Stability of early risk assessment for autism spectrum disorder in preterm infants.

The team tracked preterm babies who were born early. They gave two autism screens months apart. First they used the AOSI. Later they used the ADOS-T at 18 months. They wanted to see if the first score could predict the second.

Only half of the babies who failed the later ADOS-T had been flagged by the early AOSI. Risk labels moved around. A baby could score low at first and still show signs later.

Giserman-Kiss et al. (2020) looked at stable autism diagnosis in a mixed preschool group. They found 88% kept the label. The numbers seem opposite, but the kids were older and born on time. The two studies together say: early signs are less steady in preterm babies.

Kuang et al. (2025) took the next step. They built a computer model that uses BSID-III motor and language scores to sort very preterm infants. Their tool can say "low risk" with 93% certainty. It does not replace the ADOS-T, but it adds another check.

Capio et al. (2013) saw the same drift with Bayley-III delay labels in preterm toddlers. About 15% changed status between tests. The pattern repeats: one score is not enough for this group.

If you screen a preterm infant, mark the calendar for an 18-month rescreen even if the first score looked fine. Share this plan with parents so they expect the follow-up. Use later tools like the ADOS-T plus developmental trajectories before making any firm risk call.