Diagnostic Agreement of a Structured Telehealth Adult Autism Assessment

This preregistered retrospective study evaluated diagnostic agreement under shared-information conditions for a structured, neurodiversity-affirming telehealth adult autism assessment protocol mapped to DSM-5-TR criteria.

The sample included 100 adults who primarily communicated through spoken language and was intentionally stratified to include at least 50% Black, Indigenous, and people of color (BIPOC; 51% achieved). Evaluations generally included two 90-min structured interviews across ten DSM-5-TR-mapped domains, criterion-level decision rules, behavioral observation, questionnaires, and collateral information when available. A blinded expert reviewed de-identified assessment materials and independently classified cases as ASD or non-ASD.

Overall agreement was 86.0% (95% CI [77.9%, 91.5%]), with Cohen's κ = 0.684 (95% CI [0.531, 0.836]). The point estimate exceeded the preregistered 0.66 benchmark, although the confidence interval included lower values. ASD classification rates were 68% for original clinicians and 66% for expert review. Positive and negative percent agreement were 90.9% and 76.5%, respectively. Exploratory subgroup analyses suggested lower agreement among BIPOC participants, although estimates were imprecise.

Findings provide preliminary evidence that a structured, neurodiversity-affirming telehealth framework can yield concordant ASD/non-ASD classifications under shared-information conditions. Neurodiversity-affirming practice can coexist with structured, criterion-based diagnostic decision-making: affirming clients' experiences and reducing deficit-based assumptions does not require presuming an autism diagnosis. Because reviewers relied on substantially overlapping information, findings do not establish diagnostic accuracy or criterion validity. Independent prospective validation using separate evaluations and multiple blinded reviewers is needed.