A Big Data Approach Toward the Development of New Quantitative Autism Severity Scores from Existing Instruments

Abstract Autism spectrum disorder (ASD) symptom scores from existing measures are influenced by a range of factors including age, sex, cognitive ability, language level, co-occurring psychiatric symptoms and behavioral problems. Inadequate adjustment of assessment scores can lead to reduced specificity and mis-estimation of

autism severity. Adjustment for these factors could enhance research and clinical practice, including more accurate estimation of true relationships with other variables, including etiologic factors, and better case identification, particularly for complex cases. In addition, the majority of currently available ASD measures have

reduced ability to track change and response to interventions and are not linked to adaptive functioning, which is being increasingly recognized as one of the most relevant to long term outcomes in ASD. Considering these noted limitations, enhancing the performance of current measures could advance clinical practice and future

research. The proposed project will focus on the Social Responsiveness Scale (SRS-2), the most widely used quantitative symptom questionnaire, and the Autism Diagnostic Observation Schedule–Second Edition (ADOS- 2), the gold-standard clinician observation measure. The investigation will utilize existing multiple high-quality

large data sets and apply modern psychometric approaches in order to create new continuous-range, regression-adjusted, normative SRS-2 and ADOS-2 scores that can supplement existing scores. These scores will have better specificity for autism symptom domains, more direct links to adaptive function and, when used

alone or in conjunction with existing scores, may yield greater validity for ASD diagnosis in complex, highly comorbid cases (Specific Aims 1 and 2). By virtue of having dynamic range and being more specifically related to core ASD symptom domains, newly created SRS-2 and ADOS-2 scores could have greater utility in

longitudinal applications, including treatment studies and clinical monitoring of intervention response (Specific Aim 3). Knowledge gained from this innovative secondary data analysis will then be used to develop and pilot software-implemented scoring algorithms to account for relevant demographic, cognitive and clinical factors

(Specific Aim 4). If successful, this approach will provide a low-cost enhancement to existing, widely-used measures that can be rapidly disseminated to clinicians and scientists for improving practice and research. The proposed AREA project will also include three cohorts of undergraduate students to build a cross-departmental

and cross-institutional mental health data science training experience that can be sustained as a future undergraduate research track and as an international research collaboration focused on autism spectrum disorder. Students will be involved in all aspects of the project from database building to dissemination and the

project will build key undergraduate research infrastructure as well as augment training for the next generation of mental health data scientists.