70/100

Moderate evidence

Cite
More
EmailLinkedInX

New parent questionnaire found about twice as many autism cases as M-CHAT-R/F, but tested on only 202 children

Study title: Exploring the Potential of Parent Report for Autism/Developmental Screening at the 18-Month Visit

Evidence type
Controlled or validated comparison
Sample
n = 11,878
Sources
1 sourcePeer-reviewed
Published
1 min read

Study title: Exploring the Potential of Parent Report for Autism/Developmental Screening at the 18-Month Visit

Researchers developed and tested a parent-report screening algorithm called TADAS for the 18-month pediatric visit. Using data from 11,878 toddlers who completed M-CHAT-R/F, Q-CHAT-10-O and ASQ-3, plus extended assessments in a subsample, they trained gradient-boosting models on synthetic data and evaluated them against ADOS-2 and Mullen diagnoses in an authentic holdback sample.

Read the study abstract

Purpose To develop optimal algorithms of parent-administered items to improve the detection of autism and other developmental disorders at the 18-month visit. Methods Parents of 11,878 children aged 16-20 months completed the M-CHAT-R/F ™ , Q-CHAT-10-O, and ASQ-3 R at scheduled 18-month pediatric visits via an online system. Ninety-six children with positive screens and 314 matched controls completed additional items, including the POSI; items from the FYI and POEM data banks, and the MacArthur-Bates Communicati…

Exploring the Potential of Parent Report for Autism/Developmental Screening at the 18-Month Visit
"Mental Patients Against Stigma" by Mindy Fisher is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

The gain

In a test on 202 children, a parent-report model using vocabulary and joint attention items reached over 0.7 sensitivity and specificity for autism and developmental delay, about twice the sensitivity of M-CHAT-R/F for autism.

In brief

The resulting model, based on expressive vocabulary and joint attention items, was reported as approximately twice as sensitive as M-CHAT-R/F for autism and twice as sensitive as ASQ-3 for developmental delay, reaching over 0.7 sensitivity and specificity for both conditions. This suggests a single parent-report tool could improve early detection, but performance rests on synthetic-data training and a small validation sample of 202 children, leaving real-world generalizability uncertain.

Main points

  1. Parents of 11,878 children aged 16-20 months completed M-CHAT-R/F, Q-CHAT-10-O, and ASQ-3 at scheduled 18-month pediatric visits via an online system.
  2. 96 children with positive screens and 314 matched controls completed additional items including POSI, FYI and POEM banks, and MacArthur-Bates MCDI short-form vocabulary.
  3. Diagnostic testing used ADOS-2 Toddler Module and Mullen; ML models used tree-based gradient boosting with Boruta feature selection via Shapley values and Bayesian hyperparameter optimization.

The full story

The study collected parent-administered items through an online system at routine 18-month visits, then enriched data with 96 screen-positive children and 314 matched controls who completed POSI and vocabulary measures. Reference standard was ADOS-2 Toddler Module and Mullen.

Training used gradient boosting trees with Boruta-Shapley feature selection and Bayesian optimization on synthetic data derived from 201 authentic cases, expanded to 25,000 training and 12,500 validation cases, with final testing on 202 authentic holdback cases.

Sources

1 on file
  1. Peer-reviewed2026-09-28

Talk it over

Three questions to start on, two more if you have time
The TADAS model reached over 0.7 sensitivity and specificity, but was tested on 202 authentic holdback children after training on synthetic data. How much weight would you give that result, and why?

Start with a question

New parent questionnaire found about twice as many autism cases as M-CHAT-R/F, but tested on only 202 childrenhttps://truvace.com/item/exploring-the-potential-of-parent-report-for-autism-developmental-screening-at-t

  1. The TADAS model reached over 0.7 sensitivity and specificity, but was tested on 202 authentic holdback children after training on synthetic data. How much weight would you give that result, and why?

    Answer in the thread
  2. The models were trained on synthetic data built from 201 authentic cases, then tested on 202 authentic children. How would you judge whether a model trained mostly on made-up cases will hold up with toddlers in a routine 18-month visit?

    Answer in the thread
  3. If you were a parent at an 18-month visit, which would you want your child screened with: M-CHAT-R/F, or a parent-report tool like TADAS that has only been tested on 202 children? Why?

    Answer in the thread
If you have more time
  • The 96 screen-positive children and 314 matched controls did extra items. How might that selection of children change how well the tool works for toddlers in a routine pediatric practice?

    Answer in the thread
  • What size and kind of new sample, tested against ADOS-2 and Mullen, would settle whether TADAS really finds about twice as many autism cases as M-CHAT-R/F?

    Answer in the thread

What readers say