In this study, 77 children were evaluated at two times: initially around the age of 2 years, and then again around 4 years. Diagnostic stability based on DSM-IV clinical judgment was high, with 80% of children remaining in the same diagnostic category at follow-up; this figure is quite consistent with what has been reported in previous studies with smaller samples. Fifteen children (20%) changed from having a diagnosis on the autism spectrum to either having a diagnosis of a non-autistic developmental disorder or having no diagnosis. Also consistent with other studies, a diagnosis of Autistic Disorder was more likely to persist than a diagnosis of PDD-NOS: 85% of children who received an initial diagnosis of Autistic Disorder retained a diagnosis on the autism spectrum at age 4, whereas 47% of children initially diagnosed with PDD-NOS retained an ASD diagnosis. Similarly, fewer children with an initial diagnosis of Autistic Disorder moved to meeting criteria for no developmental disorder than children initially diagnosed with PDD-NOS (6% vs. 20%). No children who were diagnosed as non-ASD at age 2 were diagnosed with ASD at follow-up, as we did not expect to identify any children who demonstrated regression after initial screening. Of the 46 children who retained a diagnosis of ASD, only two moved from PDD-NOS to AD, whereas seven moved from AD to PDD-NOS. Thus, both in movement on and off the ASD spectrum, and movement within the spectrum, there was a general trend toward milder impairment between the ages of 2 and 4; symptomotology was more likely to improve than to worsen over time .
Movement between diagnostic categories between Times 1 and 2 evaluations
The exception to this trend was the emergence of repetitive behavior and resistance to change in some children. For the two children who moved from PDD-NOS to AD, both had no reported/observed problems with stereotyped/repetitive behaviors at the Time 1 evaluation but by the Time 2 evaluation began engaging in these types of behaviors, prompting the change to the AD diagnosis. Similarly, the children who moved from non-autistic to Autistic Disorder on the ADI-R generally did so because of the emergence of repetitive behaviors or resistance to change. In the studies summarized in , it is generally the case that studies that found a significant number of children moving onto the ASD spectrum used the ADI-R as a key diagnostic instrument, whereas the other studies found few or no children moving in this direction. Probably for this reason, the ADI-R had lower overall stability than the other three measures in the current study.
Although the children who gained an AD diagnosis generally did so because of the development of repetitive behaviors, the seven children who moved from an AD diagnosis to a diagnosis of PDD-NOS were heterogeneous in their behaviors and no clear pattern emerged to suggest a specific way in which symptoms were abating.
On the CARS, 2 children moved from having a score that was not on the spectrum at Time 1 to a score that was on the spectrum at Time 2. In each case, the Time 1 scores were within a few points of cutoff criteria for ASD and moved to just at cutoff criteria (from 26 to 30 and 27 to 30, respectively). This was not the case with the ADI-R; more children gained a diagnosis of autism for the specific reason that they developed repetitive and stereotyped behavior. Until a specific scoring algorithm is created for use with children in this age range, or scoring criteria for PDD-NOS are developed, it may be prudent to use the ADI-R in conjunction with other measures for children in this youngest age group.
Another factor that differentiated the ADI-R from the other instruments was that it was based on parent report only. The ADOS, in contrast, is based on direct child observation, whereas the CARS and DSM-IV-based clinical judgment incorporate information gleaned both from child observation and parent report. Although parents are generally excellent reporters, especially when asked about specific behaviors, and although parent report is crucial in providing information that may not be directly observable in an evaluation (e.g., reactions to peers), there were a few instances in this study where parents denied the existence of either a pathological behavior or a skill that was directly observed during the evaluation. Therefore, supplementing or modifying the ADI-R results with direct observation will result in a more accurate diagnosis. We conclude, therefore, that diagnostic stability of ASD diagnoses is high between ages 2 and 4, especially when using DSM-IV-based clinical judgment, ADOS, or CARS. Stability is also good for the ADI-R, except for children who do not meet criteria for Autistic Disorder because of few or no symptoms in the repetitive behavior domain. Were an algorithm for PDD-NOS to be derived from the ADI-R, stability would likely be high. When children move diagnostically, it is likely to be in a positive direction, both with and across the boundaries of the autism spectrum.
Sutera et al. (2007)
examined in detail the children who moved off the autism spectrum between ages 2 and 4, to determine whether child characteristics at age 2 could predict this diagnostic change. Contrary to expectations, the children who retained the ASD diagnosis were very similar to those who moved off the spectrum on most variables, including receptive and expressive language, non-verbal problem solving, and number and severity of DSM-IV autism symptoms. Motor development, however, both by parent report and child testing, was different, with the improved outcome children showing significantly better motor development at age 2.
Several limitations to the current study should be noted. First, a truly blind assessment at Time 2 would be preferable. This was not feasible in the current study, partly because parents sometimes requested to see the same clinician, and because feedback to parents involved comparing developmental and diagnostic results between the two assessments. However, developmental and ADOS testing was done by student clinicians who were generally blind to initial assessment results, and who followed specific behavioral rules for diagnosis. As reported by Ventola et al. (2006)
for the current sample of children, reliability between the ADOS, scored by the student clinicians, and the DSM-IV-based clinical judgment of the experienced clinician, was very high.
Another limitation was that the sample was drawn from Connecticut and surrounding areas in Massachusetts, Rhode Island, and New York. Early intervention services are more intensive in these areas than in many parts of the US and other countries. Future analyses will examine the relationship between type and intensity of intervention and outcome in our longitudinal sample, but the extent to which intervention played a role in the generally positive movement of the children in our sample is not yet known. In addition, the extent to which our findings can be generalized to other geographic areas is not clear.
A final question which is raised by the present findings is the youngest age at which a reliable diagnosis can be made. The mean age of our sample was 27 months, with children as young as 16 months; can reliable diagnosis be made earlier than that, and what percent of children as young as 16–17 months can be reliably diagnosed? To some extent, this will depend on the degree of developmental delay; we feel that children with all skills (motor, language, and cognitive) under a 12-month developmental level cannot be diagnosed with ASD using DSM-IV criteria. Should children under the age of 16–17 months chronological age, or 12 months developmental age, be diagnosed with ASD, using criteria developed specifically for this age group, or should the question of possible autistic development be raised without a firm diagnosis being made? These are questions for ongoing longitudinal, prospective studies of children at high risk for developing ASD to consider.