Screened, Scored, and Cleared: Why Assessment Tools Systematically Miss the Highest-Risk Neurodivergent Population

We screen for Autism and we screen for ADHD. Screening tools are developed and validated within the psychiatric field that categorises neurodevelopmental conditions by behavioural presentation. What they do not measure is what happens neurologically below awareness and observation.

A shift from psychiatry-led to neuroscience-led assessment would change the screening and detection landscape fundamentally. The tools we use determine who we find — and when we measure behaviour without measuring the system producing it, the people whose presentations are most complex are the people most likely to be missed and left undetected.

This creates four structural problems with the current assessment landscape.

Every self-report screening tool assumes the person knows what they are experiencing.

When masking develops before conscious self-awareness, the person doesn't know they are masking. A child who learned to manage eye contact and suppress sensory responses before they could recognise what they were doing has no reference point for the cost. By adulthood, those compensatory patterns are their baseline. They have not yet experienced the effortlessness of conversations without eye contact. Their screening response to whether eye contact is stressful is "no." The answer is subjectively honest and clinically incorrect.

This may be the most consequential problem that is the least discussed.

Every screening tool measures one trait at a time, which works when a condition produces a consistent pull in one direction. Autism pulls toward sameness. ADHD pulls toward novelty. The item captures the pull. The score reflects severity.

When both Autism and ADHD profiles are present, the pulls oppose each other and cancel out on the measuring instrument.

An item asks about preference for routine. The autistic system says yes. The ADHD system says no. The person needs routine and cannot sustain it. They select the middle of the scale. The score does not flag. This repeats across every domain — sustained attention, social avoidance, sensory seeking versus sensory withdrawal.

In each case, the lived reality shows two extreme positions in active conflict producing significant impairment. The score reads as moderate. The instrument cannot detect the conflict because it was designed for a single directional pull.

Co-occurrence rates are estimated between 30% and 80%. This is not a rare edge case. It is operating in a substantial proportion of every screened sample.

Anxiety and PTSD can produce social withdrawal, sensory hyper-reactivity, avoidance, emotional dysregulation, and extended recovery needs. On existing instruments, a person with chronic PTSD could exceed autism thresholds without being autistic.

The distinction determines the intervention. Environmental adaptation is the primary response to architectural neurodivergence — the system is built this way, the environment needs to fit. Psychological intervention is the primary response to acquired anxiety and trauma. When both are present, the architectural layer must be stabilised first.

Current tools identify that features are present. What they cannot reliably determine is whether a current feature is architectural — present as part of the person's neurodevelopmental profile from early life — or acquired, emerged in response to chronic environmental mismatch or sustained stress.

Current tools produce a single score at a single point in time. The features they measure — sensory sensitivity, social tolerance, communication and executive functioning — vary with the person's regulatory state. An individual assessed during a stable period will score lower than the same individual assessed during burnout. Without knowing the regulatory state at the time of assessment, the clinician cannot interpret the score.

Consider the undiagnosed individual with co-occurring profiles. Their masking is subconscious — they cannot report on traits they do not know they have. Their competing drives cancel each other out on screening items. Their presentation overlaps with acquired anxiety. They are assessed on a stable day.

The result: they score below threshold on every available instrument. They are told they do not meet criteria. They return to the environment producing the load that will eventually collapse them.

The instrument did not fail on any single item. It failed because its measurement architecture is misaligned with the presentation it needed to detect.

Current tools have enabled identification for millions of people. Their structural assumptions, however, produce a systematic bias against the population carrying the worst outcomes. Addressing this requires a different measurement architecture.

#1  Measure through pattern, not trait identification.

The person who does not know they have a trait cannot report on it. They can report on how different their energy is across settings and how long recovery takes after demands that others manage without difficulty.

#2  Measure the conflict between opposing drives, not just each pole independently.

The cancellation effect requires a dedicated measurement domain for the regulatory interaction itself.

#3  Incorporate multiple data sources.

A parent sees the recovery collapse. A teacher sees the performance. The discrepancy between them is diagnostic information.

#4  Assess developmental trajectory.

Determine whether each elevated feature has been present from early development or emerged at an identifiable time. This separates architecture from acquisition and determines intervention sequencing.

#5  Capture regulatory state at the time of assessment.

This single addition transforms the interpretation of every score that follows.

The AuDHD subgroup carries the worst outcomes of any neurodivergent population: higher rates of depression, anxiety, suicidal ideation, and burnout than those with either condition alone. Higher rates of misdiagnosis. Higher rates of inappropriate treatment. Higher rates of accumulated adverse experience across the lifespan. And the lowest identification rate — because the measurement tools designed to find them are structurally unable to detect them.

The cancellation effect means their autism scores are suppressed by their ADHD. Their ADHD scores are suppressed by their autism. Their masking is subconscious, so their self-report is inaccurate without dishonesty. Their presentation overlaps with anxiety and PTSD, and no current instrument can tell the difference. They are screened, scored, and cleared.

The measurement gap is not a psychometric limitation. It is a structural failure at a population scale. The conceptual and structural foundations for a different approach exist. What is required is the will to acknowledge the shift.