Essential behaviours for the diagnosis of DSM-5 Autism Spectrum Disorder
Carrington, Sarah × Kent, Rachel Maljaars, Jarymke Le Couteur, Ann Gould, Judith Wing, Lorna Noens, Ilse Van Berckelaer-Onnes, Ina Leekam, Susan #
Australasian Society for Autism Research (ASfAR) Conference edition:2 location:Melbourne, Australia date:4-5 December 2014
Background: Despite initial concerns about the sensitivity of the diagnostic criteria for DSM-5 Autism Spectrum Disorder (ASD), evidence is growing that the criteria provide both good sensitivity and specificity (e.g. Frazier et al., 2012; Kent, Carrington et al., 2013). The capacity of the criteria to provide good sensitivity and specificity, however, relies on careful measurement to ensure that appropriate items from diagnostic instruments map onto the DSM-5 descriptions.
Objective: To use an existing diagnostic algorithm to identify a set of ‘essential’ behaviors sufficient for reliable, accurate diagnosis of DSM-5 ASD.
Methods: Specific behaviors were identified and tested from the recently published DSM-5 algorithm (Kent, Carrington et al., 2013) for the Diagnostic Interview for Social and Communication Disorders (DISCO). Analyses were run on existing datasets (n = 335). Predictive validity was tested using ROC curves and sensitivity across age and ability was investigated in a subset of individuals (n = 190).
Results: Study 1 identified a set of 48 items with good predictive validity, but which lacked sensitivity for higher ability individuals. An adjusted item set (54 items) improved sensitivity for higher ability individuals and performance and was comparable to the published DISCO DSM-5 algorithm (Study 2).
Conclusions: This work represents a first attempt to derive a reduced set of behaviors for DSM-5 directly from an existing standardized ASD developmental history interview. Further work is required to replicate these findings and exploit their potential to contribute to more efficient and focused diagnosis.