Verification of the utility of the social responsiveness scale for adults in non-clinical and clinical adult populations in Japan

According to a recent epidemiological study [1], autistic spectrum disorder (ASD) is currently estimated to be 1% of the adult population, a figure that approximates that in the child population [2]. Recently, ASD in adulthood has attracted considerable interest in the field of general psychiatry. It has been identified that most adults with ASD living in the community still had unmet clinical needs and are socially disadvantaged [1],[3]. In line with this worldwide trend, in Japan, an increasing number of adults with ASD visit psychiatric clinics with a diverse range of chief complaints, seeking either accurate diagnosis and a medical certificate needed to receive transition support for employment or treatment of concurrent psychiatric symptoms such as depression or anxiety [4]. However, unlike in children diagnosed with ASD, clinical manifestations in adult patients first diagnosed with ASD in adulthood are often complex: deficits in social reciprocity tend to be less apparent in adults with high-functioning ASD, especially outside situations that demand responses to complex social cues, or when adults with ASD mask their deficits using compensation strategies. For these reasons, it is difficult to identify ASD symptoms underlying adulthood-onset psychopathology and differentiate ASD from other psychiatric disorders in general psychiatric practice, which can lead to misdiagnosing ASD symptoms as psychosis [5].

In the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [6], a category of pervasive developmental disorders in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) [7] has been converted to a new category of ASD in which ASD severity is quantitatively rated according to current difficulties. Further, because of a nonexistent natural boundary between affected and unaffected individuals [8],[9], the DSM-5 has a new category of social (pragmatic) communication disorder for individuals with marked deficits in social communication but who do not otherwise meet ASD criteria (i.e., those with subthreshold ASD) [6]. The availability of quick, easy-to-use, and validated screening tools for identifying autistic traits and symptoms for psychiatric patients would help make appropriate diagnoses, reduce misdiagnoses, and plan appropriate treatment or support according to individual patient needs.

To date, the few self-report questionnaires available for identifying ASD in adulthood include the autism-spectrum quotient (AQ) [10]-[12], the Ritvo Autism Asperger’s Diagnostic Scale [13], or the Ritvo Autism Asperger Diagnostic Scale-Revised [14]. By contrast, a few questionnaires such as the Social Responsiveness Scale-Adult version (SRS-A) [15], or the Autism Spectrum Disorder in Adults Screening Questionnaire [16] must be completed by another adult (e.g., family member, close friend, or professional). The SRS-A was modified from the SRS [17], a quantitative measure of autistic traits in children. The original SRS has been extensively validated in clinical and subclinical child populations as well as in general child populations not only in the U.S. [8],[15],[17]-[19] but also in Europe [20], South America [21], and Asia [9],[22]. The SRS can distinguish children with ASD from children with any other or no psychiatric disorder and is generally unrelated to IQ in the normal range [9]. The SRS, a quantitative measure of autism, is also sensitive to autistic traits and symptoms even in subthreshold ASD conditions [9]. It is extremely useful for research purposes such as genetic epidemiological research [19],[23] or in research assessing brain-behavior relationships [24]. Its utility for detecting autism-related genetic loci [25] or cross-species research [26] has been suggested.

However, at this time, only a few validation studies of the SRS-A exist [15],[27],[28]. Therefore, the main purpose of this study was to determine the score distribution of the Japanese version SRS-A in a non-clinical Japanese adult population, and to assess its factor structure, reliability, and validity. Based on our findings in clinical populations with and without ASD, optimal cutoff scores are recommended.

This study provides some evidence supporting the continuous distribution of autistic traits in a non-clinical adult population using the Japanese version of the SRS-A, and the satisfactory reliability and validity of the Japanese version SRS-A for adults aged 19 to 59 years. The Japanese version SRS-A was shown to be capable of detecting ASD and autistic traits/symptoms among a psychiatric population and also screening for ASD. The finding of continuous distribution of autistic traits in a non-clinical adult population as measured by the SRS-A and its single factor structure is closely similar to what has been observed in children [9],[18],[20]. The SRS-A provides additional evidence about the nature of the autistic spectrum [6]. Mean SRS-A scores corresponded to mean parent-rated SRS scores in Japanese children [9]. The effects of gender or age on SRS-A scores among a non-clinical adult population in this study were overall minimal, being similar to previously reported parent-rated scores in a child population aged 7 to 15 years [9]. Only in adolescents did we observe a significant gender difference with a moderate effect size. However, this male-dominant difference found in adolescent participants could be explained by age-dependent rater bias. Regarding rater effects, we could examine these for adolescent participants only due to practical restraints of the collected data. Our results demonstrated that, father ratings were significantly higher than mother ratings for adolescents. In an examination of a twin sample, Constantino and Todd (2005) reported strong parent-offspring correlations of subthreshold autistic traits as measured by the SRS [27], indicating that autistic traits are strongly heritable for the pairing. According to that study, the father-offspring correlation was higher than that of the mother-offspring pairing, and that of the father-son pairing was the strongest at 0.58. It is unclear whether such father-son similarity in social responsiveness might have affected the extremely high father ratings of their adolescent sons in this study. Given that the special status of fathers as the rater has not been observed in the U.S. standardization sample [15], pp. 44, alternatively our finding might better be explained by Japanese fathers’ high expectations of their sons approaching adulthood (i.e., they are no longer children but also not yet independent adults). The interrelationship between rater type, gender, and age remains to be replicated by larger-scale studies in Japan and in other cultures. Taken together, when interpreting the information SRS-A provides, we must keep in mind various factors, especially rater type in terms of social expectations within sociocultural contexts.

From the viewpoint of cross-cultural comparison, the score distribution in this study is comparable to that of U.S. data [15],[27]. The mean scores in our non-clinical sample rated by various raters (19-59 years; 43.3 for men, 34.3 for women) are comparable to those rated by mixed raters (18-89 years; 42.2 for men, 38.8 for women) reported in the SRS-2 Manual [15], pp. 44. As for spouse ratings of participants in early adulthood or middle age, the mean scores in our non-clinical subsamples (33.1, male 48.8%) were very similar to those of U.S. data rated by spouses (30-55 years; 31.7 for men, 30.0 for women) [27]. However, the scores in our sample were lower compared to those of German adults with typical development (19-79 years, 55.5 for mixed sex) as reported by Bölte [28]. The reason for this discrepancy between Bölte’s and our scores is not clear because rater-type details were not mentioned in Bölte [28]. As emphasized in the SRS-2 Manual [15] as well as in Bölte [28], the effect of rater type, which varies depending on an adult’s age, gender, or living situation, might be crucial and should be systematically studied in future research.

Regarding convergent validity, correlations between the SRS-A and ADOS or PARS ranged from moderate to relatively strong (the latter two of which were assessed based on direct or indirect clinician observation), and these correlations provide support that the Japanese version SRS-A measures the same clinical aspects of the autism spectrum as do validated measurements. By contrast, the weak correlation between the SRS-A and AQ-J might be because the AQ-J is self-rated, and suggests that these two questionnaires might measure different aspects of the autistic spectrum.

Although the Japanese version SRS-A capably discriminated adults with ASD from those without ASD but having any other psychiatric diagnosis such as major depressive disorder, schizophrenia, or bipolar disorder, we observed no gap but rather an overlap in the score distribution between the two clinical groups (with and without ASD) in our study. This finding is consistent with that observed for school-age children [9], although the non-ASD child clinical population in that previous study [9] included adjustment disorder, attention-deficit/hyperactivity disorder (ADHD), and anxiety and other disorders, making it more diverse than the non-ASD adult clinical population in the present study. Recent genetic, molecular, and cytologic research highlights shared contributory mechanisms between ASD and major adult-onset psychiatric disorders (i.e., major depressive disorder, bipolar disorder, schizophrenia [34], and behavioral-cognitive commonalities [5],[35]). Further, concurrent depression and anxiety symptoms, which are likely accompanied by transient psychotic symptoms, are found not only in individuals with ASD but also in those with subthreshold autistic symptoms [36],[37]. Given this transdiagnostic commonality, the overlap in the SRS-A score distribution in the present study suggests that a proportion of the non-ASD clinical population might have autistic traits/symptoms despite having subthreshold ASD, which can lead to clinical difficulties in differential diagnosis. Because such clinical uncertainty from concurrent psychiatric symptoms is likely to result in misdiagnosis that overlooks ASD, our result that the SRS-A has discriminative ability for ASD with high sensitivity would prove the clinical usefulness of the instrument as a secondary screening tool in psychiatric practice. From a therapeutic viewpoint, it is important to detect autistic symptoms of not only threshold ASD but also subthreshold autistic conditions among various psychiatric populations so that appropriate treatment based on a comprehensive clinical evaluation can be given [36],[37]. To this end, we recommend cut-off scores of 65 for men and 52 for women, which are similar to those for Japanese children [9], even though there are some adults who do not meet the diagnostic criteria of ASD above the cut-off.

This study has several limitations. First, our sample size was small, from which we examined a subsample using validated instruments measuring autistic traits and severity or IQ. The non-ASD clinical group mainly included participants diagnosed with schizophrenia and depressive or bipolar disorders, although other various psychiatric comorbidities are also common in adults with ASD, notably anxiety disorder and ADHD [38]. Replication in a larger psychiatric population including anxiety disorder and ADHD is needed. Second, our normative sample did not include individuals aged 60 or more for whom the SRS-2 manual gives higher scores [15], pp. 44. Third, we did not examine inter-rater agreement, which can explain differences due to rater type found in this study as well as test-retest reliability. Fourth, we did not examine the self-report SRS-A. A comparison between other-report and self-report questionnaires would add evidence about rater type in measuring this kind of behavior [39],[40].

This study replicated the original SRS-A study in a Japanese population and extended previous studies on the child version of the SRS to an adult population. That is, the SRS-A distributed continuously in the non-clinical population, and the other-report SRS-A rated by parents, spouses, siblings, and close friends was found to be reliable across gender and age, except in the youngest men aged 19 to 24 years. Furthermore, the SRS-A is useful for detecting ASD and autistic traits/symptoms among psychiatric patients and also for capably discriminating ASD from non-ASD psychiatric disorders such as major depression and schizophrenia. We have recommended optimal cutoff scores feasible for use in clinical settings.