Experts caution against replacing structured tools with clinical judgement in autism diagnosis

A recent study suggesting clinicians can accurately diagnose autism without structured tools has sparked debate, with experts emphasising the importance of methods like ADOS for consistency and accuracy outside specialised settings.

A recent JAMA Pediatrics paper has reignited a familiar debate in autism care: whether experienced clinicians can diagnose young children without a structured observation such as the Autism Diagnostic Observation Schedule, or ADOS. The study found that clinicians’ first impression matched the diagnosis reached after ADOS review in about 90% of cases and argued that seasoned developmental-behavioural paediatricians can often make an accurate diagnosis without the formal tool. But that conclusion, as Ronald L. Lindsay argues, is far harder to apply beyond the highly specialised academic settings in which the research was done.

Lindsay, a retired developmental-behavioural paediatrician, says the real world looks very different from the carefully controlled environments of specialist networks. In community practice, he writes, children are far more likely to be assessed by general paediatricians, family doctors, advanced practice nurses and paediatric nurse practitioners than by the small number of board-certified developmental-behavioural paediatricians still working in the United States. He contends that the study assumes a level of expertise, time and institutional support that many clinics simply do not have, and that “clinical judgement” is too variable to be treated as a uniform diagnostic method.

The strength of the ADOS, according to earlier research, is not that it replaces a clinician but that it adds structure. Studies in PubMed and elsewhere have long found strong interrater reliability, good internal consistency and useful diagnostic performance for the instrument across different modules and age groups. Research on the tool’s objectivity has also shown that it performs best under standardised conditions, while routine clinical settings can introduce variability that makes training and protocol especially important. In other words, the test is not perfect, but it helps make autism assessment more consistent and more defensible when observations are brief or the presentation is unclear.

Lindsay points to his own experience in military and community settings as evidence that structured workflows matter. He says a combination of detailed history-taking, targeted observation and a formal DSM-5 checklist can produce faster and more reliable results than impression alone, particularly when the clinician is working under time pressure. He also argues that the study’s youngest children might have been better served by toddler-focused structured tools such as the Screening Tool for Autism in Toddlers and Young Children, which are designed for short evaluations and can complement broader diagnostic work. The larger point, he says, is that accuracy comes from method, not confidence.

That is why Lindsay sees the paper’s main weakness as its broader implications rather than its data. He says the findings may be valid for a tightly knit academic network of highly experienced specialists, but they should not be read as a licence to strip structure from autism diagnosis elsewhere. The ADOS, in his view, remains important because it creates a common framework, checks bias and helps protect children from both overdiagnosis and missed diagnosis. His warning is straightforward: a system built on variable experience and limited time cannot assume that judgement alone will be enough.