![[ Visit Client Website ]](images/banner.JPG)
Saturday, May 9, 2009: 1:20 PM
Northwest Hall Room 1 (Chicago Hilton)
Background: The Autism Diagnostic Observation Schedule (ADOS; Lord et al. 2000) assesses communication, social and play skills, and restrictive/repetitive behaviours in the diagnosis of ASD. Recently, Gotham et al. (2007, 2008) revised the algorithms for Modules 1, 2 and 3 to improve diagnostic validity and to allow for increased comparability across modules. Two algorithms were derived for Module 1 (“no words,” “some words”), two for Module 2 (“younger than age 5,” “age 5 years and older”), and one for Module 3 as dividing this sample by age or language level did not result in more homogeneous groups.
Objectives: To assess agreement on ASD sub-types (autism/ASD) based on the WPS-published algorithms relative to the revised algorithms in a sample of preschool children with ASD participating in a Canadian longitudinal study.
Methods: ADOSs were administered on enrolment (Time 1 n=329; 262 Module 1, 63 Module 2, 4 Module 3) and again one year later (Time 2 n=171; 84 Module 1, 69 Module 2, 18 Module 3). Classification as autism, non-autism ASD (hereafter ASD) and non-ASD was determined with the WPS and revised algorithms at both time points. Crude agreement and kappa values (that take chance into account) were calculated.
Results: At T1, the frequency of autism and ASD was 82% and 18% using the WPS algorithms and 91% and 9% using the revised algorithms. A few children (n=10) shifted from autism to ASD or from ASD to non-ASD when the revised Module 1 “no words” algorithm was applied. However, for all other algorithms, if a child changed diagnostic categories when the revised versus WPS algorithm was used, the shift was toward a more severe diagnosis (ASD to autism, non-ASD to ASD/autism) both at T1 and T2. Kappa (agreement between algorithms) at T1 was .40. At T2, the frequency of autism diagnoses decreased for both algorithms relative to T1. However, it was still higher for the revised (85%) algorithms compared to the WPS (71%) algorithms. Kappa at T2 was .52. ADOSs were administered at T1 and T2 for 169 children, allowing examination of change in diagnostic categories. In this sub-sample, probability of a diagnosis of autism was higher at T1 and remained high at T2 for the revised algorithms relative to the WPS algorithms. Agreement in an autism classification between T1 and T2 was 87% for the WPS algorithms and 90% for the revised algorithms. Given the high rate of autism, weighted kappas were computed.
Conclusions: The revised algorithms were intended to provide a more accurate differentiation of autism/ASD from non-ASD children by being less sensitive to variability in age and language ability. This change is accompanied by a shift in diagnostic classification when the revised versus the WPS algorithms were applied, from a less (ASD) to more severe (autism) diagnostic category. This has implications for researchers who may wish to assess phenotypic heterogeneity within the ASD continuum or to measure change over time in diagnostic status. The revised algorithms may be less sensitive to heterogeneity and change over time in ASD behaviours.
Objectives: To assess agreement on ASD sub-types (autism/ASD) based on the WPS-published algorithms relative to the revised algorithms in a sample of preschool children with ASD participating in a Canadian longitudinal study.
Methods: ADOSs were administered on enrolment (Time 1 n=329; 262 Module 1, 63 Module 2, 4 Module 3) and again one year later (Time 2 n=171; 84 Module 1, 69 Module 2, 18 Module 3). Classification as autism, non-autism ASD (hereafter ASD) and non-ASD was determined with the WPS and revised algorithms at both time points. Crude agreement and kappa values (that take chance into account) were calculated.
Results: At T1, the frequency of autism and ASD was 82% and 18% using the WPS algorithms and 91% and 9% using the revised algorithms. A few children (n=10) shifted from autism to ASD or from ASD to non-ASD when the revised Module 1 “no words” algorithm was applied. However, for all other algorithms, if a child changed diagnostic categories when the revised versus WPS algorithm was used, the shift was toward a more severe diagnosis (ASD to autism, non-ASD to ASD/autism) both at T1 and T2. Kappa (agreement between algorithms) at T1 was .40. At T2, the frequency of autism diagnoses decreased for both algorithms relative to T1. However, it was still higher for the revised (85%) algorithms compared to the WPS (71%) algorithms. Kappa at T2 was .52. ADOSs were administered at T1 and T2 for 169 children, allowing examination of change in diagnostic categories. In this sub-sample, probability of a diagnosis of autism was higher at T1 and remained high at T2 for the revised algorithms relative to the WPS algorithms. Agreement in an autism classification between T1 and T2 was 87% for the WPS algorithms and 90% for the revised algorithms. Given the high rate of autism, weighted kappas were computed.
Conclusions: The revised algorithms were intended to provide a more accurate differentiation of autism/ASD from non-ASD children by being less sensitive to variability in age and language ability. This change is accompanied by a shift in diagnostic classification when the revised versus the WPS algorithms were applied, from a less (ASD) to more severe (autism) diagnostic category. This has implications for researchers who may wish to assess phenotypic heterogeneity within the ASD continuum or to measure change over time in diagnostic status. The revised algorithms may be less sensitive to heterogeneity and change over time in ASD behaviours.