Early identification of children on the autism spectrum facilitates access to support and services for children and families. Currently available autism screening tools, however, have limited accuracy and sensitivity, particularly in community-based samples. Researchers undertook a study to evaluate the diagnostic accuracy of the Social Attention and Communication Surveillance-Revised (SACS-R) and Social Attention and Communication Surveillance-Preschool (SACS-PR) tools when used in a large, community-based sample and to report estimated autism prevalence.

In phase 1 of the study, 13,511 children were monitored by maternal and child health nurses using the SACS-R at the routine 12-, 18-, and 24-month consultations. Of the 13,511 children monitored using SACS-R, 327 (2.4%) were identified as having a high likelihood of autism. Of the 240 children who completed a university assessment, 199 met autism spectrum diagnostic criteria; the other 41 had developmental or language delays.

In phase 2 of the study, 168 children were identified as at high likelihood of autism when the SACS-PR was used during the 42-month consultation. Of the 94 children assessed by the university team, 61 met autism spectrum disorder diagnostic criteria; the other 33 had developmental or language delays. No typically developing children were identified as at high likelihood of autism using either the SACS-R or the SACS-PR.

At 12 to 24 months, the SACS-R had a positive predictive value (PPV) of 83%, a negative predictive value (NPV) of 99%, sensitivity of 62%, and specificity of 99.6%. When the SACS-PR was added at 42 months, the PPV was 78%, the NPV was 99.9%, sensitivity was 96%, and specificity was 98%.