![[ Visit Client Website ]](images/banner.JPG)
Objectives: To develop a composite index, optimized for identification of adults with a clinical ASC diagnosis, using computerized measures of cognition.
Methods: Adults with (n=232) or without (n=351) a clinical ASC diagnosis completed the Autism-Spectrum Quotient (AQ) and performance measures of empathy (Reading the Mind in the Eyes Test, RMET; Karolinska Directed Emotional Faces Test, KDEFT), visuospatial ability (Embedded Figures Test, EFT; Mental Rotations Test, MRT), and nonverbal intelligence (Raven Progressive Matrices). Composite indices, each optimized for 35% or less false prediction, were formed by systematic combination of the computerized tasks administered online. Candidate classifiers were compared in terms of performance (e.g., sensitivity, accuracy) and predictive power (i.e., incremental value).
Results: Composite indices showed improved overall accuracy (max 80%) when compared to constituent tasks (max 64%). Receiver-operating characteristics identified best performance of the [RMET+KDEFT+EFT] index for general classification of adults with or without a clinical ASC diagnosis. With respect to subclassification, the [RMET+KDEFT+EFT+MRT] index best differentiated between adults with a clinical ASC diagnosis in the presence or absence of other psychiatric conditions.
Conclusions: Composite indices of cognition designed for specific autism classification performed better than constituent tasks. This use of multidimensional behavioural data supports future efforts to identify other markers of ASC, which could be incorporated into a further-improved index.