The Diagnostic and Statistical Manual of Mental Disorders, or DSM, has long been the tool most often used to render a diagnosis of autism in the United States. The upcoming version of the DSM is already drawing criticism in regard to the diagnosis of Asperger's Syndrome (AS). The current proposal will group all of the autistic disorders under one umbrella term, Autism Spectrum Disorders (ASDs), which would be accompanied by a severity scale. Though a strong argument can be made for considering AS a separate diagnostic category, as the needs of persons with it are different than the needs for those more severely affected, AS is already considered an ASD and the proposed change will likely have little impact.
Perhaps the most pressing concern about the DSM is that the reliability of diagnoses rendered by different clinicians is often poor. For ASDs, and many other mental health disorders, structured tools have been developed to aid in rendering valid diagnoses. The best diagnostic tool for diagnosing autism is thought to be the Autism Diagnostic Observation Schedule (ADOS). Administering the ADOS consists of exposing a child to a series of tasks and a diagnosis is rendered based upon how the child performs on these tasks. There are clear guidelines for scoring the child's performance on these tasks and, in general, the ADOS is considered a reliable means of obtaining a valid diagnosis. Using the ADOS is far superior to a clinician's use of the DSM. That said, the ADOS is still a clinician's impression that is translated into a diagnosis.
However, two recent studies offer some encouraging results that suggest an objective diagnosis of an ASD may possible at some point in the not too distant future. The first of these involved recording the vocalizations of children with a small device that could be attached to the child's clothing (Oller et al., 2010). The recordings were then analyzed using software that separated the child's vocal utterances from other noises. The vocal utterances were subsequently broken into "syllables." These syllables were then examined by analyzing the ratio of "syllables" to vocalizations according to 12 acoustic features. Recordings were collected for a total 232 children ranging in age from about a year and half to 4 years of age. Just over 100 of the participants were typically developing, 49 were language delayed, and 77 of them had autism. The results indicated that the research was able to detect vocal development in that the analyses predicted the age of the typically developing children. Next, the study was able to detect the impaired development of the language delayed children. That is, they were noted to have progressed in vocal skills but were behind their typically developing peers. The children with autism, however, showed little of the expected developmental progress. These findings are preliminary and require replication, preferably by an independent group of researchers. However, it is promising that objective, automated recordings of the vocalizations of children with autism could be differentiated from the vocalizations of typically developing children. Whether this technique would allow reliable detection of the presence of autism remains to be seen.
The second study involved analyses of the brain images obtained via MRIs of adults with an ASD and compared these to adults who were typically developing (Ecker et al., 2010). Numerous studies have identified neuroanatomical differences in persons with ASDs and this study set out to analyze measures of shape (e.g., folds of the brain) and volume (e.g., cortical thickness). All of the subjects with an ASD by a trained clinician and that diagnosis was confirmed with either an ADOS or an ADI which is structured interview form based on the ADOS. All of these participants could be fairly described as being on the less severely affected end of the autism spectrum. Images were obtained and analyzed imaging analysis software and 5 morphometric parameters measuring either "shape" or "volume" were obtained for both hemispheres. These measures were then further analyzed using an algorithm that was constructed to classify each subject as having an ASD or as being part of the control group. This procedure resulted in 85% of the persons with an ASD being correctly placed in the ASD group when all of the parameters were considered. This success rate increased to 90% if only the results from the left hemisphere were considered. Again, a very encouraging result, which corroborates research into neuroanatomical differences in persons with ASDs, that requires further investigation. Additionally, given that the study was performed with adults, it is unclear whether this technique would be helpful in detecting ASDs in children.
It might not be time to throw out the DSM just yet but that day may come. On the other hand, I'd argue that sound diagnoses would be better informed if clinicians used the ADOS or ADI to render a diagnosis. Nonetheless, the day may arrive soon when a more time efficient means of diagnosing ASDs is available.
Ecker et al. (2010). Describing the brain in autism in five dimensions-Magnetic resonance imaging-assisted diagnosis of autism spectrum disorder using a multiparameter classification approach. The Journal of Neuroscience, 30, 10612-10623.
Oller et al. (2010). Automated vocal analysis of naturalistic recordings from children with
autism, language delay, and typical development. Psychological and Cognitive Science: Developmental Biology, www.pnas.org/cgi/doi/10.1073/pnas.1003882107.