Show the full module text
Module 2 — RAADS-14 — A Short Autism Screener for Adults
Where the fourteen questions came from, what the three groupings underneath them ask, why the famous 97 per cent sensitivity travels everywhere and the specificity never does, and what a score over the line honestly buys you.
Autistic Self-Discovery · Part Two — The Broad Screeners · 13 min read, about 26 min with the workbook
The big idea
If you have ever taken a fourteen-question quiz at midnight, scored well over the line, and then spent three days wondering whether that meant anything at all — this module is about that questionnaire. It is a real published instrument with real numbers behind it. Some are much better than others, and the good one is the one everybody quotes.
The average non-autistic young adult in that sample scored 13.9 on a screen whose threshold is 14. In that population the cut-off was not sitting at the edge of the distribution. It was sitting in the middle of it.
The instrument
RAADS-14 — Ritvo Autism and Asperger Diagnostic Scale
Who it is for: Adults who want the quickest honest look, including at childhood
Length: 14 statements, 3–5 minutes
What you get back: One total and the band it falls in
The cut-off line: 14 of 42. Section C explains what a 14 does and does not buy you.
People usually arrive at this one because…
You want a first look and you do not have the capacity for fifty questions today.
Something has felt different since childhood, and you want a tool that actually asks about then, not just now.
You have taken a longer questionnaire before and want to see whether a shorter one says the same thing.
Someone who knows you well has raised it, and you want somewhere to start.
You are preparing for an appointment and want a number you can name out loud.
What it does well. It is the fastest honest look here, and the only one in this part that asks whether traits were present in childhood as well as now — which is exactly what an assessment will want to know.
Where it falls down. It cannot tell autism apart from other psychiatric conditions. Specificity against ADHD was 0.46, meaning more than half of non-autistic adults with ADHD screened positive.
How much weight it carries. Tier 2 of 4 — published and peer-reviewed, but researchers still disagree about what it actually measures.
A low score is not a no. Fourteen questions cannot hold a whole life. A score under the line means these particular fourteen did not catch it, not that there is nothing to catch.
FAQ
It asks about childhood and I cannot remember. What do I do?
Answer for what you can remember and treat the gaps as gaps. Old school reports, a parent or an older sibling can fill some of it in later. Nobody has a clean memory of being eight, and an assessor knows that.
What about the RAADS-R, the 80-question one?
It is the long form this was cut down from, and it is the version most sites push. There is a box at the end of the lesson explaining why this course uses the short one: given to people on their own, the RAADS-R flagged 49 of 50 people in the only service that tested it properly.
What happens to my answers?
Your screener answers and your score are kept, so that if you ever want to talk the result through with someone, it is there to look at. Your workbook is different: it saves in this browser only and never reaches us.
Step 1 — The lesson
A. Fourteen questions, cut from eighty
The RAADS-14 is not an original questionnaire. It is a short form, and knowing what it was cut down from tells you most of what you need to read your own score properly.
The parent instrument is the RAADS-R, an eighty-item scale published in 2011 by Riva Ritvo and colleagues at nine research centers across three continents. They compared 201 autistic adults with 578 people who were not autistic, set the threshold at 65, and reported 97 per cent sensitivity and 100 per cent specificity.1 Extraordinary figures, and they arrive with a detail almost never repeated alongside them: a clinician stayed with every participant while they answered, on hand to explain any unclear item. The scale was never designed to be filled in alone, on a phone, at one in the morning.
In 2013 a Swedish team — Jonna Eriksson, Lisa Andersen and Susanne Bejerot — had a narrower problem. Adult psychiatric clinics were full of people being treated for depression, anxiety and ADHD, some of whom were autistic and had never been recognized, and eighty items was too long to hand to everyone who walked in. So they cut it in three passes: down to eighteen items with the strongest discrimination, then to fourteen after four of those performed badly in psychiatric patients, then a final validation. In all, 1,233 adults took part — 643 with a psychiatric diagnosis and 590 without.2
Diagram — A · Fourteen questions, cut from eighty. The short form was built to work inside a psychiatric clinic. Everything good about it and everything awkward about it follows from that one design decision.
Read that purpose again, because it is the key to the whole module. The RAADS-14 was built to sort within a clinic — to find the autistic people among adults already unwell in some way. It was not built for the general public and certainly not for the internet. That does not make it useless to you. It makes it a tool with a designed lean, and the lean is toward catching people rather than toward being right about them. Once you know which way an instrument leans, you can read your own result properly instead of at face value.
B. Three things, not one
Fourteen items, but they do not spread evenly across autistic experience. A factor analysis of the answers produced three groupings, which the Swedish team named mentalizing deficits, social anxiety and sensory reactivity.2
Seven of the fourteen sit in the first group: reading a face or a pair of hands, knowing when to talk and when to listen, working out what somebody expects of you, taking a sentence at its literal word, seeing the detail before the whole. Four sit in the second: small talk, working in groups, finding friendship faintly mysterious, not knowing how to behave. Three sit in the third: an ordinary fabric that feels wrong against skin, covering your ears against a noise that hurts, needing to be by yourself once too much has come in.
Diagram — B · Three things, not one. Half the questionnaire is about mentalising. A fifth is sensory life. The rest is social discomfort, which is also the name of a separate condition. Nothing here asks what you are interested in.
Now look at what is not there. One item asks about being upset when the way you like to do things is suddenly changed. That is the whole of the restricted and repetitive side of autistic life, in a single question. The parent scale had four subscales and one was Circumscribed Interests.1 It did not survive the cut, and neither did most of the language subscale. So if the thing you most recognize in yourself is the interest that has organized twenty years of your life, or the particular way you use words, this questionnaire will not ask, and your total will read lower than your life would suggest. That is a gap in what the instrument covers, not a flaw you can answer your way out of.
There is a second thing worth noticing in those names. One of the three factors is called social anxiety, which is a condition in its own right, with its own treatment and its own criteria. A person with social anxiety and no autistic history will answer four of these fourteen items exactly the way an autistic person does.
A large 2024 study complicates the picture usefully. Alexandra Sturm and colleagues gave the RAADS-R and the RAADS-14 to 839 adults in four groups — formally diagnosed autistic adults, adults who identify as autistic without an assessment, adults exploring the question, and non-autistic comparisons. Both scales behaved as unidimensional instruments, with no systematic bias by age, gender, or diagnosis versus self-identification, and the researchers suggested the four response options could be cut to two without losing anything.3 So hold both findings at once. The three factors are a good way to talk about your answers. They are not three separate scores, and the RAADS-14 was never built to give you a profile.
C. Two numbers, and only one of them travels
Every page that mentions this screener mentions 97 per cent sensitivity. The figure is real. It is also only half of a pair, and the other half almost never travels with it.
Sensitivity is the proportion of autistic people a screen catches. At a cut-off of 14 or above, out of a maximum of 42, the Swedish study reported 0.97.2 Very few autistic people in that sample scored under the line, which is why a low score here carries more information than a high one.
Specificity is the other side: the proportion of non-autistic people a screen correctly leaves alone. In the same study it was 0.46 in the ADHD sample, 0.64 in the sample with other psychiatric conditions, and 0.95 in the non-psychiatric controls.2 In plain words: among adults who had ADHD and were not autistic, more than half scored at or above the very threshold being read as a sign of autism. Among psychiatric patients with neither condition, roughly a third did.
Diagram — C · Two numbers, and only one of them travels. A screen tuned this way is doing what its designers intended. The cost of missing an autistic adult in a clinic is high; a false alarm costs one more conversation. That trade is paid by everyone who screens positive and is not autistic.
The medians tell the same story from another angle. In the validating sample the median total was 32 in the autistic group, 15 in the ADHD group and 11 in the group with other psychiatric conditions.2 Look at the middle figure: the typical adult with ADHD and no autism sat above the cut-off. Fourteen out of forty-two is a low bar, and it was meant to be.
Where the screen does discriminate well is against people who are not in a clinic at all: the area under the curve was 0.99 against the 590 non-psychiatric controls, against 0.91 for other psychiatric patients and 0.88 for ADHD.2 A 2017 European review of nine adult instruments placed the RAADS-R and the RAADS-14 among the small group with satisfactory or intermediate psychometric properties on strong or moderate evidence, while noting that risks of bias and applicability limited what could be concluded.4 Both things are true. This is one of the better short instruments available, which is a statement about the field as much as about the questionnaire.
One more finding belongs here, because it matters to a great many people reading this. In the original sample, women scored higher than men on sensory reactivity in the autistic group, the ADHD group and the non-psychiatric group alike.2 A 2026 Irish study of young adults moving from child mental health services into an adult ADHD service found it again: autistic women scored higher than autistic men on every domain, and sensory sensitivities were higher in women whether or not they were autistic.5 If part of a screen separates women from men regardless of diagnosis, part of what it is measuring is not autism.
D. What a score of 14 actually buys you
These are the numbers that decide what your result is worth. They are not in the abstract, and they are not on the quiz sites.
Across 2019 and 2020, an adult psychiatric outpatient clinic in Sweden handed the RAADS-14 to 562 new patients as they arrived. 304 filled it in, and of those, 197 scored at or above 14. Forty-eight went on to a full assessment, and 26 met criteria for an autism spectrum condition.6 Read that again: roughly two out of every three new patients at an ordinary adult psychiatric clinic screened positive.
Diagram — D · What a score of 14 actually buys you. This is the real-world shape of high sensitivity and modest specificity. It is working as designed, and it is why a positive result is a door rather than an answer.
What happened next is more striking still. Susanne Bejerot, who co-authored the RAADS-14 itself, wrote to the journal with Lena Nylander to argue that the prevalence figure drawn from that study was too high. Their letter recalculates the minimum prevalence at no more than 8.6 per cent, and says of the screen that its specificity is low and that "false positives are likely a pervasive issue".7 The instrument's own author, in print, in the same journal.
Then there is the replication. In 2023 a Russian team validated a translated RAADS-14 in 1,667 adults, including 794 non-psychiatric young adults and 49 with an established autism spectrum diagnosis. Discrimination was good overall, with an area under the curve of 0.92. But at the original threshold of 14, sensitivity was 93.9 per cent and specificity 56.4 per cent, so the team recommended a cut-off of 22 instead (85.7 per cent sensitivity, 85.4 per cent specificity). The mean total in their non-psychiatric group was 13.9.8
The parent scale has been tested in an ordinary clinical queue too, and the result is bracing. Fifty adults referred to an NHS adult ADHD and autistic assessment service in Yorkshire completed the RAADS-R. Forty-nine of the fifty scored above the threshold of 65; seventeen received a diagnosis. Sensitivity 100 per cent, specificity 3.03 per cent, positive predictive value 34.7 per cent, and the authors concluded the scale lacks predictive validity for adults awaiting assessment.9 That is the RAADS-R rather than the RAADS-14, but it is the family, and it is what happens when a high-sensitivity screen meets people who are all there for the same reason.
So what does a score of 14 or over mean? Roughly what the live screener says: possible, and worth taking to somebody. Not likely, not probably, and among people who already suspect something it narrows very little down. The Royal College of Psychiatrists puts the general point plainly: a single structured questionnaire is unlikely to be reliable enough either to identify adults who are not autistic or to tell autism apart from mental health conditions.10
And a score under 14? That is the more useful half of this instrument, and it still is not a verdict. Of the 135 autistic adults in the original study, nine scored below the cut-off and two of them scored zero.2 Sensitivity of 0.97 leaves three autistic people in a hundred on the wrong side of the line. In fairness, the authors also argued the opposite way: some of their psychiatric controls may themselves have been undiagnosed autistic adults, which would make 46 to 64 per cent an underestimate.2 Nobody has settled that. It is one more reason to hold the number loosely.
INSIDE THE INSTRUMENT
RAADS-14 Screen — Ritvo Autism Asperger Diagnostic Scale, 14-item screening version
Where it came from
Derived from the 80-item RAADS-R (Ritvo et al., 2011; 201 ASD, 578 comparison, nine centers, cut-off 65, sensitivity 97%, specificity 100%, administered with a clinician present).1 Eriksson, Andersen and Bejerot reduced it in three phases across 1,233 adults, 643 with a psychiatric diagnosis and 590 without.2 Open access in Molecular Autism, 2013. Free to use.
What it is made of
Fourteen self-report items, each scored 0 to 3, total 0 to 42, three to five minutes. Factor analysis yielded mentalizing deficits (7 items), social anxiety (4) and sensory reactivity (3), full-scale alpha 0.90, every factor above 0.7.2 One item covers changed routine; circumscribed interests, repetitive movement and the RAADS-R language domain are otherwise absent. A five-item rapid form exists: cut-off 4 of 15, sensitivity 0.93, specificity 0.45 (ADHD), 0.49 (other psychiatric), 0.90 (non-psychiatric).2
How well it performs
At 14 or above: sensitivity 0.97; specificity 0.46 (ADHD), 0.64 (other psychiatric disorders), 0.95 (non-psychiatric controls). AUC 0.88, 0.91 and 0.99 respectively. Median totals: ASD 32, ADHD 15, other psychiatric 11.2 A 2017 systematic review placed the RAADS-R and RAADS-14 among the few adult instruments with satisfactory or intermediate psychometric values on strong or moderate evidence, while noting risks of bias and applicability concerns.4 A 2024 item-response study in 839 adults found both scales unidimensional with no systematic bias by age, gender or diagnosis versus identity.3
Where it was validated
Swedish adult psychiatric outpatients and non-psychiatric controls, diagnosed under DSM-IV-TR, with no confirmatory assessment and no matching for sex, age or intelligence.2 The only translation validation reporting full accuracy statistics is Russian: 1,667 adults, AUC 0.92, but specificity of 56.4% at the original cut-off of 14 and a recommended cut-off of 22 (sensitivity 85.7%, specificity 85.4%), non-psychiatric mean 13.9.8 There is no US or UK accuracy validation of the RAADS-14 itself and no general-population norms.
What it cannot do
It cannot separate autistic adults from other psychiatric patients with any confidence. Specificity of 0.46 against ADHD means more than half of non-autistic adults with ADHD screened positive.2 In a real Swedish outpatient clinic, 197 of 304 responders scored at or above 14 while 26 confirmed cases emerged from the whole screened group.6 Bejerot herself, replying in print, called the specificity of the RAADS low and false positives "likely a pervasive issue".7
It has no test-retest data, no published norms and no independently reproduced cut-off. The development paper reports internal consistency and factor structure but no stability coefficient, and there is no normative table by age or sex for any population, so a raw total cannot be converted to a percentile. The one independent validation with accuracy statistics recommended 22 rather than 14.8 Using 14 outside a psychiatric clinic imports a threshold calibrated for a base rate that does not apply.
It does not cover the restricted and repetitive domain, and it was validated against DSM-IV-TR. One item on changed routine is the whole of criterion B coverage; circumscribed interests, a full subscale on the parent instrument, were dropped.1 An interest-led, routine-led adult can screen negative here and still meet criteria on assessment. Group membership in the development study rests on the older criteria and on routine clinical diagnoses, with selection bias and absent confirmatory assessment named by the authors.2
Its accuracy in women has never been reported separately. Sex differences in subscale scores appear in both the original sample and a 2026 Irish ADHD cohort, where sensory scores were higher in women regardless of autistic status;5 a 2024 item-response analysis found no differential item functioning by gender.3 No study has published sensitivity and specificity for women alone at any cut-off.
What a clinician does with it
Use it as designed: a low-cost, high-sensitivity trigger inside a service where autistic adults are likely to be present and unrecognised. A negative screen is worth more than a positive one and still does not rule out; a positive screen prompts a developmental history, not a finding to record. NICE guidance is that a case-identification tool is not used alone,11 and the Royal College of Psychiatrists states that a single structured questionnaire is unlikely to be reliable enough either to identify adults who are not autistic or to differentiate autism from mental health conditions.10 Item-level answers beat the total.
Validity tier: 2 — peer-reviewed and published, but calibrated in one clinical sample and not independently reproduced. It is properly validated, freely available, and its sensitivity holds up; its threshold has been examined by exactly one independent laboratory, which recommended a different one, and its own author has published a caution about its specificity.
Diagram — E · What the paperwork shows. Read the blank lines rather than the ticks. This screen has never had a stability check, never had population norms, and has never had its accuracy reported for women on their own.
STRENGTHS LENS
You noticed something no questionnaire told you to notice.
Almost everyone who ends up on a page like this arrived the same way: years of small evidence, quietly collected, long before any instrument was involved. The exhaustion after a birthday party. The three shirts you will actually wear. That is sustained self-observation, and it is why the questionnaire had anything to confirm.
The fourteen items are a borrowed vocabulary for something you had already worked out. Adults self-identifying for years answer these scales much the way formally diagnosed adults do — the 2024 study of 839 people found few psychometric differences between the groups.3 Your reading of yourself is data, and it was data before the screener existed.
What about the RAADS-R, the 80-question one?
The RAADS-14 you have just taken was cut down from it. The RAADS-R has 80 questions, a total running to 240 and a cut-off of 65, and it is the version most self-assessment sites push. Its reputation rests on the original 2011 study, which reported 97 per cent sensitivity and 100 per cent specificity across nine centres. Those are extraordinary numbers, and they are real — for the way the test was given in that study, with a clinician sitting in the room.
Handed to people to fill in alone, it behaves very differently. The one independent test of it in a working NHS autism assessment service looked at 50 adults referred for assessment:
49 of 50 scored above the cut-off. 3% specificity. 0.45 area under the curve — below chance. 34% went on to a diagnosis.
A test that flags 98 per cent of the people who take it has not told them anything. The authors concluded it “lacks predictive validity and is not a suitable screening tool for adults awaiting autism assessments.”
That is why this course uses the 14-item version and not the 80-item one. Twenty minutes is a lot to ask for a result that cannot separate you from almost anyone else who was worried enough to take it. If a service asks you to complete the RAADS-R as part of an assessment, do it — in that setting, with a clinician, it is doing the job it was designed for.
→ When you have met all three general screeners, there is a table comparing them at the end of Module 4.
→ If ADHD has also been raised — or somebody has told you it is one or the other — the other course opens with the six questions the World Health Organization settled on: ADHD Self-Discovery, Module 2.
F. What helps
A screen is a doorway, and a doorway is only good for walking through. What stops people is rarely the score. When six self-identifying autistic women and gender-diverse adults were asked why they had not sought an assessment, the reasons were waiting lists, private costs, GPs acting as gatekeepers, and a well-founded fear of not being believed.12 None of those is a reflection on you.
1. Take the number to a person, not to a forum.
The RAADS-14 exists to start a conversation with someone who can take a developmental history — a GP, a psychologist, an assessment service. Bring the total and the date. Expect to talk about childhood, school reports and working life, because that is what an assessment rests on.
2. Write down the three things the fourteen questions never asked.
Write one paragraph on the interest that has absorbed you longest, one on what you do with your hands and body when content or overloaded, and one on how you actually use words. Those are real criteria and this short form drops them. An assessor will ask; your score will not have covered it.
3. Read a low score with as much care as a high one.
Nine of the 135 autistic adults in the original study scored below the cut-off, and two scored zero. In interviews with 24 autistic adults, most of them women or non-binary, participants described learning to mask from early childhood and being given mood or personality diagnoses for years before anyone asked about autism.13 A score under 14 lowers the odds. It does not close the question.
4. Name the other things that could be producing this number.
More than half of the non-autistic adults with ADHD in the development sample scored at or above 14. Social anxiety, depression, trauma and burnout push these items the same way. Write down what else has been true of the last two years — an honest assessment does that anyway, and it goes faster if you have started.
5. Keep your answers, not just your total.
Note which questions you answered 3 to, and write one real example beside each: the specific fabric, the specific conversation, the specific afternoon. Two sentences of concrete life beat a number in a first appointment, and they are much harder to dismiss.
Step 2 — Take the screener
Fourteen questions, three to five minutes, free and confidential. You get one number between 0 and 42 and a band that tells you which side of the threshold you landed on. Read section C before you read your band: the cut-off of 14 was set deliberately low so that the screen would miss almost nobody, which means a score over the line is a reason to look further rather than a finding about you.
Before you start
The RAADS-14 Screen is a published, peer-reviewed short form of the 80-item RAADS-R, developed in 2013 and validated in a Swedish adult psychiatric sample of 1,233 people. Its sensitivity is high, at 97 per cent for the cut-off of 14, but its specificity in that same study was only 46 to 64 per cent among psychiatric patients, and the one independent validation found 56.4 per cent specificity at the original threshold and recommended a cut-off of 22 instead. A score at or above 14 is therefore common among people who are not autistic, particularly people with ADHD, social anxiety or depression. It is a screening questionnaire, not an assessment, and no number it produces settles anything. Only a clinician can diagnose.
Step 3 — Your workbook
Your workbook saves to this device only — we cannot see a word of what you write. This module's workbook records your total against the live bands, then gathers the three things the fourteen questions never asked about, the alternatives worth naming, and the two sentences you would open with when you take it to someone.
Your RAADS-14 results
Took the screener? Put your total in below. Section D is the one to read alongside it - the threshold of 14 was set deliberately low. Entirely optional - skip it if you would rather just read.
Score bands: 0–13 = Below screening threshold; 14–42 = At or above the threshold
Fields: RAADS-14 · 14 items, scored 0 to 3, three to five minutes; Total score (0–42); My total (enter 0–42); Date I took it
The three things the fourteen never asked
Section F, item 2. Special interests, repetitive movement and language are real criteria, and this short form drops almost all of them. Write them out before any appointment.
Fields: The interest that has absorbed me longest, and what it looks like from outside; What I do with my hands and body when I am content, and when I am overloaded; How I actually use words - scripts, precision, literal readings, going quiet
What else could be producing this number
Section F, item 4. More than half of non-autistic adults with ADHD scored at or above 14 in the original study. Naming the alternatives is what a good assessment does anyway.
Fields: Tick: Attention and focus have been a lifelong problem too; Tick: Social situations frighten me more than they confuse me; Tick: The last two years have included depression, trauma or burnout; What else has been true of the last two years
Your threes, with a real example beside each
Section F, item 5. The item-level answers carry more than the total. Two sentences of concrete life beat a number in a first appointment.
Fields: The item I answered most strongly, and the specific afternoon it comes from; A second one, with its example; Which of the three groupings felt most like me: Mentalizing deficits - reading people / Social anxiety - groups and small talk / Sensory reactivity - texture and noise / Two of them equally / None of them quite
If your score came in low
Section D. Nine of the 135 autistic adults in the original study scored below the cut-off, and two of them scored zero. A low total lowers the odds; it does not close the question.
Fields: Did the questions feel like they were describing you, whatever the number said: Yes, closely / In parts / Not really / I could not tell; Where I think I answered the way I have learned to answer, rather than the way it is
Taking it to a person
Section F, item 1. A screen is a doorway and the only thing a doorway is good for is walking through.
Fields: Who I would take this to first; The two sentences I would open with; What is actually in the way: Cost / Waiting lists / Not being believed / Not knowing where to start / Nothing, I just have not done it
Appendix — Research companion
Peer-reviewed research
4. Baghdadli A, Russet F, Mottron L (2017). Measurement properties of screening and diagnostic tools for autism spectrum adults of mean normal intelligence: a systematic review. European Psychiatry, 44, 104-124. DOI 10.1016/j.eurpsy.2017.04.009. View the paper Systematic review of 38 articles covering 32 studies, examining nine instruments for autistic adults of mean normal intelligence: three diagnostic tools and six screening tools with eight short versions. Only the AQ-50, AQ-S, RAADS-R and RAADS-14 showed satisfactory or intermediate psychometric values supported by strong or moderate evidence, and the authors conclude that evidence for measurement properties in this population is limited overall and recommend using self-report alongside clinical expertise rather than in place of it. Limitation: the review found risks of bias and applicability concerns across the included studies that constrain what can be concluded about diagnostic properties, and it predates the independent RAADS-14 translation work.
5. Boilson M, Roarty A, Stanfield A, Berney T, Lenihan D, Doherty M, O'Dea S (2026). Autism and autism features in a young adult ADHD population, gender differences and use of the RAADS-14. Irish Journal of Psychological Medicine, advance online publication, 1-9. DOI 10.1017/ipm.2026.10213. View the paper Comparative cohort study of young adults with ADHD transitioning from child and adolescent mental health services into a specialist adult ADHD service, all assessed clinically and with the RAADS-14. Co-occurring autism was found in 53% of the cohort, high levels of autistic features were present in the ADHD-only group, and RAADS-14 subscale scores for social anxiety, mentalizing and sensory sensitivities differed significantly between the autistic and non-autistic ADHD groups. Autistic women scored higher than autistic men on every domain, and sensory sensitivity scores were significantly higher in women in both groups regardless of autistic status. Limitation: a single transitioning ADHD service cohort with no non-clinical comparison group and no reported sensitivity, specificity or predictive values, so it describes score patterns rather than screening accuracy.
2. Eriksson JM, Andersen LMJ, Bejerot S (2013). RAADS-14 Screen: validity of a screening tool for autism spectrum disorder in an adult psychiatric population. Molecular Autism, 4, 49. DOI 10.1186/2040-2392-4-49. View the paper The development and validation study for the RAADS-14, using 1,233 adults in total - 643 with a psychiatric diagnosis and 590 non-psychiatric controls - with 135 autistic adults and 508 psychiatric controls completing the abridged scale. A cut-off of 14 or above out of 42 gave sensitivity 0.97 with specificity 0.46 in ADHD, 0.64 in other psychiatric disorders and 0.95 in non-psychiatric controls; AUC was 0.88, 0.91 and 0.99 respectively; median totals were 32 (ASD), 15 (ADHD) and 11 (other psychiatric); full-scale alpha was 0.90 and factor analysis yielded mentalizing deficits, social anxiety and sensory reactivity. Nine of the 135 autistic participants scored below the cut-off, two of them scoring zero. Limitation: a single Swedish sample diagnosed under DSM-IV-TR with no independent confirmatory assessment, no matching for sex, age or intelligence, 10% of participants not stating their gender, and no test-retest reliability reported.
9. Jones SL, Johnson M, Alty B, Adamou M (2021). The effectiveness of RAADS-R as a screening tool for adult ASD populations. Autism Research and Treatment, 2021, 9974791. DOI 10.1155/2021/9974791. View the paper Fifty adults referred to the Adult ADHD and Autism Service at South West Yorkshire Partnership NHS Foundation Trust completed the RAADS-R before assessment; mean age 32.8, 70% male. Using the published cut-off above 65, 49 of the 50 screened positive while only 17 went on to receive an autism spectrum diagnosis, giving sensitivity of 100%, specificity of 3.03% and positive predictive value of 34.7%. The authors conclude that the RAADS-R lacks predictive validity and is not a suitable screening tool for adults awaiting autism assessment. Limitation: a small single-service sample of 50 already-referred adults, administered as unsupervised self-report rather than with a clinician present as in the original validation, so the collapse in specificity is partly a base-rate effect of screening people who have all already been referred.
6. Nyrenius J, Eberhard J, Ghaziuddin M, Gillberg C, Billstedt E (2022). Prevalence of autism spectrum disorders in adult outpatient psychiatry. Journal of Autism and Developmental Disorders, 52(9), 3769-3779. DOI 10.1007/s10803-021-05411-z. View the paper Consecutive new patients at an adult psychiatric outpatient clinic in Sweden between November 2019 and October 2020 were screened with the RAADS-14: 562 were approached and 304 responded, of whom 197 scored at or above the cut-off of 14. Forty-eight of those screen-positive patients completed an in-depth diagnostic assessment and 26 met criteria for an autism spectrum condition, with a further eight showing subthreshold symptoms, giving the authors a prevalence estimate of 18.9%. This is the clearest real-world picture of how a RAADS-14 cut-off of 14 behaves in an unselected psychiatric intake. Limitation: only 54% of those approached responded and only a quarter of screen-positives were assessed, so the prevalence estimate is built on a heavily self-selected subsample and was formally contested in the same journal.
1. Ritvo RA, Ritvo ER, Guthrie D, Ritvo MJ, Hufnagel DH, McMahon W, Tonge B, Mataix-Cols D, Jassi A, Attwood T, Eloff J (2011). The Ritvo Autism Asperger Diagnostic Scale-Revised (RAADS-R): a scale to assist the diagnosis of autism spectrum disorder in adults: an international validation study. Journal of Autism and Developmental Disorders, 41(8), 1076-1089. DOI 10.1007/s10803-010-1133-5. View the paper International validation of the 80-item parent scale across nine research centres on three continents, comparing 201 autistic adults with 578 non-autistic comparison participants. At a cut-off of 65 it reported 97% sensitivity and 100% specificity, test-retest reliability of r = 0.987, and four subscales: Social Relatedness (alpha 0.923), Circumscribed Interests (0.903), Language (0.789) and Sensory Motor (0.905). Limitation: a clinician remained with every participant while they completed the scale, so the reported accuracy does not describe unsupervised self-administration, and the comparison group was recruited at the same specialist centres rather than sampled from the population.
8. Skorokhodov IV, Radygina KP, Skorokhodova EY, Firsova SP, Portnova GV, Varlamov AA (2023). Development and initial validation of the Russian version of the RAADS-14: a self-report questionnaire to assess autistic traits. European Journal of Investigation in Health, Psychology and Education, 13(11), 188. DOI 10.3390/ejihpe13110188. View the paper The only independent translation validation of the RAADS-14 reporting full accuracy statistics: 1,724 respondents, 1,667 analysed, including 794 non-psychiatric young adults and 49 adults with a clinically established autism spectrum diagnosis, 13 of them ADOS-2 verified. Cronbach's alpha was 0.839, confirmatory factor analysis supported the original three factors (RMSEA 0.067, CFI 0.954), and AUC was 0.92; but the original cut-off of 14 gave sensitivity of 93.9% with specificity of only 56.4%, leading the authors to recommend a cut-off of 22 (sensitivity 85.7%, specificity 85.4%). The non-psychiatric mean total was 13.9 (SD 7.5), a point below the original threshold. Limitation: only 49 autistic participants, of whom just 13 had ADOS-2 verification, in a country with very high adult underdiagnosis and no validated local reference instrument.
3. Sturm A, Huang S, Bal V, Schwartzman B (2024). Psychometric exploration of the RAADS-R with autistic adults: implications for research and clinical practice. Autism, 28(9), 2334-2345. DOI 10.1177/13623613241228329. View the paper Item-response analysis of the RAADS-R and RAADS-14 in 839 adults across four groups: formally diagnosed autistic adults, self-identifying autistic adults without a diagnosis, adults exploring the question, and neurotypical comparisons. Both scales functioned as unidimensional instruments with no systematic bias by age, gender, or diagnosis versus identity; four items concerning adverse sensory consequences and the hidden rulebook of social interaction were the most discriminating; and few psychometric differences appeared between diagnosed and self-identifying participants. The authors recommend collapsing the four response options to two. Limitation: a convenience sample recruited online with self-reported diagnostic status and no independent diagnostic verification, so it tests measurement structure rather than diagnostic accuracy and reports no sensitivity or specificity.
Clinical frameworks and position statements
11. National Institute for Health and Care Excellence (2021). 2021 surveillance of autism (NICE guidelines CG128, CG142 and CG170). NICE surveillance report, London. View the source NICE surveillance review of its three autism guidelines, including the adult guideline CG142. It records stakeholder confirmation that, consistent with the existing recommendation, a case-identification tool such as the AQ-10 would not be used alone to screen for autism, and it notes evidence of a 64% false negative rate for the AQ-10 in adults with suspected autism, along with a correction to the published cut-off. Included here as the guideline position that short self-report screens support referral decisions rather than replace assessment. Limitation: a surveillance decision document about whether to update guidance, not a primary accuracy study, and it does not assess the RAADS-14, which is not named in the guideline.
10. Royal College of Psychiatrists (2020). The psychiatric management of autism in adults (College Report CR228). Royal College of Psychiatrists, College Report CR228. View the source UK professional-body guidance on identifying and supporting autistic adults in psychiatric services. It states that the use of a single structured questionnaire is unlikely to be sufficiently reliable either to identify adults who are not autistic or to differentiate autism from mental health conditions, and that a definitive interview capable of refuting as well as confirming the diagnosis, and of identifying co-occurring conditions, can take several hours and involve more than one professional. The RAADS-R and the AQ are both listed among questionnaires in its instruments appendix. Limitation: consensus guidance from a national college rather than a systematic review, UK-specific in its service assumptions, and it does not evaluate the RAADS-14 short form separately.
Lived experience
12. Friedman A, Paltoglou A, Sorte R (2024). A qualitative exploration of the experiences of self-diagnosed autistic women and gender-diverse individuals who are not pursuing an autism diagnosis. Neurodiversity, 2. DOI 10.1177/27546330241307828. View the source Interpretative phenomenological analysis of semi-structured interviews with six self-identifying autistic adults aged 18 to 69, four women, one non-binary and one genderfluid participant, conducted by an autistic researcher. Three themes emerged - autistic self-discovery, living without a diagnosis, and self-doubt and self-diagnosis - with participants describing arriving at self-identification through their own research, online material and autistic community contact rather than through any formal instrument. Reported barriers to seeking assessment included long waiting lists, private assessment costs, fear of not being believed, GPs acting as gatekeepers, and clinician bias against diagnosing women who mask well. Limitation: only six participants, all verbally fluent and recruited online, so it describes a specific self-identifying group and cannot speak for autistic adults who need direct support.
13. Tien I, Wolpe S, Pearson A, Seers K (2025). Creating a socially acceptable version of myself: a qualitative thematic analysis of the female and nonbinary experience of navigating the autism diagnostic system. Neurodiversity, 3. DOI 10.1177/27546330241306380. View the source Reflexive thematic analysis of semi-structured interviews with 24 autistic adults - 14 non-binary, agender or genderfluid, eight female and two male, of whom 16 were formally diagnosed and eight self-diagnosed. Participants described learning to mask autistic traits from early childhood, being misdiagnosed with mood or personality conditions before autism was considered, distrusting mental health services afterwards, and facing cost and access barriers to formal assessment; gender-diverse participants also reported clinicians unfamiliar with the intersection of autism and gender identity. Limitation: a predominantly white (79%) and highly verbal sample recruited for recorded interviews, which excludes less verbal autistic adults and limits generalisation.
Emerging or contested
7. Bejerot S, Nylander L (2022). An overestimation of the prevalence of ASD among psychiatric patients. Journal of Autism and Developmental Disorders, 52(9), 4204-4205. DOI 10.1007/s10803-022-05568-1. View the paper Letter to the editor by the senior author of the original RAADS-14 validation, with Lena Nylander, criticising the prevalence estimate drawn from the Swedish outpatient screening study. Recalculating from the 26 confirmed cases across the whole screened group of 304 responders gives a minimum prevalence of no more than 8.6%, and the letter states that the specificity of the RAADS is low at 67% and that false positives are likely a pervasive issue in that study, alongside criticism of the very low participation rate at every sampling phase. Limitation: a two-page commentary with no new data, and the 67% specificity figure it cites is drawn from a different study rather than measured in the cohort under discussion.
Peer-reviewed = checked by independent experts before publication. Clinical model = an established professional framework, not a single study.
Up next
Module 3 - AQ-50 — The Best-Known Autistic Traits Questionnaire
All modules in Autistic Self-Discovery
A number over the line is a good reason to talk. This course was built by clinicians who are part of the New Path Family. A RAADS-14 score is a starting point rather than an answer, and working out what sits underneath it — autistic traits, ADHD, social anxiety, years of masking, or some combination — is the kind of thing that goes faster with someone alongside you. Therapy for clients in California and coaching worldwide, all by telehealth, are offered by our sister company New Path Family of Therapy Centers, Inc. A conversation costs nothing and there is no pressure. Saving this for later counts too. Talk with the New Path team