Autism diagnostic interview: a standardized investigator-based instrument.

The team built the first Autism Diagnostic Interview. It is a scripted set of questions for parents.

They tried it on 16 autistic kids and 16 kids with intellectual disability only. All kids had ID.

The new interview got every diagnosis right. It flagged all autistic kids and no non-autistic kids.

This perfect score showed the tool could tell the two groups apart.

Moss et al. (2008) followed kids for seven years. Most kept their ADI-R label, but symptom scores dropped. The tool stayed stable even as kids improved.

McLennan et al. (2008) trimmed the toddler version. They dropped the behavior domain and agreement went up. The full 1989 form works best for school-age kids, not toddlers.

Lotfizadeh et al. (2020) tried billing codes alone. Codes missed many cases and called some wrong kids autistic. Their poor results remind us why we still need the gold-standard interview.

You now know the ADI started with perfect accuracy in a small, clear sample. Later work shows scores hold steady for years and can be shortened for toddlers. When you assess, use the full algorithm for ages four and up. For under-threes, skip the behavior domain. Always pair the interview with observation; billing codes are not enough.