Defining in Detail and Establishing Consensus on DSM-5 Autism Spectrum Disorder (ASD) Criteria for Case Review

Background:   The diagnostic criteria for autism and related conditions has changed based on clinical and empirical findings; however, the current DSM-5 diagnosis of ASD remains complex and behaviorally-defined. Diagnosis relies on professional evaluation of a person`s developmental history in light of the specified criteria. Most research into ASD and Pervasive Developmental Disorder (PDD) diagnostic criteria has been on the agreement of overall diagnosis, the validity of subtypes, or on factor analysis of the primary domains and specific criteria representing those domains. Within the given diagnostic taxonomy, little research has been done to specify the concepts and exemplars that serve as evidence for the criteria.  

Objectives:  To describe a process to define a comprehensive list of exemplars for each of the 7 core DSM-5 ASD criteria, and report on interrater reliability in applying these exemplars to determine ASD case status.

Methods:   The methods followed a format used for population-based prevalence record review. A team of clinicians completed an iterative process to identify specific exemplars for each of the 7 diagnostic criteria and associated features specified in DSM-5. This included mapping exemplars to the DSM-5 ASD criteria from the following sources: the Autism and Developmental Disabilities Monitoring (ADDM) Network DSM-IV-TR PDD criteria mapping; the DSM-5 text; Autism Diagnostic Observation Scale, Second Edition (ADOS-2); and Autism Diagnostic Interview-Revised (ADI-R). After each phase of mapping, expert input was sought and differences resolved. Use of the detailed criteria was piloted in South Carolina as part of a supplemental study to compare DSM-IV-TR and DSM-5 record review and direct screening with clinical evaluation on a population-based cohort of children. For the initial pilot, composite records detailing developmental evaluations (n=79) of 10 children were independently reviewed by 2 clinician reviewers to determine whether the individual behavioral exemplars could be reliably coded according to DSM-5 criteria. Primary and blinded reliability review of evaluations for approximately 210 children are underway as is piloting for use in the larger multi-site ADDM Network.

Results:   The process resulted in a case review manual detailing concepts and specific behavioral examples (over 300 exemplars) within each criteria (e.g., the example of ‘touching or acting on others without regard to other’s involvement or reaction” would be an example of “abnormal social approach” under “A1. Deficits in social emotional reciprocity”). For the first 10 cases, interrater reliability averaged 93% agreement on the DSM-5 criteria (range 80-100%), 92% on early developmental concerns, 87% on associated features, 90% on earlier PDD diagnosis, and 90% on final ASD case status. Ratings of severity levels were less consistent. Data (% agreement and Kappas) on the full reliability sample will be presented.

Conclusions: Application of the DSM-5 ASD criteria is not well-specified in clinical practice, and both researchers and clinicians may be inconsistent in terms of mapping individual behavioral exemplars to diagnostic criteria. More detailed application of exemplars within criteria and domains can be reliably defined and applied. This framework can be further evaluated for improving consistency of ASD diagnoses.