Sensitivity and specificity of proposed DSM-5 criteria for autism spectrum disorder in toddlers.

Matson et al. (2013) tested the brand-new DSM-5 rules for autism in toddlers. They looked at how many social-communication and repetitive-behavior boxes a child had to tick to get the label.

They ran ROC curves on a small US sample to find the sweet spot that catches real cases without tagging too many typical kids.

The best toddler cut-off is two-out-of-three social-communication items plus one-out-of-four repetitive items. That mix gave 93 % sensitivity and 74 % specificity.

In plain words: the rule grabs almost every toddler who truly has autism and only flags about one in four typical peers by mistake.

van Timmeren et al. (2016) tried the same DSM-5 checklist and saw lower numbers: 84 % sensitivity and only 54 % specificity. The difference comes from sample mix and how strict clinicians were; both studies still say the rule is useful.

Wong et al. (2016) in Singapore saw 91 % of preschoolers keep their autism label when switching from DSM-IV to DSM-5. Kids who lost the label usually had social issues but almost no repetitive behaviors—exactly the pattern L’s cut-off is meant to fix.

Hong et al. (2021) and Liu et al. (2026) show ADOS-2 Toddler Module keeps its accuracy when you pick the right cut-off. L’s work gives you the DSM-5 side of the same coin: use both tools together for a tighter toddler diagnosis.

Next time you evaluate a two-year-old, count social-communication and repetitive items separately. If the child hits two social and one repetitive, the DSM-5 label fits. This quick math cuts both missed cases and false alarms, saving families months of uncertainty.