Adult Autism Test Methodology

The Adult Autism Test is a screening implementation of the RAADS-14 Screen, a 14-item self-report instrument validated by Eriksson, Andersen & Bejerot (2013) in Molecular Autism and released under Creative Commons Attribution 2.0. The tool produces three sub-dimensional scores, a single composite total, a four-band severity classification, and a five-archetype profile that maps the user’s response pattern across the instrument’s factor structure.

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized in DSM-5 by persistent deficits in social communication and social interaction across multiple contexts (Criterion A) and restricted, repetitive patterns of behavior, interests, or activities (Criterion B), with onset in early developmental period (Criterion C) and clinically significant functional impairment (Criterion D). The RAADS-14 captures features relevant to Criteria A and B but does not directly assess developmental onset or functional impairment, both of which require clinical interview and history-taking. This is one of several reasons the tool is a screen and not a diagnosis.

Why we chose the RAADS-14

A range of validated adult autism screens exist. Each makes different trade-offs between sensitivity, specificity, brevity, and clinical context. The table below summarizes the alternatives we considered and why the RAADS-14 was selected.

The RAADS-14 was selected because it is the only validated, brief adult autism screen released under a license that permits free commercial deployment with attribution — Creative Commons Attribution 2.0. It also uniquely targets adults in psychiatric outpatient settings, which is the population most likely to seek out a screening tool of this kind. Its three-factor structure (Mentalizing, Social Anxiety, Sensory Reactivity) allows sub-dimensional reporting that single-factor screens like the AQ-10 do not support.

The RAADS-14 contains 14 statements rated on a 4-point Likert scale. Each item is scored 0–3, yielding total scores in the range 0–42. The response anchors, reproduced verbatim from Eriksson 2013, are:

• 3 — True now and when I was young
• 2 — True only now (refers to skills acquired since age 16)
• 1 — True only when I was younger than 16
• 0 — Never true (and never described me)

The full 14 items are reproduced below verbatim from Eriksson 2013, Table 2 (used under CC BY 2.0). Subscale assignments derive from the factor analysis reported in Eriksson 2013 Table 4. Item 6 is the only reverse-scored item; it is worded in the neurotypical direction, so its raw response is subtracted from 3 before summing.

The three subscales partition the 14 items as follows:

• Mentalizing Deficits — items 1, 4, 9, 11, 12, 13, 14 (7 items, max 21). Theory-of-mind, social cognition, detail-focused processing, change rigidity.
• Social Anxiety — items 3, 5, 6, 8 (4 items, max 12; item 6 reversed). Group functioning, social uncertainty, friendship-building difficulty, small-talk capacity.
• Sensory Reactivity — items 2, 7, 10 (3 items, max 9). Tactile sensitivity, sensory overwhelm, auditory hyper-reactivity.

Eriksson 2013 also identifies a rapid-screen sub-form using only items 1 through 5 with a cutoff of ≥4. The reported sensitivity for this sub-form is 93% and the specificity 45–49%, suggesting it functions as a coarse triage rather than a primary screen. The full 14-item form remains the recommended instrument; the 5-item rapid-screen version is mentioned here only for completeness and is not separately implemented in this tool.

Scoring proceeds in three steps: per-item value computation (with reverse-scoring for item 6), per-subscale summation, and total summation. Pseudocode for the full algorithm:

// Inputs: responses[q1..q14], each in {0, 1, 2, 3}
// Reverse-scored items
const REVERSED = {"q6"}

// Subscale partitions (Eriksson 2013, Table 4)
const MENTALIZING = [q1, q4, q9, q11, q12, q13, q14]
const SOCIAL_ANXIETY = [q3, q5, q6, q8]
const SENSORY = [q2, q7, q10]

function item_value(item):
    let raw = responses[item]
    if item in REVERSED:
        return 3 - raw     // (0->3, 1->2, 2->1, 3->0)
    else:
        return raw

function subscale_score(items):
    return sum([item_value(i) for i in items])

// Compute three subscale scores
mentalizing_score = subscale_score(MENTALIZING)    // 0..21
social_score      = subscale_score(SOCIAL_ANXIETY) // 0..12
sensory_score     = subscale_score(SENSORY)        // 0..9

// Total = sum of all subscale scores (equivalently, sum of all item values)
total = mentalizing_score + social_score + sensory_score   // 0..42

// Severity band (LBL extrapolation; only the >=14 cutoff is published)
function band(total):
    if total <= 7:  return "Minimal"
    if total <= 13: return "Mild"
    if total <= 25: return "Moderate (positive screen)"
    return "Significant"

// Archetype (first-match-wins)
function archetype(total, m, s, r):
    if total < 14:
        return "Neurotypical Range"
    if m >= 14 and m > s*1.75 and m > r*2.33:
        return "Mentalizing-Focused"
    if r >= 6 and r*2.33 > m and r > s*0.75:
        return "Sensory-Driven"
    if s >= 8 and s*1.75 > m and s*1.33 > r*1.5:
        return "Social-Anxious"
    return "Composite"
  

The reverse-scoring transformation for item 6 follows the Eriksson 2013 published rule: a response of 3 ("True now and when I was young") on item 6 contributes 0 to the total, because endorsing this neurotypically-worded item indicates absence of the autism trait. Conversely, a response of 0 ("Never true") on item 6 contributes 3 to the total. The other 13 items are scored in the unmodified direction.

The archetype thresholds derive from a simple normalization principle: the cutoff multipliers (1.75, 2.33, 1.33, 1.5) are designed to compare each subscale at parity in percentage-of-maximum terms. For example, 14/21 in Mentalizing is 67% of max; the 1.75 multiplier on Social Anxiety brings 8/12 (also 67% of max) to numerical parity with 14, because 8 × 1.75 = 14. The 2.33 multiplier on Sensory Reactivity brings 6/9 (67%) to parity with 14, because 6 × 2.33 ≈ 14. This parity-based logic ensures that "dominance" is judged against equal proportional thresholds across subscales of unequal length.

The RAADS-14 was validated in three samples reported in Eriksson 2013: an autism sample (Phase II, n = 135), a psychiatric outpatient sample (OPD, n = 213), and a non-psychiatric general-population sample (n = 590). Within the OPD sample, an ADHD sub-sample was further analyzed separately to assess discriminant validity in a population with high feature overlap.

Internal consistency

Cronbach’s alpha for the full 14-item scale was α = 0.92 in the combined ASD + OPD sample, indicating high internal consistency. Subscale-level alphas were 0.84 (Mentalizing, 7 items), 0.87 (Social Anxiety, 4 items), and 0.81 (Sensory, 3 items) — all in the acceptable-to-strong range despite the brevity of the Sensory subscale.

Sensitivity, specificity, and AUC

At the published cutoff of ≥14, the RAADS-14 demonstrated:

The pattern of high sensitivity (97% across all comparisons) and variable specificity is characteristic of a screening instrument designed to err toward false positives rather than false negatives. The 46% specificity against ADHD reflects genuine symptom overlap between the two conditions — both feature attention difficulties and social communication challenges — and is one reason adults with diagnosed or suspected ADHD should be careful interpreting a positive RAADS-14 in isolation. Co-occurring autism and ADHD ("AuDHD") is increasingly recognized; the comorbidity panel in the tool flags this directly.

Test-retest reliability

Eriksson 2013 did not report test-retest reliability for the RAADS-14. The parent RAADS-R (Ritvo 2011) reported test-retest r = 0.987 over an unspecified interval; the brief 14-item form is presumed to retain similar stability but this has not been independently confirmed at the brief-form level. This is a documented limitation. Users retaking the screen at intervals should treat short-term variability as informative rather than as a measurement artifact.

Three-factor structure

Confirmatory factor analysis in Eriksson 2013 supported the three-factor structure (Mentalizing / Social Anxiety / Sensory) with adequate fit indices. Factor loadings and inter-factor correlations are reported in Eriksson 2013 Table 4. The factor structure has not been independently replicated in non-Swedish populations; cross-cultural invariance work would strengthen confidence in the sub-dimensional interpretation.

The published RAADS-14 has a single binary cutoff: scores of 14 or above indicate a positive screen. The four-band severity scheme used in this tool (Minimal / Mild / Moderate / Significant) is an author’s-choice extrapolation beyond the published instrument, designed to provide more granular interpretive guidance to users. The bands are:

Two of the three boundary points are anchored to published data. The 14 cutoff is the Eriksson 2013 published threshold for a positive screen. The 26 boundary is set just below the autism validation-sample means to distinguish a screen-positive result that may reflect borderline or mixed presentation from a strongly screen-positive result approaching the typical autism-sample range. The 8 boundary between Minimal and Mild is the more interpretive of the three: it separates the typical non-psychiatric range (mean ≈ 4) from the typical other-psychiatric range (mean ≈ 12), with 8 being roughly the midpoint.

The four-band scheme is interpretive, not diagnostic. Users whose total scores fall close to a band boundary should not over-interpret which band they were assigned to; the published binary cutoff (above-or-below 14) is the only assertion the validation literature directly supports.

A positive RAADS-14 screen has different positive predictive value (PPV) depending on the prior probability of autism in the screened population. The instrument’s sensitivity (97%) and specificity (46–95% across comparison groups) determine the likelihood ratios; the prior depends on the user’s population.

For the OPD comparison (97% sens / 64% spec), the positive likelihood ratio is approximately 2.7 and the negative likelihood ratio is approximately 0.05. Applied to a base-rate prior of 2.2% (CDC ADDM 2023 estimate of adult autism prevalence in the United States), the posterior probability after a positive screen is approximately 5.8% — a meaningful elevation but still well below the certainty implied by lay readings of "screen positive." Applied to a base-rate prior of 30% (a self-selected population of adults specifically suspecting autism), the posterior probability after a positive screen rises to approximately 53%.

Most users of this tool will fall closer to the high-prior than the low-prior end of this spectrum. The tool does not attempt to estimate the user’s prior probability or compute a personalized PPV — this would require demographic and motivational information the screen does not collect, and would risk false precision in any case. The bands and archetype text are framed in terms of "positive screen warrants clinical consideration" rather than "X% probability of autism" precisely because that probability cannot be reliably estimated from a self-report screen alone.

The validation samples in Eriksson 2013 produced the following mean total scores. These norms are descriptive of the Swedish adult validation samples and may not generalize precisely to other populations.

Eriksson 2013 also reported sex-stratified means within the ASD validation sample: men averaged slightly lower than women, with overlapping confidence intervals. This pattern, if real, may reflect either genuine sex differences in self-rated autism traits or selection bias in the clinical referral pathway (women referred for psychiatric evaluation in adulthood may, on average, have more severe presentations because their less-severe presentations are less often caught earlier).

A short self-report screen has inherent limits that no amount of methodological care can overcome. Users should hold these limitations clearly in view when interpreting their results.

1. It is a screen, not a diagnostic instrument

The RAADS-14 is designed to identify adults who may benefit from comprehensive clinical evaluation — not to establish a diagnosis. Diagnosis of autism spectrum disorder requires structured clinical interview (typically the ADOS-2 and ADI-R), developmental history, ruling out alternative explanations, and assessment of functional impairment. None of these can be performed by a self-report instrument. A positive screen means a clinical conversation is reasonable; it does not mean autism.

2. Validation context: psychiatric outpatients, not the general population

The instrument was validated specifically in adult psychiatric outpatient settings in Sweden. Its psychometric performance in non-psychiatric populations (the general adult public, college students, employee wellness contexts) has been less rigorously characterized. The reported 95% specificity against non-psychiatric controls is encouraging but comes from a single comparison and a single national context.

3. Does not assess childhood developmental history

Autism is, by DSM-5 definition, a neurodevelopmental condition with onset in early developmental period. The RAADS-14 asks whether traits are present now, were present in childhood, or both, but it does not assess the structured developmental history (early language milestones, peer interaction in early childhood, restricted interests as a child) that clinical diagnosis requires. A positive RAADS-14 in an adult who did not display autism features in childhood may reflect alternative diagnoses (social anxiety, ADHD, post-traumatic adaptation, or others) rather than autism.

4. Does not assess functional impairment

DSM-5 Criterion D requires that symptoms cause clinically significant impairment in social, occupational, or other important areas of functioning. Many adults with autistic traits function well in carefully-chosen environments and do not meet the impairment criterion despite endorsing many RAADS-14 items. The absence of impairment does not invalidate the experience the items describe; it does affect whether the experience constitutes a clinical diagnosis.

5. Does not screen for comorbidities

Per the Lai et al. (2019, Lancet Psychiatry) systematic review, autism in adults co-occurs with anxiety disorders (~42%), ADHD (28–44%), depression (~37% lifetime), and sleep disorders (50–80%). The RAADS-14 captures none of these directly. A user with a positive RAADS-14 who is also experiencing depression or anxiety may have a distinct and treatable condition independent of (or alongside) autism. The tool’s comorbidity panel exists precisely to surface this.

6. Self-report bias and concomitant risks

Self-report instruments are subject to insight limitations: a person who has spent decades adapting to their atypicality may not consciously notice the items that describe them, while a person actively considering an autism diagnosis may notice items more readily. Alexithymia — a commonly co-occurring condition that affects awareness of one’s own emotional and cognitive states — further compromises self-report accuracy in some users.

A separate but important concern: Cassidy et al. (2014, Lancet Psychiatry) reported that adults with Asperger’s syndrome attending a specialist diagnostic clinic had elevated rates of suicidal ideation (~66% lifetime) and suicide attempts (~35% lifetime) relative to the general population. Late-diagnosed autistic adults — precisely the population most likely to be self-administering a screening tool — appear to be at the highest end of this risk distribution. The tool’s crisis banner exists for this reason. Users in distress should seek support regardless of their RAADS-14 result.

7. Does not distinguish autism from social anxiety in some cases

Eriksson 2013 specifically noted that items 5 ("I often don’t know how to act in social situations") and 6 ("I can chat and make small talk with people") do not reliably distinguish between adults with autism and adults with social anxiety disorder. Users whose elevation is concentrated in the Social Anxiety subscale should consider that social anxiety disorder is the more parsimonious explanation in many cases, and that the two conditions can also co-occur.

8. May under-detect in women, gender-diverse, and racially minoritized adults

Autism is systematically underdiagnosed in women and gender-diverse populations, in part because of camouflaging — the conscious masking of autistic traits to fit neurotypical social expectations. Hull et al. (2019) developed the Camouflaging Autistic Traits Questionnaire (CAT-Q) specifically to measure this phenomenon. Women in particular are more likely to camouflage and are more likely to have presentations that the male-skewed validation samples of most autism instruments may not fully capture. A negative RAADS-14 in a woman who suspects autism does not rule out autism. Racially minoritized adults face additional structural barriers to diagnosis and may face similar under-detection issues, although this has been less well-studied.

9. Does not capture stereotypies

DSM-5 Criterion B includes restricted, repetitive, stereotyped patterns of behavior, including stereotyped or repetitive motor movements (hand-flapping, rocking, etc.). The RAADS-14 captures rigidity (item 14, "I get extremely upset when the way I like to do things is suddenly changed") but does not directly assess motor stereotypies. A clinical evaluation will probe these directly; the screen does not.

Methodological choices in this implementation were reviewed by Eskezeia Y. Dessie, PhD, an independent clinical reviewer with expertise in adult psychiatric assessment. The review focused on the following decisions:

1. Instrument selection. Whether the RAADS-14 is a defensible choice for a free public screening tool given the alternatives.
2. Item reproduction. Verbatim reproduction of the 14 items under CC BY 2.0 with explicit citation in three places (tool page, methodology page, schema metadata).
3. Reverse-scoring of item 6. Implementation of the published reversal rule, both in the scoring engine and in the user-facing visual indication on the tool page.
4. Severity-band derivation. Whether the four-band scheme (Minimal / Mild / Moderate / Significant) is anchored honestly to published data and disclosed clearly as an extrapolation.
5. Archetype thresholds. Whether the parity-based threshold logic produces sensible classifications across the score distribution, including edge cases.
6. Care-aware framing. Whether the crisis-bar placement, the absence of a hard escalation modal, and the Cassidy 2014 disclosure in this Limitations section are appropriately calibrated — neither alarmist nor under-warning.
7. Limitations disclosure. Completeness of the limitations section; whether each limitation is accurately characterized.
8. Comorbidity panel framing. Whether the Lai 2019 prevalence figures are correctly cited and whether the cross-tool referrals serve users honestly rather than as upsell mechanisms.

The review did not extend to clinical use cases, regulatory compliance, or use of the tool in formal diagnostic pathways. This tool is educational decision support and is not validated for clinical diagnostic use.

This methodology page is versioned alongside the tool itself. Substantive changes to the scoring algorithm, severity bands, archetype thresholds, item set, or limitations disclosure are recorded here. Cosmetic and copy-editing changes are not.

• v1.0 — May 9, 2026. Initial release. Implements RAADS-14 (Eriksson 2013) verbatim under CC BY 2.0. Three subscales, four severity bands (one published, three extrapolated and disclosed), five archetypes with parity-based thresholds, comorbidity panel from Lai 2019. Independent review completed by E. Y. Dessie, PhD.

The tool’s versioning policy is conservative: any change that alters scoring outputs for the same input will increment the major version (v2.0, v3.0, etc.) and be logged here with a description of the change and its rationale. Users who took the screen in an earlier version will not see their results retroactively recomputed.

Definitions for the technical terms used on this page are maintained as separate glossary entries on LifeByLogic. Each entry is independently citable and indexed.

Why this instrument and not the AQ-10?

The AQ-10 (Allison et al., 2012) is a widely used 10-item adult autism screen derived from the 50-item AQ. It has good psychometric properties and is appropriately brief. We did not select it because the Cambridge Autism Research Centre, which holds the copyright to the AQ family of instruments, does not permit free commercial use. A consumer tool deployed on a public website with optional advertising or sponsorship constitutes commercial use under their terms. The RAADS-14 (Eriksson et al., 2013, CC BY 2.0) is among the very few rigorously validated brief adult autism screens released under a license that permits this deployment.

Why not just create a custom instrument?

A custom instrument is unvalidated by definition. It would also be ethically questionable: psychiatric screening instruments require psychometric validation in representative samples before they can produce interpretable results. Releasing a custom screen with no published reliability or validity data, on a public-facing website, in a sensitive condition area, would produce results that look authoritative but are scientifically unsupported. The premium of using a published, peer-reviewed instrument — even one with documented limitations — over a custom one is substantial.

How are the archetype thresholds chosen?

The thresholds derive from a parity-based normalization principle. Each subscale has a different maximum score (Mentalizing 21, Social 12, Sensory 9). To compare them on equal footing, the multipliers in the threshold expressions normalize each subscale to its proportional value at the 67%-of-maximum mark (which is roughly where the Mentalizing ≥ 14 threshold sits within its 0–21 range). The cutoffs of 14 (Mentalizing), 8 (Social), and 6 (Sensory) all correspond to approximately 67% of each subscale’s maximum, ensuring consistent classification logic across subscales.

What if I retake the screen and get a different score?

Short-term variability of 1–3 total points between administrations within the same week is not unusual and does not indicate measurement instability per se — it can reflect mood, recent experiences, or the order in which you’ve been thinking about the items. Variability that crosses the 14 cutoff (e.g., 12 one week, 16 the next) may be more meaningful and worth examining. Variability of 5+ points across days suggests the screen is sampling state-related variance rather than the trait-level autism dimension; in that case, the result should be interpreted cautiously and a clinical evaluation is more informative than a single screen.

Can I share my result with my clinician?

Yes. The "How to cite" section on the tool page provides citation formats in APA, MLA, Chicago, and BibTeX. The share-card section produces a plain-text summary that includes only your aggregate score, archetype, and three subscale scores — not your individual responses to each item, which the tool does not retain in any case. Clinicians familiar with the RAADS-14 will recognize the score and subscale numbers immediately.

Why doesn’t the tool ask about childhood?

The RAADS-14 response anchors implicitly capture some childhood information ("True only when I was younger than 16," "True now and when I was young"), but this is much more limited than what a clinical developmental history would gather. Adding a separate developmental history module would be a substantive change to the instrument that would require its own validation work; we have not done this. The methodology page Section 3 of Limitations addresses this directly.

Why doesn’t the tool record my email or save my results?

By design. Mental health screening data is sensitive, and the tool’s value to the user does not depend on persistence. Running the calculation entirely in the browser, with no transmission and no storage, is the simplest privacy-preserving design. A future version may add optional opt-in result-saving with end-to-end encryption; the current version intentionally does not.

Two citations are appropriate when referencing this tool in academic or professional work: one for the LifeByLogic implementation (this page and the tool), and one for the underlying instrument (Eriksson 2013), which the tool reproduces under Creative Commons Attribution 2.0.

Citing the underlying Eriksson 2013 paper is required by Creative Commons Attribution 2.0 whenever the items themselves are referenced or reproduced. Citing the LifeByLogic methodology is appropriate when referring specifically to the implementation choices documented on this page (severity-band derivation, archetype thresholds, comorbidity panel framing).

The 15 references below underpin the methodology decisions on this page. The primary instrument citation (#1, Eriksson 2013) is the most critical; the others support specific claims about validation, comorbidity, sex differences, alternative instruments, and outcomes.

1. Eriksson, J. M., Andersen, L. M. J., & Bejerot, S. (2013). RAADS-14 Screen: validity of a screening tool for autism spectrum disorder in an adult psychiatric population. Molecular Autism, 4(1), 49. doi.org/10.1186/2040-2392-4-49 · CC BY 2.0 — Primary instrument citation.
2. Ritvo, R. A., Ritvo, E. R., Guthrie, D., et al. (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.org/10.1007/s10803-010-1133-5
3. Allison, C., Auyeung, B., & Baron-Cohen, S. (2012). Toward brief “red flags” for autism screening: the Short Autism Spectrum Quotient and the Short Quantitative Checklist in 1,000 cases and 3,000 controls. Journal of the American Academy of Child & Adolescent Psychiatry, 51(2), 202–212. doi.org/10.1016/j.jaac.2011.11.003
4. Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Clubley, E. (2001). The Autism-Spectrum Quotient (AQ): evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians. Journal of Autism and Developmental Disorders, 31(1), 5–17. doi.org/10.1023/A:1005653411471
5. Hull, L., Mandy, W., Lai, M.-C., et al. (2019). Development and validation of the Camouflaging Autistic Traits Questionnaire (CAT-Q). Journal of Autism and Developmental Disorders, 49(3), 819–833. doi.org/10.1007/s10803-018-3792-6
6. Lai, M.-C., Kassee, C., Besney, R., et al. (2019). Prevalence of co-occurring mental health diagnoses in the autism population: a systematic review and meta-analysis. The Lancet Psychiatry, 6(10), 819–829. doi.org/10.1016/S2215-0366(19)30289-5 — Primary comorbidity citation.
7. Lai, M.-C., Lombardo, M. V., Auyeung, B., Chakrabarti, B., & Baron-Cohen, S. (2015). Sex/gender differences and autism: setting the scene for future research. Journal of the American Academy of Child & Adolescent Psychiatry, 54(1), 11–24. doi.org/10.1016/j.jaac.2014.10.003
8. Cassidy, S., Bradley, P., Robinson, J., et al. (2014). Suicidal ideation and suicide plans or attempts in adults with Asperger’s syndrome attending a specialist diagnostic clinic: a clinical cohort study. The Lancet Psychiatry, 1(2), 142–147. doi.org/10.1016/S2215-0366(14)70248-2 — Care-aware citation.
9. Maenner, M. J., Warren, Z., Williams, A. R., et al. (2023). Prevalence and characteristics of autism spectrum disorder among children aged 8 years — Autism and Developmental Disabilities Monitoring Network, 11 sites, United States, 2020. MMWR Surveillance Summaries, 72(2), 1–14. doi.org/10.15585/mmwr.ss7202a1
10. Constantino, J. N., Davis, S. A., Todd, R. D., et al. (2003). Validation of a brief quantitative measure of autistic traits: comparison of the social responsiveness scale with the autism diagnostic interview-revised. Journal of Autism and Developmental Disorders, 33(4), 427–433. doi.org/10.1023/A:1025014929212
11. Lord, C., Rutter, M., DiLavore, P. C., Risi, S., Gotham, K., & Bishop, S. L. (2012). Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). Western Psychological Services. — Diagnostic gold-standard reference instrument.
12. Mandy, W., & Tchanturia, K. (2015). Do women with eating disorders who have social and flexibility difficulties really have autism? A case series. Molecular Autism, 6, 6. doi.org/10.1186/2040-2392-6-6
13. Bishop-Fitzpatrick, L., Mazefsky, C. A., & Eack, S. M. (2018). The combined impact of social support and perceived stress on quality of life in adults with autism spectrum disorder and without intellectual disability. Autism, 22(6), 703–711. doi.org/10.1177/1362361317703090
14. Croen, L. A., Zerbo, O., Qian, Y., et al. (2015). The health status of adults on the autism spectrum. Autism, 19(7), 814–823. doi.org/10.1177/1362361315577517
15. Lai, M.-C., Anagnostou, E., Wiznitzer, M., Allison, C., & Baron-Cohen, S. (2020). Evidence-based support for autistic people across the lifespan: maximising potential, minimising barriers, and optimising the person-environment fit. The Lancet Neurology, 19(5), 434–451. doi.org/10.1016/S1474-4422(20)30034-X