When every child is screened, autism occurrence has not risen

Design

prospective population-based screening with diagnostic confirmation, 12 consecutive birth cohorts

Population

more than 62,000 South Korean children born 1998-2010, assessed at age 7; 881 fully assessed

Primary outcome

7-year cumulative incidence and prevalence of autism spectrum disorder at age 7

Effect

incidence 1.9% to 3.2%, prevalence 2.0% to 3.4%, with no significant trend across cohorts

Most autism prevalence figures come from administrative or service-based data, which count children who reached a service. That makes them sensitive to diagnostic criteria, referral patterns and funding incentives - all of which have moved substantially since the 1990s. This study instead screened the entire population of children entering elementary school across 12 consecutive birth cohorts in South Korea, more than 62,000 children born between 1998 and 2010 and assessed at age seven. Of 2077 screening positives, 881 completed full diagnostic assessment with structured observation, parent interview, cognitive testing and record review against DSM-5 criteria.

Seven-year cumulative incidence ranged from 1.9% (95% CI 1.5-2.4) to 3.2% (95% CI 2.7-3.8), and prevalence at seven from 2.0% (95% CI 1.5-2.6) to 3.4% (95% CI 2.7-4.0), with no statistically significant increase or decrease across the twelve cohorts. Set against a period in which service-based estimates rose steeply, the flat line is the result.

The second finding is the clinically useful one. Girls with autism were disproportionately represented among children identified in mainstream education who had never been diagnosed or received any intervention, and they had distinct presentations and substantial unmet need. That is the group your clinic is not seeing, and it is the group most likely to be described by a school as quiet and coping.

• When a parent asks whether autism is becoming more common, this is the evidence that detection explains much of the apparent rise.
• Lower your threshold for assessing girls, particularly those doing adequately at school.
• Do not use service-based prevalence trends as evidence about causes - they measure ascertainment as much as occurrence.
• Ask about social exhaustion and masking rather than relying on observed classroom behaviour.
• Estimates come from South Korean cohorts; ascertainment in Indian settings is lower still, so undiagnosed girls are likely a larger group here.

Why it matters

It removes the central factual premise from the claim that something in the environment is causing an autism epidemic.

Don't overread it

This is one country's cohorts over a defined period; it does not establish that occurrence is flat everywhere, only that a population-screened series showed no rise.

The statistics, in plain English

The confidence intervals across cohorts overlap substantially - 1.9% to 3.2% across twelve years, with each estimate's own interval spanning roughly half a percentage point. That overlap is what 'no significant change' means here: year-to-year variation is within the noise you would expect from sampling, not a trend. The design is the strength: because every child was screened rather than counted on reaching a service, changing service patterns cannot drive the numbers.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.