The modified checklist for autism in toddlers: reliability in a diverse rural American sample.

Scarpa et al. (2013) checked if the M-CHAT still works well in rural, low-income families. They gave the checklist to caregivers who were mostly minorities and had little schooling.

The team looked at how often the screen called a toddler “at risk” and whether those flags matched later diagnoses.

The M-CHAT gave too many false alarms in this group. Internal consistency was poor, meaning items did not hang together well.

Low parent education and minority status drove the high false-positive rate.

Leung et al. (2014) saw the same problem in urban Hispanic clinics: about 30% of toddlers screened positive, far above typical rates. Lung et al. (2017) added that less-educated families in Taiwan were more likely to score high yet never receive a final diagnosis.

At first glance this seems to clash with Kamio et al. (2014), who reported only 1.4% false positives in Japan. The difference is setting: Japan used the screen at a routine 18-month check-up with extra nurse coaching, while the U.S. rural and Hispanic samples relied on parents filling out the form alone.

Christopher et al. (2021) extended the worry, showing that high externalizing behaviors and younger age also drop the M-CHAT-R/F accuracy. Taken together, the tool over-identifies when caregivers have low education or speak another language, and under-identifies when kids are very young or highly active.

If you screen in rural, low-SES, or minority communities, expect inflated positives. Do not jump to referral; re-explain items, use an interpreter, or add a second brief tool. This cuts unnecessary stress and travel for families while keeping true cases in view.