Autism screening and diagnostic outcomes among toddlers born preterm.

Gestational age was available for 9,725 of 20,286 screened toddlers, and most participants before 2011 were excluded because gestational age was missing. Children who screened positive but did not attend evaluation were excluded from further analyses, so the screener accuracy results may not represent all referred children. Medical factors such as birth weight, intracranial hemorrhage, genetic conditions, and inflammatory conditions were not accounted for, and parent report did not include an exhaustive medical history. Screen-positive rates may be influenced by co-occurring conditions such as cerebral palsy, hemiparesis, or retinopathy of prematurity, which were not directly examined. Attendance differed by demographics, so the evaluated sample may be selective compared with the full referred population. Item-level screening data were not available, so blank items and item reliability could not be examined. Some children who screened negative may still have had autism, because additional missed cases are likely to be identified as children age.

Two authors, Diana L. Robins and Deborah Fein, are co-owners of M-CHAT, LLC, which licenses the M-CHAT and M-CHAT-R/F and collects royalties. Dr. Robins has a 50% share, and Dr. Fein has a 25% share. Dr. Robins also serves on the advisory board for Quadrant Biosciences, Inc. and the Program Quality Committee of the Bancroft Board. The other authors reported no potential conflicts of interest.

Do not treat the screener as equally accurate in every group. It identified all diagnosed autism cases in the extremely and very preterm groups (100% sensitivity), but only 72.65% in full-term children. It also correctly identified children without autism less often in the extremely preterm group (79.70% specificity). A negative screen does not rule out autism, and a positive screen does not confirm it.