21297
Feasibility of Developing an Algorithm to Derive Ratings of Social Communication Functioning (ACSF:SC) from ADOS Data

Thursday, May 12, 2016: 5:30 PM-7:00 PM
Hall A (Baltimore Convention Center)
S. J. Gentles1, B. M. Di Rezze1,2, P. Rosenbaum1, L. Zwaigenbaum3, M. J. C. Hidecker4, S. Georgiades5 and E. Duku5, (1)CanChild Centre for Childhood Disability Research, McMaster University, Hamilton, ON, Canada, (2)School of Rehabilitation Science, McMaster University, Hamilton, ON, Canada, (3)University of Alberta, Edmonton, AB, Canada, (4)Communication Disorders, University of Wyoming, Laramie, WY, (5)McMaster University, Hamilton, ON, Canada
Background:   Modeled on the internationally recognized Gross Motor Function Classification System for cerebral palsy, our group developed and validated the Autism Classification System of Functioning: Social Communication (ACSF:SC). This descriptive classification system—based on WHO’s International Classification of Functioning, Disability, and Health (ICF) framework—contrasts with traditional severity metrics by assessing abilities rather than deficits. The ACSF:SC allows parents or professionals to categorize children aged 3 to 5 into one of five meaningfully distinct levels of social communication functioning ability. It may be possible to extract relevant descriptive levels of social communication abilities from standardized tools routinely used in autism assessment and research (e.g., the Autism Diagnostic Observation Schedule [ADOS]). Being able to derive ACSF:SC ratings from such instruments would allow abundant earlier data to be translated into our unique validated indicator of children’s level of social communication functioning, immeasurably increasing the value of such data for research purposes. Importantly, such ratings would enable the examination of longitudinal trajectories of social communication abilities among children with autism.

Objectives:   We aim to develop an algorithm to derive ACSF:SC ratings from ADOS data. As part of this work, we will evaluate the reliability of ADOS-derived ACSF:SC ratings at progressive stages of algorithm development.

Methods:   Preliminary results were obtained using a mini-Delphi method (i.e., a small group face-to-face multiple-round consensus method): first, to identify items from the ADOS that correspond to the ACSF:SC construct of social communication, and second, to map the scores available within each relevant ADOS item to ACSF:SC ratings to determine the range of ACSF:SC levels covered. We plan an expanded Delphi method exercise (70% agreement criterion) involving our interdisciplinary team (n=6). 

Results:   In preliminary work, three team members identified 15 relevant items from ADOS Module 2 corresponding to the ACSF:SC construct of ‘social communication’; we also mapped the scores available within each relevant ADOS item to corresponding ACSF:SC ratings, determining the spread of ACSF:SC levels covered by that item. 100% consensus was reached after 2 rounds. Together all relevant ADOS items provided full coverage of the 5-level range of the ACSF:SC. Based on preliminary examination, some relevant ADOS items appear more information-rich, suggesting that weighting and possibly other ‘if-then’ conditions may be required in a final algorithm to convert relevant ADOS item scores to an overall ACSF:SC rating.

Conclusions:   We describe methods for deriving a classification of social communication functioning from ADOS data. Preliminary results suggest this is definitely feasible. We will also present findings of the expanded Delphi method, and initial version of the algorithm to be developed using anonymized ADOS assessment data. This algorithm will be validated by assessing the correlation of the ADOS algorithm-derived ratings with paired ACSF:SC ratings from an existing dataset. The methodology outlined may provide a template for developing additional algorithmic means of deriving ICF-based functioning ratings in autism (beyond social communication) from data routinely collected by commonly used instruments like the ADOS—extending opportunities for secondary data analysis. A valid algorithm will enable longitudinal examination of social communication abilities.