Validating two survey methods for identifying cases of autism spectrum disorder among adults in the community

There are no tested methods for conducting epidemiological studies of autism spectrum disorders (ASDs) in adult general population samples. We tested the validity of the Autism Diagnostic Observation Schedule module-4 (ADOS-4) and the 20-item Autism-Spectrum Quotient (AQ-20).

Randomly sampled adults aged ≥16 years were interviewed throughout England in a general population multi-phase survey. The AQ-20 was self-completed by 7353 adults in phase 1. A random subset completed phase 2, ADOS-4 assessments (n=618); the probability of selection increased with AQ-20 score. In phase 3, informant-based Diagnostic Interview Schedule for Social and Communication Disorders (DISCO) and Autism Diagnostic Interview-Revised (ADI-R) developmental assessments were completed (n=56). Phase 1 and 2 data were presented as vignettes to six experienced clinicians (working in pairs). The probability of respondents having an ASD was compared across the three survey phases.

There was moderate agreement between clinical consensus diagnoses and ADOS-4. A range of ADOS-4 caseness thresholds was identified by clinicians: 5+ to 13+ with greatest area under the curve (AUC) at 5+ (0.88). Modelling of the presence of ASD using 56 DISCO assessments suggested an ADOS-4 threshold in the range of 10+ to 13+ with the highest AUC at ADOS 10+ to 11+ (0.93-0.94). At ADOS 10+, the sensitivity was 1 [95% confidence interval (CI) 0.59-1.0] and the specificity 0.86 (95% CI 0.72-0.94). The AQ-20 was only a weak predictor of ADOS-4 cases.

Clinically recommended ADOS-4 thresholds are also recommended for community cases: 7+ for subthreshold and 10+ for definite cases. Further work on adult population screening methods is needed.