Level 1 screening for autism spectrum disorders (ASD) requires balance between minimizing false negatives, or missed cases, who miss out on needed early diagnostic and intervention services and false positives, who undergo unnecessary stress and increase burden by lengthening waiting lists and/or increasing demands for services from providers. The Modified Checklist for Autism in Toddlers (M-CHAT; Robins, Fein, & Barton, 1999) has been validated in Level 1 samples. Positive predictive value (PPV) is lower than would be optimal at .05-.11 for M-CHAT alone and .28-.65 for M-CHAT plus Follow-up Interview (FUI), although when all developmental disorders are considered, PPV is much higher.
Objectives:
A revision of the M-CHAT (Robins, Fein, & Barton, 2009) was developed to reduce the false positive rate while maintaining maximal sensitivity. The current study presents the initial validation of the M-CHAT-R.
Methods:
The M-CHAT-R is composed of 20 yes/no parent-report items. M-CHAT items were reworded to reduce misinterpretation, item order was revised to prevent parents from falling into a response set of answering “yes” to all items, examples were provided for many items, and the three worst-performing M-CHAT items were dropped. The M-CHAT-R was piloted in 3238 toddlers (50.96% male, mean age=21.69 months, SD=3.45 months, range: 14-40 months) screened during 18- or 24-month well-child visits to participating pediatric offices in metropolitan Atlanta. Parents of children who screened positive on the M-CHAT-R were offered the M-CHAT-R Follow-up Interview (FUI), and those who continued to screen positive were offered a free diagnostic evaluation. Screen positive was defined as failing two of the best7 items or any three items.
Results:
7.13% of toddler (n=231) screened positive on M-CHAT-R, and 22.46% of those who completed the FUI continued to screen positive (n=42 of 187); 23 of these 42 completed the evaluation, and 16 were diagnosed with ASD. Five additional toddlers with ASD were detected based on physician or parent concern, although four of these had screened positive on M-CHAT-R (but not FUI). It is too early to calculate sensitivity, but considering that the original M-CHAT detected 45 cases per 10,000 screened and the M-CHAT-R is detecting 49 cases per 10,000 screened, sensitivity does not appear to be reduced compared to the original M-CHAT. PPV was calculated excluding cases who declined to complete the study, and was .11 (20/178) for M-CHAT-R and .70 (16/23) for FUI. Only one child evaluated based on M-CHAT-R + FUI score had no diagnosis, bringing the PPV for all developmental delays to .96.
Conclusions:
The M-CHAT-R demonstrates promising psychometric properties, and appears to improve the rate of detection of ASD. Additional improvements may be made by developing new scoring algorithms, which should be undertaken once the sample is larger.