Caution when Screening for Autism among Socially Anxious Youth

Abstract

Social anxiety disorder (SAD) is commonly comorbid with autism spectrum disorder (ASD). Here, in a sample of 86 children and adolescents (M_AGE = 12.62 years; 68.6% male), 28 of whom were diagnosed with ASD, 34 with SAD, and 24 with comorbid ASD and SAD, we compared parent-reported scores from the Social Responsiveness Scale-Second Edition (SRS-2; Constantino & Gruber, 2012) to determine the sensitivity and specificity of the measure in cases of differential diagnosis between SAD and ASD. Results suggest that neither the subscales, nor the SRS-2 total score, consistently differed between ASD and SAD. Sensitivity and specificity analyses suggested that the SRS-2 total poorly discriminated ASD from SAD. When screening socially anxious youth for possible ASD, caution should be taken.

Autism Spectrum Disorder (ASD) is characterized by deficits in socio-emotional reciprocity, social interactions, social communication, and the presence of restricted and repetitive behaviors and interests (APA, 2013). Prevalence estimates have continued to increase over the years with current estimates approximating an ASD diagnosis in 1 out of every 54 children (Maenner et al., 2020). The high rate of psychiatric comorbidity (Supekar, Iyer, & Menon, 2017) has complicated the diagnostic picture for individuals with ASD, in particular for those with higher IQ and less impairing ASD symptoms (Iyama-Kurtycz, 2020). Social Anxiety Disorder (SAD) is one of the most common comorbid disorders associated with ASD and one of the most complicated diagnoses to differentiate from ASD given symptom overlap between ASD and SAD (Spain et al., 2018). The high co-occurrence of SAD within ASD and the degree to which symptoms overlap one another requires clinicians to consider the degree to which ASD screeners are not only sensitive, but also specific, to ASD.

SAD is characterized by persistent fears of negative evaluation; people with SAD tend to avoid situations in which negative evaluation is possible, resulting in social avoidance, social withdrawal, and social worries (APA, 2013). SAD in children and adolescents also involves anxiety related to interactions with peers, which may be expressed by crying, tantrums, clinging to parents, or failing to speak in social situations (Spence & Rapee, 2016). SAD is quite common among youth, with lifetime prevalence at age 18 estimated at 9.1% (Merikangas et al., 2010) and prevalence increasing from childhood through adolescence (Lawrence et al., 2015). In fact, SAD is the most common anxiety disorders seen among adolescents (Merikangas et al., 2010). The increase in SAD symptoms observed during adolescence is likely due to heightened social demands (van den Bos et al., Westenberg, 2014; Westenberg et al., 2007), increased time spent with peers (Blöte et al., 2015), and heightened sensitivity to peer relationships and potential negative evaluation from peers (Blöte et al., 2015).

In addition, research suggests that SAD may be one of the most common co-occurring anxiety disorders among children and adolescents with ASD (Kuusikko et al., 2008; Spain et al., 2018), with approximately 16.6% of youth with ASD demonstrating co-occurring SAD (van Steensel et al., 2011). However, this may actually under-estimate the true prevalence of SAD in youth with ASD, due to diagnostic overshadowing, which occurs when social anxious symptoms are attributed to the diagnosis of ASD and not recognized as reflecting a potential secondary co-occurring diagnosis (Grondhuis & Aman, 2012; Kreiser & White, 2014). Specifically, there is considerable phenotypic overlap between the two disorders related to social avoidance, social withdrawal, and physiological hyperarousal in social situations. Many individuals with ASD demonstrate high levels of social worries due to (accurate) perceptions of difficulties with social communication and social interactions (White et al., 2009; White, Bray, & Ollendick, 2012). Youth with ASD may present with additional risk factors for developing SAD, including social communication impairments, diminished social motivation, lack of responsiveness to others, and poor social skills. These behaviors together may increase teasing and bullying and may contribute to further social withdrawal, thus creating a bidirectional effect (Spain et al., 2018). Chang et al. (2012) found that youth with ASD and social anxiety had more severe social impairment, especially in reference to assertive (i.e., initiating social interactions) and responsive (i.e., awareness of social rules and conventions) social skills, than those without co-occurring social anxiety symptoms. Conversely, many individuals with SAD demonstrate symptoms of ASD, including social avoidance, social withdrawal, and even repetitive behaviors when communicating in certain social situations (e.g., hand wringing when meeting someone new or not knowing what to say; Towbin et al., 2005).

Age of onset is one factor frequently used to differentiate presentations of ASD from SAD, as symptoms of ASD are frequently present from early childhood (i.e., before 3-years of age; APA, 2013) and continue throughout life, whereas SAD tends to have its onset in early adolescence (i.e., around 13-14 years of age; de Lijster et al., 2017) and to be prevalent throughout adolescence). Although these age ranges seem quite distinct, diagnostic differentiations become difficult even within age, considering other factors such as symptom severity, level of impairment, and demographic factors including socioeconomic status and rurality. For instance, although the average age of a clinical diagnosis of ASD ranges from 3.1 to 4.75 years of age (Maenner et al., 2020), individuals with higher IQs and higher overall functioning are not diagnosed until an average age of 4.75 to 7.2 years, and this average age continues to climb when families reside in rural areas, and/or are below the poverty line (Mandell, Novak, & Zubritsky, 2005).

Comprehensive assessments can aid in the differentiation of psychiatric disorders (Constantino et al., 2003) like ASD and SAD. However, comprehensive psychological assessments are time consuming, with average cumulative time for administration, scoring, and interpretation between 4 to 6 hours (Camara, Nathan, & Puente, 2000). There is a need for reliable and sensitive screening instruments for use by clinicians and mental health providers to identify individuals in need of a more comprehensive diagnostic evaluation (Towbin et al., 2005). However, these screening tools need to not only accurately identify youth with ASD (sensitivity) but they also need to minimize the frequency of false positives in cases in which symptoms might be due to other psychiatric disorders like SAD (specificity).

The Social Responsiveness Scale (SRS; Constantino & Gruber, 2005) and its updated version, the Social Responsiveness Scale-Second Edition (SRS-2; Constantino & Gruber, 2012), are widely used screeners of ASD symptom severity and overall social impairments characteristic of ASD (South et al., 2017). Evidence across studies and research groups suggests the SRS reliably differentiates those with ASD from typically developing youth (Bölte et al. 2011; Cholemkery et al. 2014a; Constantino & Gruber, 2005, 2012). Although some studies have found that the SRS differentiates youth with ASD from other psychiatric conditions (Constantino & Gruber, 2005), other studies have found that sensitivity and specificity values are weaker in discriminating between those with ASD from participants with other psychiatric conditions and in particular the anxiety disorders (Bölte et al., 2011; Cholemkery et al., 2014a; Pine et al., 2008). For example, findings by Pine and colleagues (2008) suggested that youth with anxiety and mood disorders obtained higher scores using the original SRS measure (Constantino & Gruber, 2005) than children without these disorders. Two recent studies have specifically investigated the sensitivity and specificity of the SRS for ASD and anxiety disorders. Cholemkery and colleagues (2014b) determined that although youth with ASD demonstrated greater levels of social impairment relative to youth with SAD/selective mutism, the SRS overestimated the presence of ASD. Further, their findings suggested that the SRS Social Motivation subscale provided poor distinction between the ASD and the SAD/selective mutism groups. Second, and within an adult sample, South et al. (2017) found that the overall discriminant validity of the SRS-2 was poor. Similar to Cholemkery et al. (2014b), South and colleagues reported that the SRS-2 Social Motivation subscale did not distinguish those with ASD from those with anxiety. Cholemkery et al. excluded participants with co-occurring SAD or selective mutism from their ASD sample, whereas South et al. relied on self-report measures of general anxiety rather than specifically assessing for SAD symptoms. As such, prior research has not examined the sensitivity and specificity of the SRS-2 in differentiating youth with diagnosed SAD or ASD. Nor, have these studies examined the conjoint risk of comorbid ASD and SAD for social impairments as measured on the SRS-2 in youth.

The current study aimed to determine the degree to which social impairments [e.g., impairments in social communication and restricted interests and repetitive behaviors (RRBs)] differentiated youth with SAD from those with ASD as well as youth with co-occurring ASD and SAD in order to determine the sensitivity and specificity of the SRS-2. We predicted that youth with ASD+SAD would demonstrate the highest levels of social impairment, followed by those with ASD only followed by youth with SAD only. We also explored the specificity and sensitivity of the SRS-2 total score, consistent with Cholemkery and colleagues (2014b) and South et al. (2017) in differentiating the three clinical groups.

Method

Participants

Data were drawn from a university-affiliated child outpatient clinic in the southeastern United States. In terms of catchment area, the clinic serves a geographic region immediately surrounded by a large and historically underserved Appalachian rural area. The majority of participants with ASD were not previously diagnosed, consistent with prior research demonstrating that rurality may lead to delays in evaluations (Mandell et al., 2005; Smith et al., 2017). Referrals came from parents, schools, pediatricians, and other clinicians for a range of problems. For the majority of participants (n = 71), children and their primary caregivers participated in comprehensive psychoeducational assessments including multiple formats and informants (e.g., semi-structured diagnostic interviews, questionnaires, behavioral observations), and examining a range of domains (e.g., cognitive ability, processing, academic performance, behavior). For a subset of participants, data were drawn from a pre-treatment assessment session from a randomized controlled trial which examined the effects of a program to facilitate transition to adulthood for adolescents with ASD (n = 15; STEPS; White et al. 2019)¹. For demographic information on the sample, please refer to Table 1. The 86 participants were classified into the following groups: SAD only, ASD only, and ASD+SAD. Children and adolescents in the ASD only (n = 28; M Age = 12.11 years; 68% male) and ASD+SAD groups (n = 24; M Age = 13.88 years; 79% male) met diagnostic criteria for ASD, as determined by clinical evaluation and supported by the ADOS-2 and ADI-R, scored by research reliable clinicians. For the SAD only (n =34; M Age = 12.15 years; 62% male) and ASD+SAD groups, participants met criteria for a clinical diagnosis of SAD as on a semi-structured diagnostic interview [i.e., Anxiety Disorders Interview Schedule for DSM-IV-Child and Parent Versions (ADIS-IV-C/P; Silverman & Albano, 1996)]. Across the three groups, the age range of participants was 7 to 17 years, and the cognitive ability range was from 69 to 131. In terms of race, most participating youth were White (81.4%).

Table 1.

Demographic data

	SAD (n = 34)	ASD (n = 28)	ASD + SAD (n = 24)	χ2 / F
Gender (male)	21 (62%)	19 (68%)	19 (79%)	4.143
Race				13.925
White	29 (85%)	20 (71%)	21 (88%)
Black	0 (0%)	2 (7%)	2 (8%)
Latino	1 (3%)	2 (7%)	0 (0%)
Asian	1 (3%)	0 (0%)	0 (0%)
Other	3 (9%)	2 (7%)	0 (0%)
NA	0 (0%)	2 (7%)	1 (4%)
Age in years (M, SD)	12.15 (2.89)	12.11 (3.44)	13.88 (3.00)	2.732
IQ (M, SD)	99.00 (11.99)	97.81 (16.78)	101.00 (11.46)	.331

^*p < .05

^**p < .01

Note. NA = Not available; SAD = social anxiety disorder; ASD = autism spectrum disorder.

The ADIS-IV-C/P was completed with all participants, allowing us to rule out SAD for the ASD only group. For participants recruited from the assessment clinic, if the referral question was specific to possible ASD (n = 27), the ADOS-2 and the ADI-R were administered by trained and research reliable clinicians. If the referral question was not specific to ASD (n = 13), but there were other indications of possible ASD (e.g., elevations on the SRS-2, elevations on the ASD module of the ADIS-IV-P, behavioral observations), the ADOS-2 was also administered. Clinical diagnoses were made on the basis of clinical elevations on the ADOS-2 and/or the ADI-R, and paired with clinical judgment (i.e., licensed clinical psychologists with several years of experience working with individuals with ASD). Participants who met diagnostic criteria for ASD (n = 40 in total) met or exceeded cutoff on the ADOS-2 in 83% of cases. Some of the youth in the SAD group were also evaluated for ASD (ADOS-2, n = 13, 38%; of those, n = 4 were also administered the ADI-R). Only 1 participant met or exceeded cutoff on the ADOS-2 but did not meet additional diagnostic criteria per the ADI-R and thus were not diagnosed. For participants recruited from STEPS (White et al. 2019), the ADOS-2 was administered for all cases; the ADI-R was not administered.

Procedure

The study was approved by the university’s institutional review board for human subject research. Participating parents provided informed consent. Parents and children were administered a semi-structured diagnostic interview (i.e., ADIS-IV-C/P). Parents and children also completed the SRS-2 and other questionnaires, not all of which were analyzed in the present study. Cognitive ability was assessed using the Full-Scale IQ from either the Wechsler Abbreviated Scales of Intelligence, Second Edition (WASI-II; Wechsler & Hsiao-pin, 2011) or the Wechsler Intelligence Scale for Children, Fifth Edition (WISC-5; Wechsler, 2014).

Measures

Social Responsiveness Scale-Second Edition, School Age Form, Parent Report (SRS-2; Constantino & Gruber, 2012). The SRS-2 is a 65-item measure of ASD-related social impairments. Items were scaled from 1 (not true) to 4 (almost always true). Individual items comprise five subscales including Social Awareness, Social Cognition, Social Communication, Social Motivation, and Restricted Interests and Repetitive Behaviors (RRB). An overall Total Score (i.e., all five subscales) is also generated. For our primary and exploratory aims, we rely on the total raw score given its previous use in studies which focused on sensitivity and specificity of the SRS/SRS-2 (see Cholemkery et al. (2014b; South et al., 2017) and also because it is best practice psychometrically for screening purposes (Constantino & Gruber, 2012).

The Social Awareness subscale is comprised of 8 items and is used to measure an individual’s ability to recognize others’ social cues. Social Cognition is comprised of 12 items relating to interpreting social behavior of others. Social Communication is comprised of 22 items relating to reciprocal communication in social situations. Social Motivation is an 11-item subscale to assess the degree to which an individual appears motivated to engage in social interactions. Lastly, RRB is a 12-item subscale used to measure circumscribed interests and motor stereotypies (Bruni, 2014). Parent-report of their child’s ASD symptoms on the SRS-2 were analyzed as the primary dependent variables.

Following recommendations for research by Hus et al. (2013) and Cholemkery et al. (2014b), raw scores were used for all analyses; however, we also report SRS T-scores in Table 3 since they are more readily interpreted and are normed within gender. For the SRS-2 subscales, reliability estimates from the current sample were as follows: Social Awareness (α = .652); Social Cognition (α = .757); Social Communication (α = .853); Social Motivation (α =.762); and Restricted and Repetitive Behaviors (α = .860). In the full sample, alpha for the total score was .945 which suggests excellent internal consistency.

Table 3.

Means, Standard Deviations, and F statistics SRS Subscales (T-Scores), and SRS Total T-Score

	M(SD)			F
Group Variable	SAD	ASD	ASD +SAD
SRS-2 Social Awareness	56.76(11.98)	64.25(15.67)	68.33(13.10)	5.48*
SRS-2 Social Cognition	59.09(13.28)	65.18 (16.08)	67.00 (13.03)	4.46^
SRS-2 Social Communication	62.29(12.87)	64.75(15.14)	70.71(13.23)	2.69^
SRS-2 Social Motivation	63.88(13.50)	60.25(12.92)	73.67(13.51)	6.94*
SRS-2 Total	62.74 (12.13)	67.21(15.71)	71.96(13.12)	3.23*
SRS-2 RRB	71.91 (13.30)	72.64 (13.23)	76.46 (11.05)	.980

^{^}p < .10

^*p < .05

^**p < .01

Note. SAD = social anxiety disorder; ASD = autism spectrum disorder.

Anxiety Disorders Interview Schedule for DSM-IV-Child and Parent Versions (ADIS-IV-C/P; Silverman & Albano, 1996). The ADIS-IV-C/P is a semi-structured clinical interview for the diagnosis of psychiatric disorders in children and adolescents. Further, the ADIS-IV-C/P has demonstrated appropriate reliability and validity for youth with ASD (Lecavalier et al., 2014). The ADIS-IV-C/P was administered by clinicians who were trained in the differential diagnosis of ASD and SAD. In order for participants to meet diagnostic criteria for SAD on the ADIS-IV-C/P, their avoidance of social situations had to be social in nature rather than due to non-social situational aspects (i.e., social disinterest, sensory over arousal), which follows criteria outlined in past studies that distinguished core ASD impairments from SAD symptoms (Kerns et al. 2014; Leyfer et al. 2006). During both the ADIS-IV-P and ADIS-IV-C interviews, clinicians assessed the severity of the youth’s social anxiety and other psychological disorders. Clinicians assigned a clinician severity rating (CSR) on a 9-point scale (0-8, with any rating ≥ 4 indicating probable diagnosis and clinical interference). Separate clinicians administered the ADIS-IV-C and ADIS-IV-P to the child and parent, respectively. In addition to the parent interview, the ADIS-IV-C was administered only to child and adolescent participants who were ≥8 years. Final diagnoses were derived based on a composite score such that, if either interview yielded a diagnosis, that diagnosis was retained. The ADIS-IV-C/P has demonstrated acceptable reliability (child: k = .71; parent: k = .77; Jarrett, Van Meter, Youngstrom, Hilton, & Ollendick, 2016) and validity (Silverman, Saavedra, & Pina, 2001).

Statistical Analyses

In order to investigate possible group-level differences among demographic variables, we used a series of univariate analyses of variance (ANOVAs) for the continuous variables and chi square analyses for the nominal variables to determine if any variables (e.g., IQ, participant sex) needed to be included as covariates (Table 1). In order to test our primary aim, we used a series of ANOVAs using Holm’s modified Bonferroni correction, consistent with recommendations by Jaccard and Guilamo-Ramos (2002). Preliminary assumption testing was conducted to check for normality, linearity, univariate and multivariate outliers, homogeneity of variance, and multicollinearity. For the ANOVAs, group (i.e., ASD only, ASD+SAD, SAD only) was the independent variable, and the dependent variables were the five SRS-2 subscale scores which comprise the SRS-2 total raw score. Post-hoc contrasts were computed using Sidak adjustments for multiple comparisons in order to increase power while also accounting for Type I error, especially when there are several comparisons. This analytic approach is also consistent with Cholemkery and colleagues (2014b). These analyses were all conducted with IBM SPSS Statistics Version 25. We also conducted exploratory supplementary analyses of the ANOVAs to determine if restriction of the sample to youth 10 years and older (i.e., closer age to onset of SAD) affected our primary findings (see Supplementary material). Two one-sided t-tests for equivalence (TOST; Lakens et al., 2017) compared mean SRS-2 total raw scores in groups for which there were nonsignificant differences to determine the extent of similarity in the SRS-2 impairments endorsed between the groups. We relied on the TOSTER package in R for the TOST analyses. For the post-hoc equivalence testing, the lower and upper bounds are effect size quantities (Cohen’s d) which are specified to represent the range within which potential effect sizes are considered equivalent to a non-meaningful result. Consistent with effect size guidelines offered by Cohen (1988), we explored whether total levels of social impairment were similar within large effect bounds (lower bound = −.80 and upper bound = .80).

We also examined scores in terms of the recommended guidelines for clinical interpretation (mild, moderate, or severe concern) as reported in the SRS-2 manual. Sensitivity and specificity was calculated in order to determine the diagnostic validity of the SRS-2 total score in differentiating the three clinical groups (Exploratory Aim). Per recommendations by Cicchetti and colleagues (1995), levels of diagnostic accuracy were classified as follows: < 0.70 is poor, < 0.79– 0.70 is fair, > 0.89–0.80 is good, and ≥ 0.90 is excellent.

Results

Group descriptive statistics (see Table 1) indicated no problems with skewness or kurtosis. Further, there were no significant differences between the SAD only, ASD only, and ASD+SAD groups in terms of participant sex, χ² = 4.143, p = .387 (phi = .219); age, F(2, 83) = 2.732, p = .071 (ηρ²= .062), IQ, F(2, 76) = .331, p = .719 (ηρ² = .009), or race, χ² = 13.925, p = .455 (phi = .402). Given there were no differences on demographic variables, all subsequent analyses were conducted without covariates. There were no significant group-level differences between the youth participating in STEPS versus the youth receiving the comprehensive psychoeducational assessments on cognitive ability. Although these two groups significantly differed on age (t = 6.54, p <.001) and participant sex (χ² =8.32, p = .016, phi = .311), neither age nor sex were associated significantly with any of the dependent variables of interest (ps = .518-.947); therefore, neither age nor sex were used as covariates.

Univariate ANOVAs examined group status (i.e., SAD only, ASD only, and ASD+SAD) as the independent variable and the five SRS-2 subscale raw scores and the SRS-2 total raw score as the dependent variables (Table 2). For the SRS-2 total raw score, univariate F-tests showed there was a significant difference between the groups. As can be seen in Table 2, post hoc contrasts including Sidak correction showed the ASD group did not differ significantly from children and adolescents with SAD (M difference = −10.670, t = 1.188, p = .240, d = .300) for the SRS-2 total raw score (Table 2). TOST analyses suggested that the SAD and ASD groups demonstrated similar social impairment levels within large effect bounds (d=.80, p = .039). However, youth within the ASD+SAD group demonstrated a significantly higher mean SRS-2 total raw score compared to the SAD only group (M difference = −28.549, t = −3.186, p = .009, d = −.838. TOST analyses suggested that the ASD only and ASD+SAD groups demonstrated dissimilar, albeit non-statistically significant differences, in the total levels of social impairment within large effect bounds (d = .80, p = .107). Univariate F-tests showed there was a significant difference between the three groups for all subscales except for the RRB subscale (Table 2). However, with application of Holm’s modified Bonferroni correction, the only significant group differences which remained statistically significant was for the SRS-2 Social Awareness and Social Motivation subscales. At the subscale level, post-hoc contrasts suggest that the SAD and ASD groups differed significantly on Social Awareness and Social Cognition, with the ASD group having significantly higher scores compared to the SAD group. Between-group differences also emerged between the ASD and ASD+SAD groups for the Social Motivation subscale, with the ASD+SAD group demonstrating significantly higher scores, per post-hoc contrasts.

Table 2.

Means, Standard Deviations, and F statistics SRS Subscales (Raw Scores) and SRS-2 Total

Group Variable	SAD	ASD	ASD +SAD
	M(SD)			F	Significant Post hoc Comparisons
SRS-2 Social Awareness	7.74(3.99)	10.54(5.07)	11.63(4.30)	6.05*a	ASD > SAD* ASD+SAD > SAD*
SRS-2 Social Cognition	10.82(7.18)	16.32(9.48)	15.79(7.17)	4.46*	ASD > SAD* ASD+SAD > SAD^
SRS-2 Social Communication	21.38(11.65)	26.32(12.99)	30.75(13.44)	3.95*	ASD+SAD > SAD*
SRS-2 Social Motivation	12.41(6.50)	11.46(6.21)	18.38(7.60)	7.94**a	ASD+SAD > SAD* ASD+SAD > ASD*
SRS-2 RRB	17.41 (8.58)	17.61 (7.50)	23.25 (12.78)	3.09^	ASD+SAD > SAD^
SRS-2 Total	64.12 (31.35)	74.79(39.38)	92.67(36.61)	4.54*	ASD+SAD > SAD*

^{^}p < .10

^*p < .05

^**p < .01

Note. SAD = social anxiety disorder; ASD = autism spectrum disorder.

^aSuperscript suggest significant difference following Holm’s correction for experimentwise error rate

Post hoc testing using Sidak correction.

Closer examination showed that the mean total raw score on the SRS-2 for the SAD group (M = 64.12) exceeded the suggested cutoff for clinical concern, meaning scores in this range are generally associated with clinically significant ASD (i.e., mild range; Figure 1). Follow-up analyses indicated that 59% of youth with SAD scored above the clinical threshold. Specifically, 6 parents reported scores in the mild range, 8 in the moderate range, and 6 in the severe range per the SRS-2 total raw score. For the ASD group, 68% of participants exceeded the clinical threshold suggestive of ASD. Specifically, 9 parents reported their child demonstrated scores in normal limits (32%) whereas 6 participants demonstrated scores in the mild range, 4 in the moderate range, and 9 in the severe range (see Figure 1). For the ASD+SAD group, only 4 participants demonstrated parent-reported SRS-2 total raw scores within normal limits (i.e., 16.67%). Follow-up analyses indicated that 83% of the sample with ASD+SAD scored above normal limits (mild range, n = 4; moderate range, n = 5; severe range, n = 11). Per the SRS-2 manual (Constantino & Gruber, 2012), the suggested cutoff is a raw score of 62 (i.e., score that optimizes sensitivity and specificity as a screener indicating the diagnosis may be present and further testing is warranted). In our sample, a raw score of 62 resulted in sensitivity of .67 and a specificity value of .50. As such, 67% of participants diagnosed with ASD demonstrated parent-reported SRS-2 score of at least 62. The specificity of the parent form of the SRS-2 indicated that 50% of participants without ASD obtained a score of at least score of 62².

Figure 1.

Percentage from each Diagnostic Group within the Recommended Guidelines for Interpretation of SRS-2 Total Scores

Discussion

Although previous studies have investigated the ability of the SRS/SRS-2 to differentiate ASD and SAD (e.g., Cholmkerey et al., 2014b), few studies have addressed comorbid presentations of ASD+SAD, even though this presentation may occur in as many as 40-50% of youth with ASD (van Steensel et al., 2011; White et al., 2009). Our study focused on a combined outpatient sample and examined ASD against SAD and their combination which were reliably established diagnostically. This study explicitly addressed the sensitivity and specificity of the SRS-2 in differentiating the diagnoses of ASD only, SAD only, or ASD+SAD in children and adolescents.

Overall, results for our primary aim suggest that both at the subscale level, as well as the SRS-2 total raw score, the SRS-2 does not adequately differentiate between the ASD only, SAD only, and ASD+SAD groups. Following a series of univariate ANOVAs with application of alpha corrections, the group differences were significant only for the SRS-2 Social Awareness and Social Motivation subscales. Per post-hoc contrasts, the ASD group demonstrated significantly higher (indicating greater impairment) scores in the areas of Social Awareness and Social Cognition, but not in the areas of Social Motivation, Social Communication, or RRB. This pattern suggests that some SRS-2 subscales are more discriminating of ASD than others (i.e., some evidence for sensitivity). Consistent with other research showing lack of ASD-specificity of the Social Motivation scale (e.g., Cholemkery et al. 2014b; South et al., 2017), we suggest that this scale reflects active social avoidance, which can stem from disinterest or from social anxiety (e.g., Capriola, Maddox, & White, 2016; Factor et al., 2017; White et al., 2009).

Lack of differentiation between the SAD only vs. ASD only groups for the Social Communication and RRB subscales is fairly novel. Past research by Bellini (2006) has found that for adolescents with ASD, poor social communication abilities were related to increased social anxiety. Specifically, impaired social skills (one component of social communication impairment) leads to negative peer interactions, which may ultimately lead to social anxiety. Social communication impairments, however, are not unique to ASD and are often present among youth with SAD (Halls, Cooper, & Creswell, 2015) as well as youth with other disorders (Towbin et al., 2005), which might explain the lack of significant differences between the SAD only and ASD only groups in our study. Our lack of findings are consistent with extant research by Hartley and Sikora (2009) who found that none of the stereotyped domain criteria determined ASD membership reliably. Previous works have noted that RRBs are equally present in anxiety disorders as ASD (Spiker et al., 2011). For example, anxious children reportedly engage in restricted interests (as a result of negative social experiences; Spiker et al., 2011), are more likely to use redundant, repetitive, and rigid hand movements, and demonstrate ritualized behaviors, all of which can be conceptualized as coping and anxiety-reducing strategies (Lang et al., 2015). Taken together, RRBs are equally present and interfering in SAD as ASD, even though they remain diagnostic for only the ASD group.

Regarding the effects of comorbid diagnoses, post-hoc contrasts demonstrated that the ASD+SAD group had significantly higher scores than the SAD group for the overall SRS-2 total raw score as well as three of the SRS-2 subscales: Social Awareness, Social Communication, and Social Motivation (i.e., evidence for sensitivity). At the subscale level, post-hoc comparisons demonstrated there was also marginally significant differences between the ASD+SAD group and the SAD group for the Social Cognition and RRB subscales. These results suggest that youth with comorbid ASD+SAD possess severe social deficits characteristic of ASD. TOST equivalence testing indicated the ASD+SAD group and the ASD only group were actually dissimilar on the SRS-2 total raw score (i.e., lack of discrimination between groups). However, the differences between the groups were marginally significant per TOST equivalence testing, suggesting that the current analyses may be underpowered to detect significance for a large effect. Only the Social Motivation subscale revealed differences between ASD+SAD and ASD groups; the comorbid group demonstrated higher raw scores than the ASD only group (i.e., evidence for discrimination). These findings suggest that co-occurring SAD compounds the social avoidance present in youth with ASD.

As noted by South and colleagues (2017), the SRS and SRS-2 offer promise as dimensional measures of ASD symptoms in research and have established diagnostic validity for screening purposes when compared to typically developing youth. However, our study found evidence for poor sensitivity and specificity of the SRS-2 in differentiating ASD and SAD. Sensitivity and specificity results indicated poor accuracy when using the clinically suggested raw score of 62 (as recommended by Constantino & Gruber, 2012). Our specificity findings are largely consistent with past evidence by Aldridge, Gibbs, Schmidhofer & Williams (2012) for poor specificity and an unacceptably high rate of false positives (i.e., 92% of participants in their sample without an ASD diagnosis fell within the ASD range on the SRS). As defined by Norris and Lecavalier (2010) and Cicchetti and colleagues (1995) sensitivity rates of 70-80% are the minimum standard for screening purposes, while specificity values should be ≥ 80%. This was not the case for our sample, as sensitivity and specificity values were lower and outside acceptable ranges. Our results suggest that the SRS-2 total raw score poorly discriminated when collapsed across the three diagnostic groups as well as when the ASD+SAD combined group was excluded. These results, albeit preliminary, highlight the need for caution when interpreting SRS-2 results given the SRS-2 total raw score did not differentiate sufficiently youth with SAD from those with ASD. This is especially critical given the SRS-2 is widely used across a variety of clinics, including comprehensive assessment clinics, but also disorder-specific clinics such as anxiety-specific clinics or ASD-specific clinics (e.g., Aldridge et al, 2012). Given the prevalent use of this measure and the poor sensitivity/specificity for disentangling ASD and SAD, it is recommended that additional screeners are needed to properly rule in/out ASD and SAD. Recent studies have also indicated that the SRS-2 may, instead, be best understood as a general screener used to indicate global distress or symptom severity (Korzeniewski et al., 2017) and may also be elevated even in the externalizing disorders (e.g., oppositional defiant disorder and conduct disorder, Cholemkery et al., 2014a, as well as general behavior problems, Hus, Bishop, Gotham, Huerta, & Lord, 2013).

Results of this study must be considered in light of several methodological limitations. Although our sample size meets minimum requirements posed by Kraemer (1992), we were likely underpowered in the current study to detect smaller effects. Although not unique to our study (cf, Factor et al., 2017), another noteworthy limitation is the relatively modest internal consistency in our sample for the Social Awareness subscale which might suggest that the subscale-level items are not measuring the same construct. This could potentially be attributed to the Social Awareness subscale’s high number of reverse-coded items as well as the smallest number of items relative to other SRS-2 subscales (i.e., half of scale items are reverse-coded). Although the authors of the SRS-2 encourage caution with interpreting subscales alone, we found that this limitation was not unique to our study (Cholemkery et al., 2014b; Factor et al., 2017; South et al., 2017). The majority of participants in this study were relatively cognitively high functioning, Caucasian, and within the ages of 7-17 years, therefore our findings might not readily generalize across diverse races and ethnicities, those with lower cognitive abilities, or younger children/adult populations. Lastly, we combined data from two separate samples to bolster sample size, therefore not all participants received a comprehensive psychological assessment. Although group-level differences were not related to any of the dependent variables, replication within well-matched samples is important for future research.

In spite of these limitations, our findings offer important implications for screening for possible ASD among youth who have symptoms of SAD. Only a few studies have examined the overlapping symptomatology of ASD and SAD (Cholemkery et al., 2014b; Tyson & Cruess, 2012). In our study, we also focused on differences among a combined outpatient sample which is a more rigorous test than examining specificity and sensitivity between ASD youth and typically developing youth. Results from our preliminary aim suggest that the SRS-2 misclassified many children and adolescents diagnosed with ASD, and participants with SAD were mistakenly misclassified as having social impairments indicative of ASD. Specifically, for the SRS-2 total raw score as well as at the subscale-level, youth with SAD did not significantly differ from those with ASD. Our findings also suggest that social motivation impairments associated with ASD are exacerbated by the co-occurrence of social anxiety. In sum, when using an ASD screener, like the SRS-2, researchers and clinicians should be cautious about risk of false positives. This is especially so when assessing higher functioning youth and those who present with symptoms consistent with SAD.