This study set out to develop and test a short self-evaluation questionnaire for assessing ASD in adults without intellectual disability. The RAADS-14 Screen, derived from the RAADS-R, showed good psychometric properties, and takes only a few minutes to complete, compared with up to 1 hour for the original RAADS-R. Notably, individuals with ASD scored high in all three domains comprising the RAADS-14 Screen, showing mentalizing deficits, increased social anxiety, and sensory oversensitivity, thus capturing characteristics typical of ASD. Thus we suggest the RAADS-14 Screen is a useful tool for screening adult psychiatric outpatients for an unrecognized ASD, which often exists comorbidly with other disorders such as major depression, obsessive-compulsive disorder, ADHD, and various anxiety disorders [2, 3].
The RAADS-14 Screen shows excellent discrimination abilities in the non-psychiatric population. Furthermore, because of its established capability to distinguish ASD from OPD, which may obscure the clinical picture, the RAADS-14 Screen also has extended screening utility. In the final sample, a cut-off score of 14 or above in the RAADS-14 Screen correctly identified 97% of the participants with ASD and excluded an incorrect ASD diagnosis in 46 to 64% of patients with ADHD or OPD. However, as described in the Introduction, some patients with OPD may also have undiagnosed ASD; it is therefore likely that some participants in the psychiatric control groups should have been diagnosed with ASD. If this is the case, the aforementioned specificity of 46 to 64% may be an underestimate.
Although the short versions of AQ are equally good at discrimination in the non-psychiatric population, they have not been tested as to whether they are useful for distinguishing ASD from overlapping psychiatric disorders. Moreover, the different versions of AQ aim to assess autistic traits in the general population, whereas the RAADS-14 Screen aims to spot ASD. Evaluations of the abridged versions of AQ have resulted in mean scores reaching around 50% of the total AQ score in a normal population sample for the 28 item version, and 28% in the 10 item version [6, 7], whereas healthy controls only reach a mean score of 9% of the total RAADS-14 Screen score. In addition, sensory reactivity, included in the diagnostic criteria for DSM-5, is not included in the AQ 28 item version, and comprises only one item in the AQ-10 version, compared with three items in the RAADS-14 Screen.
The first five items in the RAADS-14 Screen could serve as initial rapid screen. A cut-off score of 4 or above out of a possible maximum score of 15 correctly diagnosed 93% of the patients with ASD, while 45 to 49% of the psychiatric controls were (presumably) correctly excluded as having ASD. These first items may provide sufficient information when time is very limited, or when the patient is reluctant or unable to respond to many questions.
Out of 135 patients with ASD in phases II and III, 9 scored below the cut-off for ASD, including 2 individuals that scored zero points. As it seems highly unlikely that a person with ASD would not have any of the core symptoms mentioned in the questionnaire, other reasons such as lack of insight, misinterpretation, or unwillingness to endorse any of the statements are more likely. This highlights the benefits of having a clinician present during the completion of the questionnaire, both for explaining the statements to the patient if needed, and for assessing the congruity between the observed symptoms and the responses in the questionnaire.
Endorsement patterns in the different diagnostic groups
For validity, based on the magnitude, pattern, and significance of difference of the items between the ASD group and the clinical control groups, the RAADS-14 Screen has good validity for criteria and convergence. With regard to the ADHD group, the differences were smaller, although still clinically meaningful. This lower difference could be an indicator of some items being discriminators for early-onset neurodevelopmental disorders in general, rather than ASD specifically. It could also be due to the fact that many individuals with ADHD have co-existing autistic traits [12, 13].
The items in the social anxiety domain overlapped with symptoms suggestive of social anxiety disorder. The responses to two of these items (I often don’t know how to act in social situations’ (item 5) and ‘I can chat and make small talk with people’ (item 6)) were similar between the patients with social anxiety disorder and the ASD group. People with ASD exhibit deficits in social skills, and often experience peer rejection, which may induce social anxiety. In addition, a large difference in the total RAADS-14 Screen scores meant that the ASD and social anxiety disorder groups were clearly distinguishable. Compared with patients with other anxiety disorders, the ASD group scored higher on all the social anxiety items, indicating to the importance of taking all domains into consideration when using the RAADS-14 Screen.
Based on Cronbach’s alpha, the internal consistency of the factors in the identified three-factor structure exceeded 0.7 for all clinical groups, the recommended minimum level for group comparisons in clinical studies . These factors were characterized as mentalizing deficits, sensory reactivity, and social anxiety. In the DSM-5 for ASD , the A criteria ‘Persistent deficits in social communication and social relatedness across contexts’ include descriptions that are congruent with eight items in the mentalizing deficits and social anxiety domains on the RAADS-14 Screen. The B criteria ‘Restricted, repetitive patterns of behavior, interests, or activities’ correspond with a total of five items in the mentalizing deficits and the sensory reactivity domains. Stereotypies is the only DSM-5 item that is not covered by the RAADS-14 Screen; however, self-reported stereotypies may not be a valid option for estimating their presence, because individuals are often unaware of having them. Together, these results suggest that the RAADS-14 Screen will be useful when diagnosing people according to the DSM-5 criteria for ASD (Table 3).
The mentalizing deficit domain of the RAADS-14 Screen includes all the items from the original RAADS-R circumscribed interests’ domain, half of the items from the social relatedness domain, and one language item. The items from the circumscribed interests RAADS-R domain were all related to an ability to change from an internal to an external focus, which fits well into a domain of mentalizing deficits. In addition, four of the social relatedness items, relating to understanding social codes and reading body language, and one language item assessing the mentalizing skill of making a distinction between the literal and intended meanings of sentences, fit nicely into the new mentalizing deficit domain. Another four social relatedness items and one language item describe feelings of social awkwardness and inabilities, thus they formed the second domain, social anxiety. Finally, the sensory motor domain of the RAADS-R was not compromised by the new factor structure, but none of the pure motor function items in the RAADS-R was sufficiently discriminative of ASD to be included in the RAADS-14 Screen. Although poor gross motor skills are common in ASD, it is by no means specific to any psychiatric disorder, and children with emotional, behavioral, and pervasive developmental disorders often exhibit gross motor problems .
In the non-psychiatric disorders group, females had fewer mentalizing deficits and social anxiety than males. Notably, a reverse trend was found in the ASD group, with females scoring higher than males in the mentalizing deficits domain. Both findings are consistent with the Swedish RAADS-R validation study . The former findings support the extreme male brain theory for autism, that, in general, males have more cognitive autistic traits than females . The reason for poorer mentalizing skills in females with ASD compared with males is intriguing. Possibly, females with ASD have greater insight into their behaviors, and thus endorse autistic symptoms more readily than the males. The scores of the psychiatric controls in the mentalizing deficits and social anxiety domains were independent of gender. However, females scored higher than males in the sensory reactivity domain, across all groups, supporting the gender differences in sensitivity to noise and touch that have been reported in earlier studies [18, 19].
Some methodological limitations should be noted. Selection bias is always a crucial issue in clinical research. In order to include a representative population of patients, a desirable design for a validity study would be to collect data from all psychiatric patients visiting a clinic in a certain time period; however, this design does not protect against selection bias. For prospective research studies, informed written consent from the patient is a prerequisite, and this may result in attrition due to a reluctance to participate . Further selection bias could be related to the variability in frequency of consultations; some patients rarely visit the clinic even if they are severely ill, whereas others are frequent visitors. In the present study, selection bias is difficult to estimate because the collection of patient data was performed by a number of clinicians across the country, and through web pages directed towards people with psychiatric diagnoses. However, the option to respond via a web survey may enable responses from patients who, for different reasons (such as poor executive skills), rarely visit the clinic, thus possibly improving the representativeness of our sample. The inclusion of a number of highly specialized psychiatric clinics was also chosen to broaden the spectrum of patients compared with a general psychiatric clinic. Another limitation is the lack of confirmatory assessments to confirm or exclude diagnoses, but Swedish clinical practice is to use reliable diagnostic instruments. The truthfulness of the web respondents is supported by the fact that their scores were not different from those of the clinical samples. The 16 suspected outliers in the non-psychiatric group indicate that some of these subjects actually have an undiagnosed psychiatric disorder. Although a weakness of the study, it makes it plausible that the specificity in this group is even better. Although the ASD sample was not matched for gender, age, or intelligence in the comparison samples, the age distributions of all three psychiatric samples were roughly the same. Moreover, 10% of the participants did not state their gender; this may have had an influence on the gender difference results. Finally, as the DSM-5 was introduced after the completion of the current study, the RAADS-14 Screen was inevitably validated in psychiatric patients diagnosed according to the DSM-IV-TR. However, all items included in the RAADS-14 Screen are applicable for the DSM-5 ASD criteria, suggesting a utility for the RAADS-14 Screen beyond DSM-IV. The RAADS-14 Screen print-out version with instructions to the clinician is available as an Additional file 1.