Two-Stage Autism Screening

Background: The Modified Checklist for Autism in Toddlers (M-CHAT, Robins, Fein, Barton, & Green, 2001) is a widely used early autism screener; however, it is rarely followed up with measures for school-age children, such as the Social Communication Questionnaire (SCQ; Rutter et al., 2003) or the Social Responsiveness Scale (SRS; Constantino et al., 2003). Such two-stage screening would best assess the M-CHAT’s predictive power.

Objectives: The primary goals of the study were to (1) complete a two-stage screening of twins born from 1998-2004 who were recruited from birth records or via statewide case-finding; (2) consider use of both threshold and full range M-CHAT scores in predicting later SCQ and SRS scores; and (3) identify specific items indicative of false positive M-CHAT scores.

Methods: Recruitment occurred in a step-wise fashion. First, mothers of >3000 twins from a statewide, unselected population born from 1998-2004 completed items similar to the six critical M-CHAT items when the twins were between 2 and 3 years of age. Second, twin pairs in which at least one of the twins (1) had at least two endorsed M-CHAT items (n=14 pairs) or (2) received subsequently a community autism spectrum diagnosis, regardless of M-CHAT score (n=16 pairs) were invited to participate in another screening phase when the twins were at least 4 years of age. The SRS and SCQ composed the second phase of screening.

Results: Of 29 individual twins scoring at or above the published threshold of two endorsed M-CHAT items, 12 did not meet or exceed threshold on either the SCQ or the SRS (with the SRS cutoff of 76 T-score). Shifting to a higher threshold of at least three endorsed MCHAT items (which reduced the above-threshold N from 29 to 11), only two children did not meet threshold on either the SCQ or the SRS, and the other nine met threshold on both the SCQ and SRS. For the threshold of two M-CHAT items, we determined that M-CHAT items assessing interest in other children, sharing, and following gaze most frequently characterized children who would later be designated as “false positives,” whereas the item assessing eye contact contributed to “true positive” outcomes. Total M-CHAT scores were moderately correlated with both the SCQ (r=.64) and SRS (r=.53) for males (n=42), but not for the much smaller group of females (SCQ, r=.34; SRS, r=.15). These correlations suggest that the M-CHAT is predictive across the range of scores and may be useful as a dimensional measure. Limitations of sample size, lack of follow-up of children with sub-threshold M-CHAT scores, and possible age-dependence of findings will be discussed. A second follow-up of the sample with the Autism Diagnostic Observation Scale will also be described.

Conclusions: A slightly higher M-CHAT threshold than is usually employed may minimize false positives, and the full range of M-CHAT scores may be predictive of later SCQ and SRS scores. Endorsement of particular M-CHAT items may be associated with false positive screening outcomes.