Detecting subgroups in children diagnosed with pervasive developmental disorder - not otherwise specified.

Sparaci et al. (2015) looked at toddlers who got the catch-all label PDD-NOS at age two.

They fed many test scores into a cluster program to see if hidden groups popped out.

The team checked each child again around age four to learn who kept an ASD label and how they scored on IQ tests.

The numbers sorted the toddlers into three clear subgroups, not a messy blur.

One group shifted to classic autism, one to milder ASD, and one lost the label while their IQ scores split high, mid, and low.

The clusters at age two predicted these later paths better than the original PDD-NOS tag.

Rivard et al. (2026) ran the same math on preschoolers with ASD and also found three clusters, showing the pattern holds past toddlerhood.

Journal et al. (2024) zoomed in on early social-communication skills and again landed on three toddler clusters that forecast later development, proving the idea works when you cut the data finer.

Bitsika et al. (2018) seems to disagree: they saw only two clusters that differed mainly in severity, not kind. The gap is age. Vicki studied older kids where symptoms may smooth into a line, while Laura worked with two-year-olds where distinct types still stand apart.

If you assess a very young child who does not fit classic autism, do not stop at PDD-NOS. Pull scores from language, play, social, and IQ tasks and let a quick cluster rule place the child into one of three lanes. This gives families a clearer picture of likely kindergarten needs and helps you pick targets that match that lane instead of using one-size-fits-all goals.