Observation-centered approach to ASD assessment in Tanzania.

Doctors in Tanzania needed a way to spot autism when no local tests existed. They gave the Childhood Autism Rating Scale-2 (CARS2) to 41 kids . Half had known autism, half had general delays.

They watched each child play, talk, and move. Then they scored 15 items like relating to people and body use. No fancy lab gear—just clipboards and trained eyes.

CARS2 scores cleanly split the two groups. The autism group averaged 34 points, the delay group 21. A cut-off of 30 caught 90 % of autism cases and ruled out 86 % of delays.

Inter-rater agreement was 92 %. The tool worked even when kids spoke Swahili, not English.

Nah et al. (2014) found CARS scores did NOT predict how autistic kids looked six years later. That sounds like a clash, but the two studies asked different questions. The Tanzania paper shows CARS2 can separate autism from delay right now. Yong-Hwee asked if early CARS scores forecast later skills—they don’t.

Meimei et al. (2022) reviewed telehealth screens that reach kids through a phone. Their paper and this one both push low-cost answers where specialists are scarce. Telehealth gives you remote help; CARS2 gives you in-person clarity when the internet fails.

Nickerson et al. (2015) call for open-source tools. CARS2 is commercial, but the team shared their Swahili prompt sheet and cut-off free of charge—answering that call on the ground.

If you work in rural clinics, refugee camps, or low-income schools, you can borrow this game plan. Bring the CARS2 manual, translate the item descriptions into the local language, and score kids during natural play. A 30-point cut-off gives a quick red-flag for referral until full evaluation arrives. You just turned a 45-minute observation into a culturally fair screen.