Show the full module text
Module 7 — Reading the Report
The anatomy of a report section by section, which section answers which question, how to read a score without over-reading it, what the severity levels do and do not mean, and what to do about a factual error.
The Assessment · Part Three — After
The big idea
Short on capacity today? The big idea: A report is not a verdict written down. It is a document with about seven parts, and each part answers a different question and has a different shelf life.
Most people read it the way you would read an exam result: find the line with the diagnosis, feel something, put it in a drawer. That is the one section other people will barely use. The parts that will still be doing work in five years — when an employer, a university, a prescriber or a therapist wants something from it — are the description of how you function and the recommendations, and those are the sections that get skimmed.
This module is a way of reading it that treats it as what it is: a working document about you, written by somebody else, that you are entitled to check.
Step 1 — The lesson
A. The anatomy, and which question each section answers
Length varies enormously — some services issue three pages, some forty. Section names vary too. The functions do not.
Section. Referral question. The question it answers. Why was this assessment done, and what was I asked to answer? What to check. That it is the question you asked. Module 1’s one-sentence exercise is the test.
Section. Background and history. The question it answers. What was described, and by whom? What to check. Facts, dates, and attribution — what you said, what an informant said, what a document says.
Section. Instruments administered. The question it answers. What evidence was collected, in what modality, from whom? What to check. Which version of each tool, who completed it, and the dates.
Section. Results. The question it answers. What did those instruments return? What to check. That every score has a sentence around it, not just a number.
Section. Clinical formulation. The question it answers. How does the assessor make sense of the whole picture? What to check. Whether it says what converged, what did not, and what weight was given to each.
Section. Diagnosis and rationale. The question it answers. Which criteria were judged met, on what evidence? What to check. That evidence is mapped to criteria, and the differential is named.
Section. Recommendations. The question it answers. What should happen next, and who needs to act? What to check. Whether they are concrete enough for somebody else to implement.
Three of those deserve more than a table row.
The referral question governs everything after it. A report answers the question it was set. If you went in asking why you cannot hold a job past eighteen months and the report answers is this person autistic, it may be a perfectly good report and still not be the document you needed. This is much easier to fix at the draft stage than afterwards.
The formulation is the most valuable section and the least read. It is where the assessor does the actual work Module 1 described: five kinds of evidence, weighed against each other, including the places they disagreed. If the formulation is a paragraph restating the diagnosis, you have been given a conclusion without its reasoning — and reasoning is what you paid for.
The rationale should be legible to a stranger. Not “the presentation is consistent with autism spectrum disorder” but which criteria were judged met, from what evidence, and what else was considered and set aside. English guidance requires a comprehensive assessment to enquire into a specific list — core features present in childhood and continuing into adulthood, developmental history where possible, functioning at home and in work or education, other neurodevelopmental conditions, mental and physical health, sensory sensitivities — and to assess for differential diagnoses and coexisting conditions. Those requirements are, in effect, a contents list. If a whole heading is missing from the report, that is a fair question to ask.
A good report also has a limitations paragraph: what could not be established, and why. No informant. Records unavailable. Remote administration of an instrument standardised in person. A limitations paragraph is a sign of care, not weakness.
Diagram — A · One flash, one steady flame. A report is not a verdict written down. It is a document with about seven parts, and each part answers a different question and has a different shelf life. The referral question asks why this assessment was done and what the assessor was asked to answer; check that it is the question you asked, because the referral question governs everything after it, and a mismatch is much easier to fix at draft stage than afterwards. Background and history asks what was described and by whom; check facts, dates and attribution — what you said, what an informant said, what a document says. Instruments administered asks what evidence was collected, in what modality and from whom; check which version of each tool, who completed it, and the dates. Results asks what those instruments returned; check that every score has a sentence around it and not just a number. The clinical formulation asks how the assessor makes sense of the whole picture, and it is the most valuable section and the least read: it is where five kinds of evidence are weighed against each other, including the places they disagreed, and if it is a paragraph restating the diagnosis you have been given a conclusion without its reasoning. Diagnosis and rationale asks which criteria were judged met on what evidence; the rationale should be legible to a stranger, mapping evidence to criteria and naming the differential. Recommendations asks what should happen next and who needs to act; check that they are concrete enough for somebody else to implement. Most people read a report the way you would read an exam result: find the line with the diagnosis, feel something, put it in a drawer. That is the one section other people will barely use. The parts that will still be doing work in five years — when an employer, a university, a prescriber or a therapist wants something from it — are the description of how you function and the recommendations, and those are the sections that get skimmed. Ask for a functional-impact paragraph you could hand over on its own, and ask for it before the report is finalised.
B. Reading scores without over-reading them
A standardised score is a position relative to a comparison group. It is not a quantity of you. Everything else in this section follows from that one sentence.
Ask who the comparison group was. A score says where you sit against the people the instrument was normed on. If that sample was mostly children, or mostly men, or drawn from a country and a decade you have nothing in common with, the number still computes and means less than it appears to.
Treat a score as one vantage point, because that is measurably what it is. In a study at a specialised adult autism diagnostic service, 110 adults completed a repetitive-behaviours questionnaire about themselves while a relative or long-term friend completed the informant version. Self and informant totals correlated at 0.71. Both versions also correlated with the clinician’s structured-interview score for the same domain — at 0.45 to 0.54. Same people, same construct, three vantage points, moderate agreement between all of them. That is not error in any of the three. It is what happens when you measure a life from three directions.
Diagram — B · Three views of the same jug. A standardised score is a position relative to a comparison group. It is not a quantity of you, and everything else follows from that one sentence. Ask who the comparison group was: a score says where you sit against the people the instrument was normed on, and if that sample was mostly children, or mostly men, or drawn from a country and a decade you have nothing in common with, the number still computes and means less than it appears to. Treat a score as one vantage point, because that is measurably what it is. In a study at a specialised adult autism diagnostic service, 110 adults completed a repetitive-behaviours questionnaire about themselves while a relative or long-term friend completed the informant version. Self and informant totals correlated at 0.71. Both versions also correlated with the clinician's structured-interview score for the same domain, at 0.45 to 0.54. Same people, same construct, three vantage points, moderate agreement between all of them. That is not error in any of the three; it is what happens when you measure a life from three directions. No score is a diagnosis — English guidance for ADHD states that a diagnosis should not be made solely on the basis of rating scale or observational data, and describes rating scales as helpful added tools. And never quote a bare score at anybody: not to an employer, not to a doctor, not to yourself at three in the morning. If the report gives a range or a confidence interval, that range is part of the finding, and it is there precisely because the instrument's makers know a single number overstates its own precision. Scores also age faster than descriptions. A number from four years ago tells a reader what an instrument returned on a particular Tuesday; a paragraph describing what you cannot sustain in an open-plan environment is still true.
No score is a diagnosis. English guidance for ADHD states that a diagnosis should not be made solely on the basis of rating scale or observational data, and describes rating scales as helpful added tools. Module 5 showed what happens when a threshold is treated as an answer: in 88 adults, the best-known observation schedule ran at 92 per cent sensitivity and 57 per cent specificity.
Never quote a bare score at anybody. Not to an employer, not to a doctor, not to yourself at three in the morning. If the report gives a range or a confidence interval, that range is part of the finding, and it is there precisely because the instrument’s makers know a single number overstates its own precision.
And scores age faster than descriptions. A number from four years ago tells a reader what an instrument returned on a particular Tuesday. A paragraph describing what you cannot sustain in an open-plan environment is still true. When you eventually need to hand something over, the second kind of writing is what you will want.
C. Severity levels, and what they do not mean
If your report carries a severity level, this section is the most important one in the module. If it does not — and many adult reports do not — that is not a defect.
The levels came in with the fifth edition of the diagnostic manual in 2013 and are defined by levels of needed support across the two diagnostic domains. As reproduced in a 2026 study of how they are used, they are: “requiring support” (Level 1), “requiring substantial support” (Level 2), and “requiring very substantial support” (Level 3). The same paper notes they arrived “with minimal description of the criteria for categorizing three levels of severity” — and records that “There was no change regarding severity levels in the latest DSM update, the DSM-5 Text Revision (DSM-5-TR), released in 2022.” A report citing either edition should therefore carry the same three labels.
Diagram — C · The same plant, different weather. If your report carries a severity level, this is the most important thing in the module; if it does not, and many adult reports do not, that is not a defect. The levels came in with the fifth edition of the diagnostic manual in 2013 and are defined by levels of needed support across the two diagnostic domains. As reproduced in a 2026 study of how they are used, they are: requiring support (Level 1), requiring substantial support (Level 2), and requiring very substantial support (Level 3). The same paper notes they arrived with minimal description of the criteria for categorizing three levels of severity, and records that there was no change regarding severity levels in the latest update, the text revision released in 2022 — so a report citing either edition should carry the same three labels. Now the three things a level does not tell anybody. It does not say how autistic somebody is: there is no such quantity. A review of how severity is conceptualised found that research stratifying people by autism severity is mostly about core symptomatology, whereas the manual's version is based on level of needed support, which is also impacted by cognitive, language, behavioral, and adaptive functioning. Two different concepts wearing the same word. A person can be described as needing less support and still experience the traits intensely, because support need is a function of the person and their circumstances, and circumstances are half of it. It is not stable: support needed changes with environment, life stage, health, whether accommodations exist and whether anyone is currently helping, so a level records a judgement about a point in time. Read it as a photograph, not a measurement of your permanent nature. And Level 1 is not mild autism, and it is not a lower rank — two people given the same level can need entirely different things. If a level in your report is going to be read by somebody who makes decisions about you, the sentence you want alongside it is the one about what you actually need, in what setting, to do what.
Now the three things a level does not tell anybody.
It does not say how autistic somebody is. There is no such quantity. A review of how severity is conceptualised found that research stratifying people by autism severity is mostly about core symptomatology, whereas the manual’s version “is based on level of needed support, which is also impacted by cognitive, language, behavioral, and adaptive functioning.” Two different concepts wearing the same word. A person can be described as needing less support and still experience the traits intensely — support need is a function of the person and their circumstances, and circumstances are half of it.
It is not stable. Support needed changes with environment, life stage, health, whether accommodations exist and whether anyone is currently helping. A level records a judgement about a point in time. Read it as a photograph, not a measurement of your permanent nature.
It is applied inconsistently by the professionals writing it. Population surveillance data in the United States found that fewer than half — 40.4 per cent — of children with a documented autism diagnosis had any severity level specified in their records at all, with variation between sites from 4.8 per cent to 73.2 per cent. The authors’ conclusion is the sentence to carry: use of the levels “varied widely, limiting their potential utility in identifying needed services and supports.” An earlier study classified 726 participants as mildly, moderately or severely impaired on three different bases — autism symptoms, cognitive skills and adaptive functioning — and found the classifications did not line up, which is why its authors called for “a clearly elucidated method of classifying level of support in ASD diagnosis.”
Both of those samples were children, and that matters — it means there is even less standing behind a level assigned to an adult, not more.
Diagram — D · The tags that never got tied. The third thing a severity level does not tell anybody is that it is applied consistently, because it is not. Population surveillance data in the United States found that fewer than half — 40.4 per cent — of children with a documented autism diagnosis had any severity level specified in their records at all, with variation between sites from 4.8 per cent to 73.2 per cent. The authors' conclusion is the sentence to carry: use of the levels varied widely, limiting their potential utility in identifying needed services and supports. An earlier study classified 726 participants as mildly, moderately or severely impaired on three different bases — autism symptoms, cognitive skills and adaptive functioning — and found the classifications did not line up, which is why its authors called for a clearly elucidated method of classifying level of support in ASD diagnosis. Both of those samples were children, and that matters: it means there is even less standing behind a level assigned to an adult, not more. The practical consequence is not that a level is worthless. It is that a level cannot be relied on to carry your needs to somebody who has to act on them, and the sentence describing what you actually need, in what setting, to do what, can.
So: Level 1 is not “mild autism”, and it is not a lower rank. Two people given the same level can need entirely different things. If a level in your report is going to be read by somebody who makes decisions about you, the sentence you want alongside it is the one about what you actually need, in what setting, to do what.
D. The part other people will actually use
Here is the practical hinge between this module and the next.
The diagnosis line is what you will remember. The sections a stranger will act on are the description of function and the recommendations — because that is how the systems that provide anything are built. English criteria for ADHD, for instance, require symptoms to cause “at least moderate psychological, social, or educational or occupational impairment” and to be occurring in “2 or more important settings”. The evidence about how you function across settings is therefore already in the report by necessity. It may simply be scattered through it, in a form nobody outside can use.
So read the recommendations as a specification, and check three things.
Are they concrete? “Reasonable adjustments should be considered” is not an adjustment. “Written instructions following verbal briefings; advance notice of schedule changes; a non-open-plan working location” is.
Do they cover what was found alongside the diagnosis? A set of published recommendations for diagnosing autism in adults says clinicians should tailor interventions to include psychoeducation for newly diagnosed autistic adults, and should “provide evidence-based treatment recommendations to manage possible co-occurring mental health needs.” The co-occurring things are usually where the immediate suffering is.
Is there a functional-impact paragraph you could hand over on its own? If not, ask for one. It is a reasonable request, it is far easier to add while the report is being finalised, and Module 8 is about what it then unlocks.
Two administrative points that cost people months. Ask who else receives a copy and whether anything is sent to your medical record automatically — that is your decision to make knowingly. And keep the original file, because some institutions require a report to be less than a certain number of years old and you may need the same document more than once.
E. A factual error in the report
Separate three things before you write to anybody, because they have three different remedies.
- A factual error. A wrong date, a job you never had, a sibling who does not exist, a condition attributed to you that you do not have, a quotation you did not say.
- A disagreement with a judgement. The formulation, the weight given to something, the conclusion itself.
- An omission. Something you reported that is not in the document.
Facts have a route. Under the United Kingdom’s retained General Data Protection Regulation, Article 16 states that a person “shall have the right to obtain from the controller without undue delay the rectification of inaccurate personal data concerning him or her” — and, in its second sentence, “the right to have incomplete personal data completed, including by means of providing a supplementary statement.” That second sentence is the remedy for omissions, and it is the one people never quote.
In the United States, the health-privacy rule gives an individual the right to have a covered entity “amend protected health information or a record about the individual in a designated record set”. The entity may refuse — including where it determines the record “Is accurate and complete” — and must act on a request “no later than 60 days after receipt”. If it refuses, it must permit you “to submit a written statement disagreeing with the denial”, and must tell you how to complain.
The shape common to both, and worth having even if neither applies where you live: facts get corrected; judgements usually do not — but your disagreement can be attached to the file so that it travels with the document. That is a real outcome. A future reader sees both.
Diagram — E · Bound in under the same cord. Separate three things before you write to anybody, because they have three different remedies. A factual error: a wrong date, a job you never had, a sibling who does not exist, a condition attributed to you that you do not have, a quotation you did not say. A disagreement with a judgement: the formulation, the weight given to something, the conclusion itself. And an omission: something you reported that is not in the document. Facts have a route. Under the United Kingdom's retained General Data Protection Regulation, Article 16 states that a person shall have the right to obtain from the controller without undue delay the rectification of inaccurate personal data concerning him or her — and, in its second sentence, the right to have incomplete personal data completed, including by means of providing a supplementary statement. That second sentence is the remedy for omissions, and it is the one people never quote. In the United States, the health-privacy rule gives an individual the right to have a covered entity amend protected health information or a record about the individual in a designated record set; the entity may refuse, including where it determines the record is accurate and complete, and must act on a request no later than 60 days after receipt. If it refuses, it must permit you to submit a written statement disagreeing with the denial, and must tell you how to complain. The shape common to both, and worth having even if neither applies where you live: facts get corrected; judgements usually do not — but your disagreement can be attached to the file so that it travels with the document, and a future reader sees both. How to do it: write once, in a numbered list — page, the sentence as written, the correction, and the evidence. Keep opinion out of the factual list and put any disagreement in a clearly separate paragraph, so that the correctable items cannot be dismissed along with the arguable ones. Ask for a corrected version, and ask what happens to copies already issued. And do it early: correcting a document before it goes to an employer, a university or a health system is a different task from retrieving it afterwards. None of this is legal advice, the specifics differ by country and often within one, and the two instruments above are quoted as examples of shape rather than as rules that apply to you.
How to do it. Write once, in a numbered list: page, the sentence as written, the correction, and the evidence. Keep opinion out of the factual list — put any disagreement in a clearly separate paragraph, so that the correctable items cannot be dismissed along with the arguable ones. Ask for a corrected version, and ask what happens to copies already issued.
Do it early. Correcting a document before it goes to an employer, a university or a health system is a different task from retrieving it afterwards.
None of this is legal advice, the specifics differ by country and often within one, and the two instruments above are quoted as examples of shape rather than as rules that apply to you.
F. What helps
1. Read the formulation first and the diagnosis line last.
It is the section that shows the reasoning, including where the evidence disagreed. If it is thin, that is the thing to ask about.
2. Check that the referral question is the one you asked.
Compare it against the sentence you wrote in Module 1 about what you wanted the assessment to change. A mismatch is much cheaper to fix at draft stage.
3. Go through the history with a pen, and number every factual error.
Page, sentence as written, correction, evidence. Facts in one list; disagreements in a separate paragraph.
4. Never send a bare score anywhere.
A number without its comparison group, its date and the sentence around it will be over-read by whoever receives it.
5. If there is a severity level, read it as support needed at a point in time.
Not a rank, not a quantity of autism, not a permanent property. And if a decision-maker will see it, make sure a plain description of what you need sits next to it.
6. Ask for a functional-impact paragraph and concrete recommendations before the report is finalised.
This is the part other people act on, and it is far easier to request now than to extract later.
Module 8 is what a diagnosis does and does not change — the documentation that actually gets accommodations, what the evidence says changes after a late diagnosis, and what does not.
Up next
Module 8 — What a Diagnosis Does and Does Not Change