Having seen the diagnosis of autism by developmental peds experts, I don't think that any algorithm can replace that final judgment of what neurodevelopmental diagnosis is most fitting.
Primary care physicians don't struggle in applying an algorithm for probable autism diagnosis to come to that conclusion. Working diagnoses of autism are used all the time in the meantime.
Autism and related conditions affect social functioning by definition. As such, there is a REASON that it takes months to get into a specialist - because an all day multidisciplinary in depth assessment is needed by folks familiar with all manner of developmental conditions, to tease this apart. It's a social setting. It's not a backlog because people are confused and need an AI algorithm applied for them. There are also other algorithmic and validated tools, such as Autism Detect.
Suspecting autism and related conditions is not difficult. Pinning it down, and not only that, developing an effective and comprehensive treatment plan, moreso. This relies not only on physician experience, but factoring in complex social determinants and relationships that AI cannot. Can a single mom with a very large and violent teen with autism handle outbursts? Is a higher dose of sedating medication appropriate, particularly if is causes weight gain in an already overweight patient? She's the guardian and primary caregiver. Is AI going to take the place of her judgment about what she is willing to do and thinks best?
There are borderline cases and cases with more complexity.
The issues around waiting for a more formal diagnosis is multifold, and part of it is not, primary care physicians are incapable and need AI algorithm help. Some of it has to do with government and health insurance red tape around payment or qualifying for specific treatments. Resources are limited. The idea is to be certain of a certain diagnosis before committing certain resources.
ADHD vs autism vs autism with ADHD are different beasts. Right or wrong, or if it would even help ADHD to apply some treatments for autism such as in classroom support, is not the point. The PCP and AI might make one determination, but in some sense it can be a checkbox to have an expert weigh in. Then again, that's how complex cases with misdiagnosis can creep in, if we treat evals as a checkbox where AI assisting PCPs can replace the developmental peds.
It's actually kind of infuriating to me the suggestion that these issues are for the reason given and that this is where AI would be particularly helpful. I guess I would need to see data that the AI was validated and did better than 1) primary care physicians and 2) developmental peds. I suspect not, and even if so, patients will be best served by developmental peds creating a treatment plan in conjunction with a multidisciplinary team and family.