Show the full module text
Module 2 — Who Assesses You, and Why It Matters
Which professions can diagnose adult autism and ADHD, how that changes at every border, and what actually predicts a sound assessment once you are past the job title.
The Assessment · Part One — Before you book
The big idea
Short on capacity today? The big idea: Module 1 ended on an uncomfortable finding. Two clinicians, the same manual, the same criteria, the same person — and agreement on ADHD that ran at 0.71 at one research site and 0.45 at another. The conclusion drawn there was that reliability is not a property of the diagnosis. It is a property of the diagnosis as practised, in a particular place, by particular people.
If that is true, then “who is doing this” is not administrative detail to be settled after you have chosen a service. It is the largest variable you can still control.
This module is about how to think about that variable — and it will end up somewhere most people do not expect. The profession on the letterhead matters much less than two things that are rarely advertised: how much specific experience the person has with adults like you, and how much structure they work inside.
Step 1 — The lesson
A. There is no universal answer to “who is allowed”
The first thing to say plainly: the question which professions can diagnose adult autism or ADHD has no international answer. It has a local one, and the local one changes at every border, sometimes at every state line.
Two examples, chosen because both can be read from the rule-making body’s own published text.
England, Wales and Northern Ireland. The national clinical guideline for ADHD states that “ADHD should only be diagnosed by a specialist psychiatrist, paediatrician or other appropriately qualified healthcare professional with training and expertise in diagnosing ADHD” — and then adds what that diagnosis must rest on: a full clinical and psychosocial assessment across more than one setting of the person’s everyday life, a full developmental and psychiatric history, observer reports, and assessment of mental state. The same guideline says services should form multidisciplinary specialist ADHD teams and, explicitly, separate teams or clinics for adults.
The parallel guideline for autism in adults does not name a gatekeeping profession at all. It describes a team. In each area there should be a specialist community-based multidisciplinary team for autistic adults, and its membership “should include” psychologists with training and experience in working with autistic adults, nurses, occupational therapists, psychiatrists, social workers, speech and language therapists, and support staff. Seven roles. No single one of them owns the diagnosis.
The United States. There is no national rule. Scope of practice is written state by state, in each state’s licensing statute. In California, for instance, the practice of psychology is defined in the Business and Professions Code, and the section spelling out what that practice covers says it includes, among other things, “assessment, diagnosis, prevention, treatment, and intervention.” A licensed psychologist there is doing the thing their statute names when they diagnose. In another state the wording — and therefore the answer — may differ.
The honest general statement, then, is this: in most jurisdictions more than one profession can diagnose these conditions, and in almost none is it only psychiatrists. Clinical and educational psychologists, psychiatrists, paediatricians, specialist nurses, and in some systems speech and language therapists and occupational therapists working within a team, all appear in that list somewhere.
Diagram — A · Seven roles, and no gatekeeper. The first thing to say plainly is that the question which professions can diagnose adult autism or ADHD has no international answer. It has a local one, and the local one changes at every border, sometimes at every state line. Two examples, both readable from the rule-making body's own published text. In England, Wales and Northern Ireland, the national clinical guideline for ADHD states that ADHD should only be diagnosed by a specialist psychiatrist, paediatrician or other appropriately qualified healthcare professional with training and expertise in diagnosing ADHD — and then adds what that diagnosis must rest on: a full clinical and psychosocial assessment across more than one setting of the person's everyday life, a full developmental and psychiatric history, observer reports, and assessment of mental state. The same guideline says services should form multidisciplinary specialist ADHD teams and, explicitly, separate teams or clinics for adults. The parallel guideline for autism in adults does not name a gatekeeping profession at all. It describes a team: in each area there should be a specialist community-based multidisciplinary team for autistic adults, and its membership should include psychologists with training and experience in working with autistic adults, nurses, occupational therapists, psychiatrists, social workers, speech and language therapists, and support staff. Seven roles, and no single one of them owns the diagnosis. In the United States there is no national rule at all: scope of practice is written state by state in each state's licensing statute. In California, the practice of psychology is defined in the Business and Professions Code, and the section spelling out what that practice covers says it includes, among other things, assessment, diagnosis, prevention, treatment, and intervention. A licensed psychologist there is doing the thing their statute names when they diagnose; in another state the wording, and therefore the answer, may differ. The honest general statement is that in most jurisdictions more than one profession can diagnose these conditions, and in almost none is it only psychiatrists.
But there is a second question hiding behind the first, and for practical purposes it is the more important of the two.
B. The question that decides more than the law does
Not may this person diagnose me? but whose document will the place I need it for actually accept?
Those come apart more often than people expect, and the gap costs money.
- A prescription needs a prescriber. If you are seeking an ADHD assessment partly because you want to try medication, the profession matters concretely: a diagnosis from a non-prescribing professional is a valid diagnosis and does not by itself get you a prescription. The English guideline puts the boundary in one sentence — all medication for ADHD should only be started by a healthcare professional with training and expertise in diagnosing and managing ADHD. Where a diagnosing psychologist and a prescribing physician are two people, you will need both, and it is better to know that before the first appointment than after the last.
- Employers, universities and benefits systems set their own evidence rules. Some accept a report from any suitably qualified clinician. Some name professions. Some require the report to be less than a certain number of years old, or to include a specific functional-impact section. None of them publish this where you would naturally look.
- A public health system may not recognise a private diagnosis for onward care, even where it is perfectly valid as a diagnosis. This is a funding rule rather than a clinical one, and it is the single most common expensive surprise in this whole area.
The practical instruction is short: find out what the specific door you need requires, in writing, before you book. One email. It is the highest-return thing in this module.
Diagram — B · Two questions, and the gap between them. There is a second question hiding behind the first, and for practical purposes it is the more important of the two: not may this person diagnose me, but whose document will the place I need it for actually accept? Those come apart more often than people expect, and the gap costs money. A prescription needs a prescriber: if you are seeking an ADHD assessment partly because you want to try medication, the profession matters concretely, because a diagnosis from a non-prescribing professional is a valid diagnosis and does not by itself get you a prescription. The English guideline puts the boundary in one sentence: all medication for ADHD should only be started by a healthcare professional with training and expertise in diagnosing and managing ADHD. Where a diagnosing psychologist and a prescribing physician are two people, you will need both, and it is better to know that before the first appointment than after the last. Employers, universities and benefits systems set their own evidence rules: some accept a report from any suitably qualified clinician, some name professions, some require the report to be less than a certain number of years old or to include a specific functional-impact section, and none of them publish this where you would naturally look. And a public health system may not recognise a private diagnosis for onward care, even where it is perfectly valid as a diagnosis — a funding rule rather than a clinical one, and the single most common expensive surprise in this whole area. The practical instruction is short: find out what the specific door you need requires, in writing, before you book. One email. It is the highest-return thing in this module.
C. What training actually predicts
Now to the part that matters more than the letterhead.
Start with a study designed to isolate exactly this. Researchers recorded seven administrations of the standard autism observation schedule and had a group of experts, working through a consensus process that included minute-by-minute video analysis, define a gold-standard coding for every case. They then showed those same recordings to a large group of clinicians drawn from routine care and collected 189 codings.
The result: high variance. Ordinary clinicians, using the best-known instrument in the field, coded the same recorded person differently from one another — and the accuracy of a person’s coding depended on their experience with that instrument, on features of the case, and on how well the assessment had been administered in the first place. The authors’ conclusion is blunt about what follows: specialisation in autism diagnosis has to be required, and specialised services are needed to guarantee a qualified differential assessment.
Read that carefully, because it is easy to take the wrong lesson. The finding is not that the instrument is worthless. It is that an instrument is a thing a person operates, and operating it well is a skill that is acquired by doing it a great deal, with feedback, in a setting where somebody checks.
Now the mirror image, from Module 1. When eleven clinicians across eight ordinary clinical sites each rated recorded administrations of a structured diagnostic interview — five raters per recording, no research-reliability training — agreement on the diagnostic classification came out at kappa 0.83, with agreement on the individual criterion domains between 0.91 and 1.00. These were not specialists in a research unit. They were clinicians in clinical services, working inside a defined interview.
And when the structure is taken away, it collapses. In the Australian study Module 1 quoted, 27 health professionals each watched two of nine recorded assessments and rated the person against the diagnostic criteria. There was 100 per cent agreement on the classification for only three of the nine cases, and only 24 per cent of the clinicians reached good or excellent agreement — kappa above 0.6 — with the original assessment team.
Put the three studies side by side and a single sentence falls out:
What predicts a sound assessment is not the profession. It is specific experience with the instrument and the population, and structure that does not depend on the individual’s memory or mood on the day.
That sentence is the reason Module 3 is a list of questions rather than a list of providers.
Diagram — C · What predicts a sound assessment. Start with a study designed to isolate exactly this. Researchers recorded seven administrations of the standard autism observation schedule and had a group of experts, working through a consensus process that included minute-by-minute video analysis, define a gold-standard coding for every case. They then showed those same recordings to a large group of clinicians drawn from routine care and collected 189 codings. The result was high variance: ordinary clinicians, using the best-known instrument in the field, coded the same recorded person differently from one another, and the accuracy of a person's coding depended on their experience with that instrument, on features of the case, and on how well the assessment had been administered in the first place. The authors' conclusion is blunt about what follows: specialisation in autism diagnosis has to be required, and specialised services are needed to guarantee a qualified differential assessment. The finding is not that the instrument is worthless. It is that an instrument is a thing a person operates, and operating it well is a skill acquired by doing it a great deal, with feedback, in a setting where somebody checks. Now the mirror image. When eleven clinicians across eight ordinary clinical sites each rated recorded administrations of a structured diagnostic interview — five raters per recording, no research-reliability training — agreement on the diagnostic classification came out at kappa 0.83, with agreement on the individual criterion domains between 0.91 and 1.00. These were not specialists in a research unit; they were clinicians in clinical services, working inside a defined interview. And when the structure is taken away it collapses: in the Australian study, 27 health professionals each watched two of nine recorded assessments, there was 100 per cent agreement on the classification for only three of the nine cases, and only 24 per cent of the clinicians reached kappa above 0.6 with the original assessment team. Put the three studies side by side and a single sentence falls out. What predicts a sound assessment is not the profession. It is specific experience with the instrument and the population, and structure that does not depend on the individual's memory or mood on the day.
D. Specialist versus generalist, on the evidence
The next question people ask is whether a generalist can do this at all. The evidence gives a more interesting answer than yes or no.
The cleanest test comes from a Canadian diagnostic study in which 17 general paediatricians and a multidisciplinary specialist team each assessed the same children, blind to each other’s conclusions. 106 children were referred; 72 of them (68 per cent) received an autism diagnosis from the specialist team.
Measured against the team, the paediatricians’ assessments had a sensitivity of 0.75 and a specificity of 0.79. But the numbers that matter to a person in the queue are the predictive values. The positive predictive value was 0.89 — when a paediatrician said autism, the specialist team agreed 89 per cent of the time. The negative predictive value was 0.60 — when a paediatrician said not autism, the specialist team agreed only 60 per cent of the time. The study also found that the paediatrician’s own certainty predicted accuracy: higher certainty was associated with a correct call.
Two population warnings before anyone carries that into their own case. These were children under five and a half, in Ontario, referred by the paediatricians themselves. Nothing in the study is a measurement of what a generalist achieves with a thirty-eight-year-old adult, and the adult case is harder, not easier, because there is no live developmental picture to observe and decades of adaptation on top. It is quoted here for the shape of the finding, not the digits.
And the shape is worth having, because it repeats everywhere the question is asked:
A generalist confirming the obvious case is usually right. A generalist ruling the case out is a much weaker signal — and if the person in front of them is an adult, a woman, articulate, or someone who has spent thirty years learning to pass, weaker still.
Diagram — D · A yes that holds, and a no that leaks. The next question people ask is whether a generalist can do this at all, and the evidence gives a more interesting answer than yes or no. The cleanest test comes from a Canadian diagnostic study in which 17 general paediatricians and a multidisciplinary specialist team each assessed the same children, blind to each other's conclusions. 106 children were referred; 72 of them, 68 per cent, received an autism diagnosis from the specialist team. Measured against the team, the paediatricians' assessments had a sensitivity of 0.75 and a specificity of 0.79 — but the numbers that matter to a person in the queue are the predictive values. The positive predictive value was 0.89: when a paediatrician said autism, the specialist team agreed 89 per cent of the time. The negative predictive value was 0.60: when a paediatrician said not autism, the specialist team agreed only 60 per cent of the time. The study also found that the paediatrician's own certainty predicted accuracy, with higher certainty associated with a correct call. Two population warnings before anyone carries that into their own case. These were children under five and a half, in Ontario, referred by the paediatricians themselves. Nothing in the study is a measurement of what a generalist achieves with a thirty-eight-year-old adult, and the adult case is harder rather than easier, because there is no live developmental picture to observe and decades of adaptation on top. It is quoted here for the shape of the finding, not the digits. And the shape is worth having, because it repeats everywhere the question is asked: a generalist confirming the obvious case is usually right, while a generalist ruling the case out is a much weaker signal — and if the person in front of them is an adult, a woman, articulate, or someone who has spent thirty years learning to pass, weaker still. What that costs in the adult population has also been measured. In two Italian university centres, 161 adults received a first autism diagnosis. The median age at diagnosis was 23, with a range from 18 to 55, and their median age at their first evaluation by a mental health professional was 13 — a gap of eleven years, in people who were already in the system. Only 33.5 per cent had never received a psychiatric diagnosis before; the rest had been given one or more other diagnoses first, most often intellectual disability, psychoses, personality disorders and depression. Those adults were not failed by an absence of clinicians. They were seen, and seen repeatedly, by professionals whose training did not include this question.
What that costs, in the adult population, has also been measured. In two Italian university centres, 161 adults received a first autism diagnosis. The median age at diagnosis was 23, with a range from 18 to 55 — and their median age at their first evaluation by a mental health professional was 13. A gap of eleven years, in people who were already in the system. Only 33.5 per cent had never received a psychiatric diagnosis before; the rest had been given one or more other diagnoses first, most often intellectual disability, psychoses, personality disorders and depression.
Those adults were not failed by an absence of clinicians. They were seen, and seen repeatedly, by professionals whose training did not include this question.
E. Why the person matters more than the instrument
Which returns us to Module 1, and lets us finish the argument it started.
The English ADHD guideline contains a sentence that could serve as the epigraph for this whole course: a diagnosis should not be made solely on the basis of rating scale or observational data. Rating scales are described in the same paragraph as helpful added tools. Tools. The judgement is somewhere else.
That is not modesty. It is what the evidence in section C forces. An instrument administered by someone who has run it forty times is a different instrument from the same form administered by someone running it for the third time, and the 189-coding study is a direct measurement of that difference.
There is one more thing worth knowing, and it is written into guidance rather than inferred from it. The English autism guideline explicitly tells services to consider obtaining a second opinion, including referral to another specialist team, where there is disagreement about the diagnosis inside the team, disagreement with the person or their family, a lack of local expertise, or a complex co-occurring condition.
Sit with that for a moment. The guideline anticipates that a service may not have the expertise the case needs, and instructs it to say so. If the system’s own rules treat one assessor’s conclusion as revisable, you are entitled to treat it the same way. A second opinion is not an accusation. It is a step the guidance describes.
Diagram — E · Six questions, in the order you ask them. One: ask who, specifically, will do the assessing, by name and by role. Not our clinical team. One question: which professional will conduct my assessment, what is their registration, and who signs the report? A service that cannot answer that on the phone has told you something. Two: ask how many adults they assess, not how many years they have practised. Years of general practice is the weakest of the available proxies; volume with your population is the strongest available signal, and the coding study is why. Roughly how many adult autism or ADHD assessments does this assessor complete in a year is a fair, answerable question. Three: ask what structure the assessment runs on — which structured interview, which rating scales, whether an informant is involved, and whether more than one person contributes to the conclusion. Structure is what took ordinary clinicians from 24 per cent agreement to 0.83. Four: settle the prescribing question before you book, if medication is part of why you are going. If the assessor does not prescribe, ask who does, whether they accept this assessor's diagnosis, and what that costs; two-step routes are perfectly normal, and discovering one late is not. Five: check what the receiving institution actually requires — employer, university, insurer, benefits system — in writing, before payment. This is the single most common way people buy the wrong assessment. Six: treat a no from a generalist differently from a no from a specialist service. The predictive-value asymmetry is the reason: confirmation from a generalist tends to hold, exclusion from a generalist is the weaker of the two calls, and the case for a second opinion is correspondingly stronger. One more thing is written into guidance rather than inferred from it: the English autism guideline explicitly tells services to consider obtaining a second opinion, including referral to another specialist team, where there is disagreement about the diagnosis inside the team, disagreement with the person or their family, a lack of local expertise, or a complex co-occurring condition. If the system's own rules treat one assessor's conclusion as revisable, you are entitled to treat it the same way. A second opinion is not an accusation; it is a step the guidance describes.
F. What helps
1. Ask who, specifically, will do the assessing — by name and by role.
Not “our clinical team.” One question: which professional will conduct my assessment, what is their registration, and who signs the report? A service that cannot answer that on the phone has told you something.
2. Ask how many adults they assess, not how many years they have practised.
Years of general practice is the weakest of the available proxies. Volume with your population is the strongest available signal, and the coding study is why. Roughly how many adult autism or ADHD assessments does this assessor complete in a year? is a fair, answerable question.
3. Ask what structure the assessment runs on.
Which structured interview, which rating scales, whether an informant is involved, and whether more than one person contributes to the conclusion. Section C is the argument: structure is what took ordinary clinicians from 24 per cent agreement to 0.83.
4. Settle the prescribing question before you book, if medication is part of why you are going.
If the assessor does not prescribe, ask who does, whether they accept this assessor’s diagnosis, and what that costs. Two-step routes are perfectly normal. Discovering one late is not.
5. Check what the receiving institution actually requires.
Employer, university, insurer, benefits system. In writing, before payment. This is the single most common way people buy the wrong assessment.
6. Treat a “no” from a generalist differently from a “no” from a specialist service.
The predictive-value asymmetry in section D is the reason. Confirmation from a generalist tends to hold. Exclusion from a generalist is the weaker of the two calls, and the case for a second opinion is correspondingly stronger.
Module 3 turns all of this into the practical question — which routes exist, what they cost, how long they take, and the list of questions to put to any service before you commit.
Up next
Module 3 — Choosing a Service, and What to Ask