Sign Up
Login
Try for Free
Login
The CARS-2 (Childhood Autism Rating Scale, Second Edition) is a clinician-completed rating scale that quantifies autism symptom severity from direct observation, record review, and collateral information. Psychologists and other trained clinicians use it as one severity input within a broader autism evaluation, never as a standalone diagnosis. This page covers how to write up CARS-2 results, with a fictional sample and a results-section template.
Psychologists, physicians, school psychologists, and other trained clinicians; publisher qualification level C
Autism evaluation teams, physicians, IEP and early-intervention teams, payers naming CARS-2 such as Nevada Medicaid and TRICARE, parents
300 to 550 words for the CARS-2 section · rating 5 to 10 minutes after observation and record review
Clinician-rated autism severity rating scale
Autism diagnostic evaluations, ages 2 and up; ABA authorization, school, and support-service requests where a severity rating supports the picture
Published by Western Psychological Services (2010); current edition as of July 2026; described here for write-up purposes, no test content reproduced
The Childhood Autism Rating Scale, Second Edition (CARS-2) is a clinician-completed rating scale that quantifies autism symptom severity from direct behavioral observation combined with record review and collateral information. Western Psychological Services published it in 2010, and it is the second edition of the original CARS, one of the most widely used and studied autism rating scales. It has three forms: the Standard Version (CARS2-ST), for individuals younger than 6 and for older individuals with communication difficulties or below-average estimated IQ; the High-Functioning Version (CARS2-HF), for verbally fluent individuals aged 6 and older with estimated IQ above 80; and an unscored Questionnaire for Parents or Caregivers (CARS2-QPC) that gathers developmental information to inform the clinician's rating. Each scored form has 15 items covering functional areas, rated by the clinician on a graded response scale, and takes about 5 to 10 minutes to complete once the observation and background information are in hand. It spans ages 2 and up. As of July 2026 the 2010 edition remains current, with no CARS-3 announced.
Beneath the total score sits the fact most competing pages get wrong: the CARS-2 standard scores and percentiles are referenced to a clinical sample of 1,034 individuals with autism spectrum disorders, not a nationally representative population sample. A percentile therefore expresses where a person falls relative to other diagnosed or referred individuals, not relative to the general population, and the write-up should phrase it exactly that way. The second load-bearing point is scope: the CARS-2 is a severity rating, not a standalone diagnosis, and the authors and publisher position it as informing a diagnosis rather than making one. Only the clinician completes the ST or HF rating booklet; parents and caregivers complete the QPC, never the scored forms. A defensible autism diagnosis under DSM-5-TR integrates the CARS-2 with a direct observation, often the ADOS-2, a developmental-history interview such as the ADI-R, an informant trait scale such as the SRS-2, and cognitive and adaptive testing, all feeding a whole evaluation. This page reproduces none of the test's items, rating anchors, or cutoff values.
Autism diagnostic clinics, developmental-behavioral pediatricians, child and school psychologists, speech-language pathologists, occupational therapists, and early-intervention teams use the CARS-2 as the clinician-rated severity component of an autism evaluation, choosing the Standard or High-Functioning form by the person's age, verbal fluency, and estimated cognition. Its short rating time and modest training burden make it usable across a range of professions and settings, including places where access to an ADOS-2-trained specialist is limited. It rarely stands alone: convention pairs it with a direct observation, a developmental-history interview such as the ADI-R, an informant trait scale such as the SRS-2 or a behavior rating scale, and cognitive and adaptive measures such as the Vineland-3. The whole set anchors a psychological evaluation report. One boundary belongs in the report itself: the commercial CARS-2 was not validated for use without live, in-person observation, so a remotely gathered rating belongs in a limitations line, not a silent footnote.
No statute, payer, or publisher mandates a format for the CARS-2 section of an autism evaluation. The sequence below is the convention experienced evaluators converge on because it survives review, and because it is built around the rules the top search results routinely break: report the severity level in your own words and keep the item ratings, rating anchors, and band cutoff values out, and state plainly that a CARS-2 percentile is referenced to a clinical sample of individuals with autism, not the general population. Leaving the protected content out is a test-security obligation under APA Ethics Standard 9.11 and WPS policy, not an oversight. Each section below carries the pitfall that most often undermines it.
Instrument, form, and administration. Name the CARS-2 and its 2010 edition, the form used (Standard or High-Functioning) and why it was chosen (age, verbal fluency, and estimated IQ), who completed the rating and their training, and that it was one component of the evaluation. Pitfall: naming the CARS-2 with no form, or writing it up as though the rating alone produced the diagnosis.
Sources and basis of the rating. State who observed the individual, over what period and in what setting, and which collateral informed the rating: records, the caregiver questionnaire, and teacher or parent report. Pitfall: a severity number with no stated basis, so a reader cannot tell whether it rests on a brief clinic look or a rich multi-source picture.
Severity result, reported in words. State the severity level in plain language tied to observed behavior and support needs. If you cite the standard score or percentile, say plainly that it is referenced to a clinical sample of individuals with autism rather than the general population. Pitfall: reporting a percentile as if it were population-based, or printing the item ratings, rating anchors, or band cutoff values.
Integration with observation and history. Place the CARS-2 next to the direct observation, developmental history, and any rating scales, and say where the sources agree and where they diverge. Pitfall: when the CARS-2 and an ADOS-2 disagree, averaging them or picking one; reconcile the divergence in words, for example a masking child who presents below threshold on office observation but shows a clear developmental history.
Diagnostic formulation under DSM-5-TR. State the diagnostic determination and the clinical judgment behind it, with the CARS-2 named as one severity input among several. Pitfall: a diagnosis that rests on the CARS-2 alone. The publisher and authors position the scale as informing a diagnosis, not making one.
Limitations and validity. Note anything that bears on interpretation: a form choice made at the age or IQ boundary, rater training and the scale's inherent subjectivity, single-setting sampling, masking, in-person versus remote gathering, the thinner evidence base for the High-Functioning form, and limited validation in adults and in women and girls. Pitfall: silence on limitations, especially for high-masking presentations where the rating can under-detect autism.
Recommendations linkage. Tie each recommendation, whether supports, services, or further assessment, to a specific finding. Pitfall: recommendations that could follow any evaluation and trace back to nothing in the rating.
CARS-2 RESULTS SECTION Instrument/edition: CARS-2 (2010) Form: [Standard (CARS2-ST) / High-Functioning (CARS2-HF)] Form selected because: [age + verbal fluency + estimated IQ] Rated by: [name, credentials, training] Date: [ ] BASIS OF THE RATING [Who observed, over what period and setting; collateral used: records, caregiver questionnaire (CARS2-QPC), teacher/parent report] SEVERITY RESULT (narrative, in your own words) [Severity level tied to observed behavior and support needs. If a percentile is cited, state it is referenced to the scale's clinical sample of individuals with autism, not the general population. No item ratings, anchors, or band cutoff values.] INTEGRATION WITH OBSERVATION AND HISTORY [Direct observation, developmental history, rating scales; state where they converge or diverge, and reconcile divergence rather than averaging it] DSM-5-TR FORMULATION [The diagnostic determination and the clinical judgment behind it; the CARS-2 is one severity input, not the diagnosis] LIMITATIONS AND VALIDITY [Form choice at the boundary, rater subjectivity, single setting, masking, in-person vs remote, thinner HF evidence] RECOMMENDATIONS LINKAGE [Each recommendation tied to a specific finding] Evaluator signature / credentials: Date:
Free to use and share, no signup. The PDF includes a one-page cheat sheet with section-by-section pitfalls and a pre-sign checklist; the DOCX is the blank results-section skeleton, ready to adapt.
Scenario: a verbally fluent 10-year-old is referred by their school team for social difficulty and rigidity, with average estimated cognition and a developmental history of subtle but persistent social-communication differences. The High-Functioning form was rated as the clinician-severity component of a broader autism evaluation. This is the CARS-2 section only, and it deliberately reports the severity level and narrative behavior without item ratings, anchors, or band cutoff values, which are protected. All details are fictional.
Client: M.R., 10 · Referral: school team, social difficulty and rigidity · Evaluator: L. Okafor, PhD, Licensed Psychologist · Rating date: 07/13/2026 · Report date: 07/17/2026
Instrument, form, and administration: The Childhood Autism Rating Scale, Second Edition (CARS-2) was completed as one component of an autism evaluation. The High-Functioning form (CARS2-HF) was selected because M.R. is a verbally fluent 10-year-old with average estimated cognition, the population the HF form was designed for. The rating was made by the examining psychologist from a clinic observation, record review, and the completed Questionnaire for Parents or Caregivers.
Basis of the rating: Ratings drew on a 60-minute clinic session, a classroom report, and the caregiver questionnaire. M.R. was cooperative and articulate. Conversation ran largely one way, with limited reciprocal give-and-take, an intense and detailed interest that recurred across topics, discomfort with unexpected changes to routine, and reported sensitivity to certain sounds.
Severity result: The CARS2-HF placed M.R. in the moderate range of autism-related behaviors. The converted percentile fell near the middle of the distribution relative to the scale's clinical sample of individuals with autism, not the general population, so it describes where M.R. sits among referred and diagnosed peers rather than among all same-age children. Consistent with test-security convention, item ratings, anchors, and band cutoff values are not reproduced here.
Integration with observation and history: The rating converges with the developmental and informant data. The caregiver history documents early and persistent difficulty sustaining friendships, and a completed social-communication rating scale fell in the elevated range across home and school. The observation-based severity is best read as a floor rather than a ceiling: subtle, well-compensated presentations can produce lower office-based ratings that must be weighed against pervasive real-world impairment.
DSM-5-TR formulation: Integrating the CARS-2 severity rating with the developmental history, the informant rating scale, and clinical judgment, M.R. meets DSM-5-TR criteria for autism spectrum disorder, requiring support (Level 1). The CARS-2 is one input to this determination, not the determination itself.
Limitations: The rating reflects behavior sampled largely in a single clinic session and rests on clinician judgment, which the scale shares with other autism measures. The High-Functioning form has a thinner published evidence base than the Standard form, and subtle presentations in cognitively able children can be under-detected, so the finding is read alongside history and collateral report rather than in isolation. No remote administration was involved.
Recommendations linkage: The reciprocal-communication findings support recommendation 1 (social-communication support at school with structured peer opportunities). The rigidity and sensory findings support recommendation 2 (classroom predictability and sensory accommodations). The diagnosis supports recommendation 3 (eligibility and service discussions by the school and payer teams applying their own criteria).
This sample is fictional and for educational purposes. It does not describe a real person or record, and the severity level, percentile, and details are invented for illustration and correspond to no real child or record.
Writing these after every session? BastionGPT drafts complete notes from bullets, dictation, or a transcript.
Generate a note from bulletsWrite the CARS-2 section knowing which decision it will feed, and label the strength of each requirement honestly. No jurisdiction legally requires the CARS-2, and no authority makes a CARS-2 score alone sufficient for a diagnosis, special-education eligibility, or insurance authorization. DSM-5-TR governs the diagnosis, and the AACAP practice parameter (Volkmar and colleagues, 2014) frames autism diagnosis as a comprehensive, multidisciplinary process rather than any single instrument. What names the CARS-2 is payer or program policy. In the United States, Nevada Medicaid's ASD diagnosis certification form FA-11F (updated June 2025) lists the CARS-2 alongside the ADOS-2 and GARS-3 as a diagnostic tool whose score a provider may enter, and North Carolina's 2026 draft Clinical Coverage Policy 8F proposed adding the CARS-2 standard and high-functioning forms to its accepted list, though that draft closed public comment in June 2026 and is not final. Aetna's clinical policy bulletin and TRICARE's Autism Care Demonstration both name the CARS-2 among diagnostic tools used with clinical assessment, all of which is PAYER POLICY, not a clinical mandate. Educational eligibility is separate again: under IDEA a school does not require a medical diagnosis or any specific instrument, and 34 CFR 300.304(b)(2) bars any single measure as the sole criterion, which is LAW. In Canada, the Canadian Paediatric Society (2019) treats CARS-2 findings as one contributor that cannot be used alone to diagnose autism (CONVENTION), and in Australia the Autism CRC National Guideline (2018, updated 2023 and NHMRC-approved) recommends standardized assessment integrated by a skilled team without mandating a specific tool (POLICY and CONVENTION). Testing and evaluation time is billed under the psychological and developmental testing code families by bare code number, for example 96112, 96113, 96130, 96131, 96132, 96133, 96136, and 96137 in the United States, as payer policy with plan-specific rules. The format is a convention; the severity rating, the narrative, and the integration are the requirement.
Edition currency and the qualification boundary are the CARS-2's live defensibility questions. As of July 2026 the 2010 edition remains current, with no CARS-3 announced; this is worth re-checking each year, because publishers revise instruments without long notice. Two facts about the norms belong in the write-up. First, the standard scores and percentiles rest on a clinical sample of 1,034 individuals with autism assembled for the 2010 edition, which predates DSM-5 and is not a representative population sample, so a percentile is standing among referred individuals, not a population percentile. Second, who may use the scale: the CARS-2 is qualification level C, and WPS qualifies people to purchase the test, not to use it, which is set by licensing boards and scope of practice. The High-Functioning form is newer and less validated than the Standard form, its specificity drops against non-autism clinical comparison groups, and validation in adults and in women and girls is limited, so reports leaning on the HF form should say so. The commercial CARS-2 is also not validated for administration without live, in-person observation; research adaptations for remote use exist but are not the published instrument, so a remote rating is a stated limitation. CARS and CARS-2 are trademarks of Western Psychological Services (WPS). BastionGPT is not affiliated with, or endorsed by, the publisher. This page reproduces no test items, stimuli, norms, or scoring materials.
There is no payer audit series for autism severity write-ups; the accountability here is psychometric and ethical, and it is pointed. The CARS-2 was reviewed in The Nineteenth Mental Measurements Yearbook (2014, test review by K. K. Malcolm), the correct Buros volume to cite. The peer-reviewed evidence is usable but bounded: a 2019 systematic review and meta-analysis of 24 studies and 4,433 participants found the CARS internally consistent and its sensitivity acceptable, but its specificity was not, and the authors concluded it should be used alongside other confirmatory tools. The errors below are write-up errors, not administration errors. The BastionGPT Clinical Advisory Board sees the same ones most often in CARS-2 report reviews:
BastionGPT is specifically trained, tuned, and clinically tested on psychological and psychoeducational evaluation reports.
See how clinicians use it day to day on the AI therapy notes page.
Many BastionGPT users report saving more than 90 minutes per day on documentation.
HIPAA-compliant with a signed BAA on every plan. Your data is never used to train models. BastionGPT drafts, you review and sign.
The CARS-2 is a clinician-rated severity scale, not a diagnostic test. The examining clinician rates functional areas from direct observation, record review, and collateral information, and the ratings sum to a total that converts to a standard score and percentile. The essential thing to understand, and to write down, is that those conversions are referenced to a clinical sample of individuals already diagnosed with autism, not the general population, so a percentile describes where a person falls among referred and diagnosed peers, not among all same-age children. Report the severity level in words tied to observed behavior, and keep the item ratings and cutoff values out of the document. A CARS-2 result is one input a clinician integrates into a whole evaluation under DSM-5-TR.
The manual's rule is age, verbal fluency, and estimated cognition. The Standard form (CARS2-ST) fits individuals younger than 6, and older individuals with communication difficulties or below-average estimated IQ; the High-Functioning form (CARS2-HF) fits verbally fluent individuals aged 6 and older with estimated IQ above 80. The two forms are scored separately and are not interchangeable. At the margin, a child who is 6 with emerging language, or whose estimated IQ sits near the boundary between the forms, name the judgment in the report and note that the form chosen can shift the severity classification. No authority prescribes a tie-breaker, so document the reasoning.
No law in the United States, Canada, or Australia requires the CARS-2, and no authority makes a CARS-2 score alone sufficient for a diagnosis or for eligibility. DSM-5-TR governs the diagnosis. What names the CARS-2 is payer or program policy: Nevada Medicaid's diagnosis form lists it as a diagnostic tool whose score a provider may enter, a 2026 North Carolina Medicaid draft proposed adding both CARS-2 forms to its accepted list, and Aetna and TRICARE name it among tools used with clinical assessment. Educational eligibility is separate, since under IDEA a school requires no specific test. The honest sentence for a report is that the CARS-2 is convention and, for some payers, policy, not a legal mandate. It pairs naturally with the ADOS-2.
No, for two reasons. First, a single historical cut score, such as the 30 that circulated for decades in DSM-IV-era research on the standard form, was an identification threshold, not a diagnosis, and it does not transfer to the High-Functioning form, which is scored separately. Second, the CARS-2 is a severity rating that informs a diagnosis rather than making one. A 2019 meta-analysis found the scale's sensitivity acceptable but its specificity not, and concluded it should be used alongside other confirmatory tools. Report the severity in words and integrate it with a direct observation and developmental history.
They do different jobs and are strongest together. The CARS-2 is a brief clinician severity rating built from observation and collateral; the ADOS-2 is a semi-structured direct observation and is often the payer or research benchmark; the ADI-R is a structured caregiver interview about developmental history; and the SRS-2 is a quantitative trait scale that helps flag masking when observation sits near the threshold. The CARS-2 is faster and cheaper and usable by a wider range of professionals, but best practice uses it as a supplement, not a substitute, for a structured observation plus history. Each is an input; none is the diagnosis.
No. The rating anchors, item content, and band cutoff values are protected test materials: psychologists are ethically obligated to maintain test security under APA Ethics Standard 9.11, and WPS enforces the same through its purchase agreements. Reproducing the anchors or band tables also degrades the norms, which assume the person has not been coached on them. A report carries the severity level, your narrative description of behavior, and the interpretation, never the item ratings or cutoff values. Raw test data, as distinct from test materials, are releasable only under Standard 9.04 with an appropriate release, and still route through your test-security obligations.
Not with the published instrument. The commercial CARS-2 was not designed or validated for use without live, in-person observation, so a remotely gathered rating should be described as a limitation rather than reported as a standard result. Research adaptations exist: Sanchez and Constantino validated an observation-only CARS-2 protocol in person, and a remote version has been piloted against the in-person ADOS-2 in a small sample, with promising but preliminary agreement. Those are research adaptations, not the published scale. When a remote observation is unavoidable, name the method and its limits in the report.
As of July 2026 the CARS-2, published in 2010, is the current edition, and no CARS-3 has been announced by WPS or found in any credible source. Because the standard scores rest on a clinical sample assembled for the 2010 edition, which predates DSM-5, interpret percentiles as standing among referred individuals and re-check for an edition change periodically. Test on the current edition, name it and the year in the report, and treat any future edition change as a stated limitation.
Yes. Paste a summary of the form you used, the severity level, and your behavioral observations, and it drafts the CARS-2 section for your review: the severity in narrative form, the percentile framed against the clinical sample rather than the population, and the integration with observation and history laid out so you can confirm it. It can also cross-check a draft you wrote for the gaps reviewers flag, a percentile written as population-based, a severity level treated as a diagnosis, a standard-form cut applied to the High-Functioning form, or protected content that slipped in, and produce a plain-language summary for parents and referrers. BastionGPT is HIPAA-compliant with a signed BAA on every plan, your data is never used to train models, and because you paste only your summary and observations, no item content leaves your records.
The instrument facts and compliance claims on this page trace to these sources, last verified July 2026:
Educational content, not legal or billing advice. Sample notes are fictional. Follow your organization's policies and your board, payer, and jurisdiction requirements.