Describing the Profile of Diagnostic Features in Autistic Adults Using an Abbreviated Version of the Diagnostic Interview for Social and Communication Disorders (DISCO-Abbreviated)

Although many individuals do not receive a diagnosis of ASD until late adolescence or adulthood, relatively little is known about the profile of individuals who are first diagnosed with ASD as adults. In the current study, comparison of individuals diagnosed in adulthood and a sample of children revealed that the profiles in the two groups were similar overall, but with some distinctive differences.

The frequency of ASD features was generally lower in the adult sample, with the exception of a few items, such as ‘does not share in others’ happiness’ and ‘talks about repetitive themes’ (Fig. 3). Those items with higher frequency in the adult sample included behaviours less likely to be present in young children; for example, ‘does not share in others’ happiness’ is not coded in children younger than 7 years.

Both the adult and child samples exhibited more impairments in the social-communication domain of DSM-5 compared with the domain of restricted and repetitive patterns of behaviour (RRB). Moreover, within the social-communication domain, more difficulties were reported for both children and adults in socio-emotional reciprocity (A1) and in deficits in developing and maintaining relationships (A3) compared with non-verbal communication (A2). These effects were more pronounced in the adult sample, and were reflected both in terms of the percentage of items and in the percentage of individuals meeting threshold (Fig. 1). The relative lack of impairment in non-verbal communication has diagnostic significance. The DSM-5 criteria specify impairment in all three social-communication subdomains; the relatively good non-verbal communication skills within this adult sample could therefore mean that they do not qualify for a diagnosis of DSM-5 ASD. By contrast, difficulties in non-verbal communication are represented in two distinct subdomains within different domains of the ICD-10/DSM-IV-TR criteria; moreover, it would be possible to receive a diagnosis according to the ICD-10/DSM-IV-TR criteria without demonstrating impairment in either of these subdomains. Consequently, relatively good non-verbal communication skills in these areas would be less of a barrier to diagnosis according to ICD-10/DSM-IV-TR.

Evidence of difficulties in social interaction relative to both communication difficulties and RRBs supports findings from other studies involving adults, but who were diagnosed in childhood. For example, Billstedt et al. (2007) reported a higher proportion of difficulties related to social interaction compared with communication, RRBs, and emotional problems/maladaptive behaviour. In contrast, Shattuck et al. (2007) found that adults had more difficulties with non-verbal communication than younger individuals. Evidence of difficulties in non-verbal communication was not replicated in the current study, both when focusing on the most severe manifestation of behaviours (‘ever’ codes) and when examining the current profile (see Supplementary materials). These findings suggest that in a sample who were predominantly diagnosed as adults, non-verbal communication difficulties had never been as marked as difficulties with socio-emotional reciprocity and in developing and maintaining relationships.

Within the RRB domain, there was a slightly different profile of behaviour in the child and adult samples. Children had more behaviours in the subdomains relating to repetitive motor movements, use of speech, or objects (B1) and restricted, fixated interests (B3), both compared with the other subdomains and in comparison with adults. Adults had a higher percentage of behaviours in the subdomains relating to insistence on sameness/inflexible routines or rituals (B2) and restricted, fixated interests (B3) compared with both repetitive motor movements, use of speech, or objects (B1) and hyper- or hypo-reactivity to sensory input (B4).

The differing patterns of behaviour within the RRB domain are consistent with known developmental changes in two types of RRBs (e.g. Evans et al. 1997; Uljarević et al. 2017). Behaviours described in B2 and B3 have previously been described as higher-level RRBs, whilst behaviours described in B1 and B4 may be considered lower-level and more characteristic of children and lower-ability individuals (e.g. Prior and Macmillan 1973). Consistent with this argument, the child sample had a higher proportion of behaviours relating to stereotyped or repetitive movements (B1) than the adult sample. Moreover, the only subdomain in which adults had a higher frequency of items than children was B2 (insistence on sameness/inflexible routines or rituals). As such, the profile of RRBs seen in the child and adult samples are somewhat consistent with what might be expected for the two age-groups.

Despite these differences in the patterns of behaviour, the percentage of adults and children meeting the thresholds within each RRB subdomain (at least one behaviour present) was more similar. The greatest differences between the two samples were in B3 and B4. Although adults had a relatively high mean percentage of items in the subdomain relating to highly fixated interests (B3), this is the RRB subdomain with the highest threshold (see Fig. 1) and was, therefore, the subdomain in which the fewest adults met the threshold. Moreover, despite the relatively low proportion of behaviours relating to hyper- or hypo-reactivity to sensory input in both groups, the percentage of adults and particularly children who met the threshold for this subdomain was relatively high (75% for adults and 97.2% for children), indicating that the majority of individuals had at least one sensory symptom. Although this finding is consistent with Billstedt et al. (2007), the percentage of adults with at least one sensory symptom was lower in the current sample (75% compared with 93% as reported by Billstedt et al.). This finding may, again, be related to the nature of the sample, the majority of whom were diagnosed as adults, although this interpretation must be viewed with caution due to the limited size of the sample.

Given that the DSM-5 criteria require difficulty in just two of the four RRB subdomains, the potential implications of the differing RRB profiles for the child and adult samples for diagnosis are less striking initially than for the social communication domain. These differences do, however, highlight differences in the types of RRBs that might be expected in individuals presenting for diagnosis as adults. Rather than a focus on lower-level RRBs, these findings suggest that adults are more likely to present with difficulties relating to a lack of flexibility in their behaviour as demonstrated by adherence to routines and rituals, as well as fixated interests.

Exploration of the behavioural profile at the level of individual items identified some key behaviours that remained salient in the adult profile and could, therefore, be of significance for the identification of ASD in adults. While these were predominantly social–communication items, two additional behaviours from the RRB domain were present in over 50% of adults; these were a ‘limited pattern of self-chosen activities’ and ‘collects objects’. Some of the strongest items in adults were part of the ‘signposting set’, a set of items identified as being highly discriminating for children with ASD relative to those with language impairment or intellectual disability (Carrington et al. 2015). Although adults on average had fewer of the behaviours described in the signposting set than children, they still, on average, scored on over half of the items, indicating that this item set may still have potential in signposting when referral for ASD may be appropriate in adulthood.

When considering the potential implications of the findings from this study for the diagnosis of adults, it is important to consider whether the potential barriers to diagnosis that have been identified are from the DSM-5 criteria or from the interview used. Although the DISCO is a comprehensive interview designed to obtain a broad and detailed developmental history, the DISCO Abbreviated includes only a subset of items from the full interview. While those items were identified based on their predictive validity, the tool was developed and tested using a sample consisting predominantly of children. Moreover, those adults and adolescents included in the test samples for the abbreviated algorithm had been diagnosed in childhood or adolescence. It may, therefore, be necessary to identify further items for inclusion within the DISCO Abbreviated that may be more characteristic of individuals who do not seek diagnosis until adulthood, such as more items focusing on the types of RRBs that appear to be more characteristic of this adult sample. Nevertheless, the finding of fewer non-verbal communication difficulties in adults is consistent with the results reported by Billstedt et al. (2007), who used the non-abbreviated DISCO and also included items that were not part of the diagnostic algorithms. While further development of the DISCO Abbreviated and other interview tools will be important in supporting the diagnosis of adults, the DSM-5 specification that diagnosis is dependent on the presence of non-verbal communication difficulties may present a barrier for those seeking diagnosis in adulthood.

The majority of participants in the adult sample did not bring an informant with them for the interview with the DISCO Abbreviated. Although clinical guidelines for diagnosis recommend that a detailed developmental history should be conducted with an informant who has known the individual throughout their life and is able to comment on their behaviour, both currently and during childhood (National Institute for Health and Care Excellence 2011), such informants are not always available for adults seeking a diagnosis. Where they are able to attend, their memory of events during the individual’s childhood may lack detail. While these interviews can be done with the individual themselves, as they have been in the current study, they may lack detailed and accurate memories of their early development, or, for a subset of individuals, may not have detailed insight into their own difficulties. The inclusion of both self- and other-informant assessments in the current study was intended to better reflect the diagnostic process within adult services, and addresses a recognised gap in the literature (Mandy et al. 2018). Nevertheless, the limited number of participants within each of the two groups (with informant and self-informant) prevented formal comparison of the profile of behaviour described in these different interview approaches; however, the frequency with which items were endorsed by self-informants was comparable with those who brought an informant with them (see Fig. S5). Moreover, when total scores were calculated, using the algorithm, none of the overall subdomain or domain scores differed. Further investigation of potential differences, including comparison of self- and other-informant interviews for the same individual, will enable identification of behaviours that may be under- or over-reported by different informants.

Both the child and adult samples were diagnosed according to the ICD-10/DSM-IV-TR criteria. All individuals in the child sample had a diagnosis of Autistic Disorder or Childhood Autism. By contrast, the adult sample also included individuals with a diagnosis of Asperger Syndrome. While this diagnostic discrepancy could potentially account for differences observed in the behavioural profiles of the two groups, it may be a genuine reflection of the nature of these two samples. Evidence from children diagnosed with an ASD in the UK suggests that those with Asperger Syndrome are, on average, diagnosed later than those with autism (Crane et al. 2016; Howlin and Asgharian 1999). As such, a higher rate of Asperger-like presentations may be expected in those seeking diagnosis as adults. Nevertheless, the child sample selected for the current study was selected on the basis of a diagnosis of Autistic Disorder or Childhood Autism, and as such, it would be important to also draw comparisons between the adult sample, and children who fit the ICD-10/DSM-IV-TR criteria for Asperger Syndrome.

Many adults with ASD may have learned to camouflage certain areas of difficulty, for example, by learning to make eye contact during conversations and using pre-prepared social scripts. Moreover, many adults report learning to suppress repetitive motor mannerisms, particularly whilst interacting with others. While such techniques may have enabled them to function more effectively within complex social environments—thus potentially accounting for the delay in seeking support—the use of these techniques may also lead to individuals under-reporting their own difficulties, or to others underestimating the occurrence and impact of those difficulties. Social camouflaging is thought to be more pronounced in females than males with ASD (e.g. Hull et al. 2017; Lai et al. 2017), which could, therefore, impact on the relative rate of diagnosis for males and females. The DISCO interview, however, includes questioning to check for camouflaging and learned or rehearsed strategies. Coding is applied only with respect to what would be natural behaviour when no strategies are in place. Self-informants are also likely to be aware of their use of strategies to discuss in the interview. This might have mitigated against finding a gender difference in this study. While the current sample included too few females (n = 28) to draw meaningful comparisons with males (n = 52), a growing body of research has been focusing on the so-called ‘female profile’ of ASD. For example, it has been suggested that in young children, repetitive behaviours or circumscribed interests around dolls may be misinterpreted as pretend play, while older females with ASD may exhibit apparently benign repetitive behaviours, such as constantly reading a specific set of books to the detriment of social interaction (Halladay et al. 2015). Such behaviours may be undetected, potentially contributing to reports of lower levels of repetitive behaviours in females than males with ASD (e.g. Mandy et al. 2012). As such, it will be important to identify behaviours characteristic of adult females with ASD to facilitate their diagnosis and subsequent access to appropriate support.

The current study is the first use of an abbreviated form of the DISCO as a standalone clinical interview, and thus represents an important step in facilitating diagnosis within a more clinically feasible time frame. The findings indicate both similarities and differences between the child and adult profiles of ASD, with the differences highlighting potential barriers to diagnosis according to DSM-5 criteria for higher ability adults, which may also have prevented them from being diagnosed as children. A key difference between the child and adult samples was the lower rate of non-verbal communication difficulties in the adult sample. This relative strength in adults could represent a particular barrier to diagnosis according to the DSM-5 criteria that would not be as pronounced for the ICD-10/DSM-IV-TR criteria. In cases where a lack of impairment in non-verbal communication is the only contraindication of an ASD diagnosis, it may be prudent, therefore, to consider ‘relaxing’ the rules for the social-communication domain of DSM-5 to require impairment in just two of the three subdomains. Furthermore, although the current study has found that a ‘signposting set’ of items with high predictive validity in children also has the potential to signpost diagnosis in adults, it will be important to identify behaviours that are more characteristic of this population of able individuals with ASD, who may be missed in childhood, to facilitate their diagnosis and access to support at an earlier age.