9
10 Ever Use of other’s body to communicate S 15.4 143.1*** 16.5 75.2
11 Ever Unusual preoccupations R 5.3 39.1*** 9.2 34.8
12 Ever Repetitive use of objects R 6.7 73.2*** 17.0 57.8
13 Ever Circumscribed interests R 1.2 0.7 NS 29.1 33.0
14 Ever Unusual sensory interests R 7.5 83.2*** 17.5 61.5
15 Ever Hand and finger mannerisms R 30.1 167*** 6.8 68.8
16 Ever Complex body mannerisms R 8.8 97.4*** 19.0 67.3
17 Ever Self-injury 9.2 57.4*** 5.9 36.5
18 Ever Attached to objects 6.5 59.6*** 11.7 46.1
19 Current Friends S 7.3 85.2*** 21.4 66.5
All items are for age 4–5 years
20 Age 4–5 Social chat C 19.4 162.4*** 19.6 82.5
21 Age 4–5 Imitation C 5.9 66.0*** 18.4 57.3
22 Age 4–5 Pointing to express interest C 2.5 21.3*** 42.4 65.2
23 Age 4–5 Gestures C 0.6 4.9* 56.6 45.6
24 Age 4–5 Nodding to mean yes C 7.0 83.3*** 28.6 73.7
25 Age 4–5 Head shaking to mean no C 3.9 43.7*** 31.6 64.1
26 Age 4–5 Eye gaze S 8.0 70.1*** 10.2 47.6
27 Age 4–5 Social smiling S 13.1 73.2*** 4.9 40.3
28 Age 4–5 Showing and directing attention S 6.3 59.8*** 12.6 47.6
29 Age 4–5 Offering to share S 12.3 116.1*** 13.2 65.0
30 Age 4–5 Seeking to share enjoyment S 14.6 113.4*** 8.7 58.3
31 Age 4–5 Offering comfort S 14.2 115.8*** 9.8 60.5
32 Age 4–5 Quality of social overtures S 1.8 6.6** 20.9 32.0
33 Age 4–5 Range of facial expression S 18.5 94.2*** 4.4 45.9
34 Age 4–5 Spontaneously join in social games C 10.2 79.8*** 8.3 47.8
35 Age 4–5 Pretend or make-believe games C 17.8 156.6*** 18.4 80.1
36 Age 4–5 Interest in children S 5.0 60.0*** 28.4 66.5
37 Age 4–5 Response to other children’s approaches S 9.7 82.8*** 9.7 51.0
38 Age 4–5 Look up and pay attention 13.4 102.6*** 8.3 54.6
39 Age 4–5 Imaginative play with peers S 23.1 176.9*** 17.5 83.0
40 Age 4–5 Group play S 83.5 171.8*** 2.0 62.4

SCQ: Social Communication Questionnaire; OR: odds ratio; TD: typically developing; ASD: autism spectrum disorder; C: communication subscale;
R: repetitive behaviors subscale; S: social interaction subscale.
aItem no. 1 documents whether or not the child has phrase speech and does not have scoring value.
bAnalyses for items 2–7 were based on fewer subjects (n ranging from 295 to 300) due to items being skipped by parents because of lack of

sufficient language.
*p < 0.05; **p < 0.01; ***p < 001.

cooperative with peers), with a median OR of 12.3 (item ORs ranged from 1.2 (item 13; unusually intense special
29; age: 4–5 years; offering to share). For the 8 items on interests) to 30.1 (item 15; motor mannerisms), with a
the repetitive behaviors subscale (all ever codings), the median value of 7.4. Thus, on average, social
6 Autism 00(0)

interaction items had a stronger association with case

status than did items in the other two domains.

Correlations between SCQ scores

We computed the total score in two ways. First, we com-
puted a nonverbal total by summing the scores for items
8–40 (excluding the six items (2–7) that require sufficient
language). Next, we calculated the total as previously for
nonverbal subjects but as the sum of items 2–40 for ver-
bal subjects. Pearson’s correlation between these two
scores was 0.983 (p < 0.01) in the overall sample and
0.961 (p < 0.01) among ASD children only, suggesting
that no imputation technique to adjust for unequal num-
ber of items between verbal and nonverbal total scores
was necessary. In the whole sample, there was a strong
correlation between social and communication scores
(Pearson’s r: 0.731; p ⩽ 0.001) and slightly lower corre-
lation between the repetitive behaviors score with either
the social score (Pearson’s r: 0.522; p ⩽ 0.001) or the
communication score (Pearson’s r: 0.521; p < 0.01). Figure 2. Discriminant validity of the Social Communication
Questionnaire (SCQ) total score in the receiver operating
When analyses were repeated in the ASD subsample
characteristic (ROC) curve analysis (n = 412).
(n = 206), the same pattern emerged, although associa-
tions were in general weaker, especially with the repeti-
tive behaviors score. The corresponding coefficients Discussion
were as follows: for the social and the communication
Our study is the first to assess the use of the SCQ screening
scores (r = 0.593; p < 0.01), for the repetitive behaviors
tool in an Arabic-speaking population, which worldwide is
and the social scores (r = 0.178; p = 0.01), and for the
estimated to number 420 million individuals (United Nations
repetitive behaviors and communication scores (r = 0.293;
Department of Economic and Social Affairs, 2017). The need
p < 0.01). No significant correlation was found with age.
to validate a screening tool in Arabic was recognized by the
All results were the same when nonparametric Spearman’s
authors while planning a country-wide autism epidemiologi-
r coefficients were calculated (data not shown).
cal study for the Qatari elementary school population. The
Modified Checklist for Autism in Toddlers (M-CHAT) is the
Discriminant validity of SCQ scores in ROC only available validated screening tool for Arabic populations
(Mohamed et al., 2016; Seif Eldin et al., 2008). However, the
analyses and cutoff performances
M-CHAT is restricted to a young and narrow age group,
We used ROC analysis to summarize the overall discrimi- which made it inadequate for our planned study. We initially
nant validity of the SCQ. For the total score, the area under translated and piloted the Social Responsiveness Scale (SRS),
the curve (AUC) was 0.95 (95% confidence interval (CI): which is among the most widely used quantitative parent/
0.93–0.97) (see Figure 2). For the subscale scores, the teacher report tools for use in general population, educational,
AUCs were 0.923 (95% CI: 0.898–0.949) for the social and clinical settings (Bölte et al., 2008; Constantino and
interaction score, 0.872 (95% CI: 0.838–0.905) for the Gruber, 2005). After analyzing the pilot data, however, we
communication score, and 0.856 (95% CI: 0.819–0.893) noted a large proportion of partially completed forms, a fact
for the repetitive behaviors score. that many parents attributed to the “long” time needed to
For the established cutoff of 15, sensitivity and specificity complete the scale and the multiple, at times confusing,
were 0.796 and 0.966, respectively. For a cutoff of 12, the response formats for each item. We then switched to the SCQ,
values for sensitivity and specificity were 0.893 and 0.893, previously known as the Autism Screening Questionnaire
respectively. In-between values for the cutoff were associated (Berument et al., 1999; Rutter et al., 2003), because it required
with similar overall performance as defined by the sum of the less time to complete and the response to each item was
sensitivity and specificity-1 (Youden’s index). Thus, for cut- binary: yes or no. Moreover, and contrary to the SRS that
offs ranging from 11 to 15, the sensitivity varied between focuses on current behavior only, the SCQ screens the lifes-
90.3% and 79.6%, whereas the specificity varied between pan of the individual, thus allowing highly suggestive ASD
85.4% and 96.6%. Within that range, Youden’s index had a features to be included in the screening score even though
narrow range of variation between 75.7% and 78.6%. they may reflect past, but no longer current, behaviors. As
Aldosari et al. 7

such, the SCQ better approximates the longitudinal develop- showed comparably high sensitivity and specificity, including
mental perspective recommended for autism in diagnostic the Turkish version (sensitivity = 0.94 and specificity = 0.84)
schemes such as the DSM and the International Classification and the Mandarin Chinese version (sensitivity = 0.957 and
of Diseases and that is also embodied in diagnostic tools such specificity = 0.825) (Avcil et al., 2015; Guo et al., 2011). Age
as the ADI-R. and gender effects within this primary school age range were
The strength of our study resides in the large sample size absent or minimal, a result that supports the use of a single
and the inclusion of subjects from two Arabic-speaking coun- threshold to screen boys and girls within that age range. Thus,
tries, Qatar and Saudi Arabia, which have many similarities in our study shows that, for the established cutoff of 15, the sen-
their demographic and ethnic characteristics. These countries sitivity and specificity of the Arabic SCQ was comparable to
also have in common high consanguinity rates among their figures from previous studies and supports the use of the
populations in the order of 40% to 60% (Tadmouri et al., Arabic SCQ questionnaire as a screening instrument for epi-
2009). Qatar’s population is estimated to be around 2.6 mil- demiological purposes in primary school age samples. In
lion and the Kingdom of Saudi Arabia 32 million. Conducting addition, we also determined that the SCQ total score achieved
population research on ASD, especially in this region of the slightly better discrimination between children with ASD and
world, can be challenging due to the potential stigma associ- a comparison group than did SCQ subscores, supporting use
ated with ASD and other developmental conditions and lim- of the whole scale rather than a shorter version, especially
ited research infrastructure. Another feature of our study is the because the full SCQ is well accepted and can be rapidly
translation process, which was performed in collaboration completed by caregivers.
with an author-assigned reviewer with multiple rounds of One limitation of our study is that, due to our sampling
translation and back-translation to ensure linguistic equiva- procedures, children with ASD had mostly moderate-to-
lence and cultural appropriateness while maintaining the severe impairments, whereas control children were without
screening performance of the original English version. learning or behavioral problems. It is therefore possible that
We included children between the ages of 5 and 12 years the strong discriminant ability obtained with the SCQ to dif-
because this is the age group in which autism diagnosis is ferentiate our two samples may have been slightly overesti-
most reliable (Charman et al., 2005). This age group is also mated. If so, use of the SCQ in our population-based study to
similar to that targeted by our large epidemiological study as screen more representative samples of both cases and con-
well as other major population studies such the Korean preva- trols might show a lower performance. However, the high
lence study (Kim et al., 2011), the Mexico study (Fombonne levels of specificity and sensitivity obtained in this prelimi-
et al., 2016), and the studies of the Centers for Disease Control nary study were robust and should help maintain good psy-
and Prevention (Baio et al., 2018). Controls were matched by chometric properties in a different sampling context. A
age and sex to enable a better comparison between the two second limitation is that samples were recruited for conveni-
groups. The SCQ showed a high internal consistency coeffi- ence rather than for being representative of each site or
cient as measured by Cronbach’s alpha coefficient for the recruitment source, and thus, results might be sensitive to
total as well as the subscores, which was comparable to other some undocumented selection biases. However, the large
SCQ validation studies (Avcil et al., 2015). sample size and the recruitment across multiple sources and
As expected, SCQ scores and subscores were signifi- sites should have protected our study against substantial
cantly different between the two groups. The correspond- biases and atypical findings. Finally, absence of ASD among
ing effect sizes for these differences were very large, more controls was evaluated by parental and teacher reports only.
so for the total SCQ score, although the subscores also However, in the unlikely eventuality that some TD children
showed highly significant differences. Notably, the social might have had ASD, it would have contributed to decrease
interaction scores and items appeared to have the best dis- SCQ discriminant accuracy rather than the other way around.
criminant properties compared to the other two domains.
Thus, the social scale subscore had the largest overall dis- Conclusion
criminant validity as measured by the effect size and the
AUC from the ROC curves, and it had the highest reliabil- Our study suggests that the Arabic version of the SCQ can
ity. In examining item-level discriminant ability, social differentiate children with clinically diagnosed ASD from
interaction items showed the more robust associations TD children by use of the established cutoff value of 15.
with case-control status compared with items from the Therefore, the Arabic version of the SCQ will be useful in
other two domains, a result that is in line with social symp- clinical settings for screening children suspected to have
toms being at the core of autism impairment. autism as well as for executing epidemiologic studies.
When inspecting performances associated with different
cutoffs, the sensitivity and specificity were 0.796 and 0.966 Acknowledgements
for the cutoff of 15. The original SCQ validation study The authors would like to thank all the staff of the autism centers
showed sensitivity of 0.86 and specificity of 0.78 for the same and schools who contributed in distributing and collecting the
cutoff (Berument et al., 1999). Few translated versions SCQ forms. They also would like to thank Western Psychological
8 Autism 00(0)

Services (WPS) staff for their help during the process of translat- Dawson G, Rogers S, Munson J, et al. (2010) Randomized, con-
ing and reviewing the Arabic SCQ. They acknowledge Jennifer trolled trial of an intervention for toddlers with autism: the
Holmes, ELS, for medical editing. early start Denver model. Pediatrics 125(1): e17–e23.
Eaves LC, Wingert HD, Ho HH, et al. (2006) Screening for
Funding autism spectrum disorders with the social communication
questionnaire. Journal of Developmental and Behavioral
The author(s) disclosed receipt of the following financial support
Pediatrics 27(2 Suppl.): S95–S103.
for the research, authorship, and/or publication of this article:
Fombonne E (1991) The use of questionnaires in child psy-
The study was funded by the Qatar National Research Fund
chiatry research: measuring their performance and choos-
(NPRP 6-093-3-024).
ing an optimal cut-off. Journal of Child Psychology and
Psychiatry, and Allied Disciplines 32(4): 677–693.
References Fombonne E, Marcin C, Manero C, et al. (2016) Prevalence of
Allen CW, Silove N, Williams K, et al. (2007) Validity of the autism spectrum disorders in Guanajuato, Mexico: the Leon
social communication questionnaire in assessing risk of survey. Journal of Autism and Developmental Disorders
autism in preschool children with developmental problems. 46(5): 1669–1685.
Journal of Autism and Developmental Disorders 37(7): Guo YQ, Tang Y, Rice C, et al. (2011) Validation of the autism
1272–1278. spectrum screening questionnaire, mandarin Chinese ver-
American Psychiatric Association (APA) (2013) Diagnostic sion (CH-ASSQ) in Beijing, China. Autism 15(6): 713–727.
and Statistical Manual of Mental Disorders. 5th ed. Kim YS, Leventhal BL, Koh YJ, et al. (2011) Prevalence of
(DSM-5). Available at: https://doi.org/10.1176/appi.books autism spectrum disorders in a total population sample. The
.9780890425596.744053 American Journal of Psychiatry 168(9): 904–912.
Avcil S, Baykara B, Baydur H, et al. (2015) The validity and reli- Mohamed FE, Zaky EA, Youssef A, et al. (2016) Screening of
ability of the social communication questionnaire-Turkish Egyptian toddlers for autism spectrum disorder using an
form in autistics aged 4–18 years. Türk Psikiyatri Dergisi Arabic validated version of M-CHAT; report of a community
26(1): 56–64. https://doi.org/10.5080/u7298 -based study (Stage I). European Psychiatry 34: 43–48.
Baio J, Wiggins L, Christensen DL, et al. (2018) Prevalence of Peters-Scheffer N, Didden R, Korzilius H, et al. (2012) Cost com-
autism spectrum disorder among children aged 8 years— parison of early intensive behavioral intervention and treat-
autism and developmental disabilities monitoring network, ment as usual for children with autism spectrum disorder
11 sites, United States, 2014. MMWR Surveillance Summary in the Netherlands. Research in Developmental Disabilities
67: 1–23. 33(6): 1763–1772.
Berument SK, Rutter M, Lord C, et al. (1999) Autism screen- Rutter M, Bailey A and Lord C (2003) Social Communication
ing questionnaire: diagnostic validity. The British Journal Questionnaire: Manual. Los Angeles, CA: Western
of Psychiatry 175(5): 444–451. Psychological Services.
Bölte S, Poustka F and Constantino JN (2008) Assessing autistic Seif Eldin A, Habib D, Noufal A, et al. (2008) Use of M-CHAT
traits: cross-cultural validation of the Social Responsiveness for a multinational screening of young children with autism
Scale (SRS). Autism Research 1(6): 354–363. in the Arab countries. International Review of Psychiatry
Charman T, Taylor E, Drew A, et al. (2005) Outcome at 7 years 20(3): 281–289.
of children diagnosed with autism at age 2: predictive valid- Tadmouri GO, Nair P, Obeid T, et al. (2009) Consanguinity and
ity of assessments conducted at 2 and 3 years of age and reproductive health among Arabs. Reproductive Health 6:
pattern of symptom change over time. Journal of Child 17.
Psychology and Psychiatry, and Allied Disciplines 46(5): United Nations Department of Economic and Social Affairs
500–513. (2017) World population prospects the 2017 revision.
Constantino JN and Gruber CP (2005) Social Responsiveness Scale Available at: http://www.un.org/en/development/desa
(SRS). Los Angeles, CA: Western Psychological Services. /population/