What makes a neurodiversity profile valid?

A shared language is not a measurement instrument

Under proposed SEND reforms, Individual Support Plans (ISP) will become a statutory requirement across Targeted, Targeted Plus, and Specialist tiers. Designed to ensure every child and young person with SEND receives a standardised profile without needing a formal diagnosis, these plans will be co-produced by schools and parents, and will travel with the pupil digitally between settings.

The structure underpinning this mandate is set out in Appendix D. Five developmental areas replace the traditional four areas of need, breaking down into 26 named components. This taxonomy will directly shape the updated Code of Practice, organise the National Inclusion Standards, and form the procurement gateway for the digital library of identification tools.

As a shared language for teachers, Appendix D is a genuine improvement on the four areas of need. Children moving between support layers will be seen more consistently. That matters.

But a statutory digital record that follows 1.7 million children and feeds every needs assessment in England is not a communication device. It's a measurement instrument. Read Appendix D as a measurement specification (which anyone building a tool against it must now do), and four serious problems emerge.

The department says it will keep refining the definitions with experts. None of these problems are fixed today. Schools will be writing plans before 2028.

Four structural flaws in the Appendix D taxonomy

1. Some components can't be told apart

The social and emotional area lists six components, but three of them cover substantially the same ground: understanding and managing emotions, regulating emotions, and identifying and expressing feelings. These aren't three distinct constructs. At best they're two, more likely one, split across three labels.

Appendix D then contradicts itself. The prose description regroups those same items into different clusters than the component list does. It also introduces a seventh component that doesn't appear in the list at all: recognising social rights and responsibilities such as dignity, respect and safety. That's a values construct, not a developmental one. Is it in scope? Out of scope? A supplier building to this specification has to guess.

This matters because internal consistency is the foundation of a valid score. If the components within an area don't hold together as a coherent construct, adding them up produces a number, not a measurement. No amount of factor analysis rescues a taxonomy that can't agree with itself on what it contains.

2. The document says the constructs move

Appendix D explicitly states that sensory processing is dynamic rather than fixed: that how a child responds to sensory input shifts with context, wellbeing and emotion. It says development in each area can move forward, pause or feel uncertain, and that this is normal.

This is a statement about variation within a single child, and it changes what a good instrument must do. If the thing you're measuring moves, one score is a point estimate of a moving target. You can't tell whether a child has genuinely changed or whether you happened to measure on a bad Tuesday, unless the instrument publishes test-retest reliability and a minimal detectable change figure.

Very few school-facing profiling tools publish either. The reform has quietly written this requirement into its own taxonomy without anyone noticing it is a requirement.

3. The five areas are correlated by design

Appendix D states that the areas are interconnected and that progress in one influences the others, using communication strengthening social and emotional development as its example. That's almost certainly true. It also means a five-area profile is not five independent measures.

Display the five areas as separate dashboard columns and you're double-counting, by an amount nobody has quantified, because no factor analysis has been published. Confident-looking data built on unexamined correlation is still sand.

4. There is one age expectation in the entire taxonomy

Speech sounds get one: most children have acquired them by age five. Nothing else does. The same 26 components are listed for the entire 0 to 25 system.

Appendix D half-acknowledges this. It notes that the early years are a sensitive phase and that many four-year-olds are still learning to sustain attention. But forming relationships at two and forming relationships at 22 are different constructs, measured by different instruments, against different norms. A tool offering one item set for every age isn't covering the taxonomy. It's averaging over it.

How to tell an instrument from a questionnaire

There's an international standard for judging whether a measurement instrument is fit for purpose: COSMIN (COnsensus-based Standards for the selection of health Measurement INstruments). It emerged from a multi-country Delphi study and is what systematic reviewers reach for when assessing an outcome measure.

Reliability and responsiveness are the two properties Appendix D has quietly made load-bearing, and the two commissioners ask about least.

Every profiling pitch you'll hear leads with population data: dashboards, benchmarks, needs assessments. But population data is a by-product. The unit that matters is one child, measured repeatedly.

Appendix D tells you plainly that a single child's profile fluctuates with context, wellbeing and emotion. If an instrument can't give you the same answer for a stable child two weeks apart, aggregating it doesn't fix that problem. It hides it.

Get the within-child measurement right and the population view comes free. Do it the other way round and you get a very confident dashboard built on sand.

The three words to check first

Appendix D names proprioception, interoception and the vestibular sense as the internal senses within sensory development. Almost no profiling tool in school-use, measures any of the three.

This matters beyond compliance. ICD-11 now treats sensory hyper- and hypo-reactivity as a core feature of autism rather than an associated difficulty. A sensory profile missing three of its named components will under-identify autistic children, and will miss most of all the children whose sensory differences are internal and therefore invisible to a classroom observer.

Search any specification you're given for those three words. It takes 30 seconds and it's remarkably discriminating.

Case study: evaluating the Portsmouth Neurodiversity Profiling Tool

This is worth examining first, as it represents the approach most commissioners will encounter and is being adopted well beyond Portsmouth.

Its nine strands cover speech and language, energy levels, attention, impulse control, emotion regulation, sensory processing, flexibility and adaptability, empathy and systemising, and motor skills.

Piloted with 50 families between March 2021 and April 2022 (46 completed), it was rolled out across Portsmouth in July 2022. School staff complete the profile alongside the family following a two-hour training session.

The strengths of the concept are genuine. Co-produced with Portsmouth Parent Voice, it provides families and schools with a shared language and identifies need without making a formal diagnosis a prerequisite for support. None of that is in question.

What is missing, however, is the psychometric evidence: there are no published studies on content validity, factor analysis, internal consistency, test-retest reliability, inter-rater agreement, criterion validity, responsiveness, or measurement invariance across ethnic groups.

The nine strands were not established through the COSMIN methodology. If two staff members rate the same child, do they agree? If a staff member rates a stable child a month apart, do the scores match? There are currently no published answers.

Even setting psychometrics aside, there is a more immediate issue for those who have already procured it: the nine strands do not map onto Appendix D. "Energy levels" are completely absent from the taxonomy. Meanwhile, working memory, planning and organisation, the three internal senses, and the distinction between fine and gross motor skills are all present in the taxonomy but missing from the tool.

Consequently, any area using the Portsmouth tool faces an inevitable mapping exercise, regardless of how the evidence base unfolds.

This does not make early, needs-led profiling inherently wrong. It simply means the instrument itself must be robust enough to justify the claims being made for it.

A commissioner's checklist: seven questions for your next specification

The National Inclusion Standards will feature a digital library of identification tools, supported by up to £15m to expand the evidence base by 2028, alongside a £4m UKRI project on early identification and strengths-and-needs assessment. These represent the key procurement gateways, and the standards applied at these checkpoints are everything.

Put these seven questions in writing.

• ·Which of the 26 components in Appendix D do you measure with a validated instrument, and which do you not? Request this mapping as a component-by-component table. Any supplier unable to provide this within a week has not completed the necessary work.
• ·Where is the published psychometric evidence for each construct? Look for content validity, structural validity, reliability, and criterion validity citable in the COSMIN database or a published systematic review. Evidence residing solely in the supplier's marketing materials is not independent evidence.
• ·What is your test-retest reliability, and how much change is large enough to be meaningful? Appendix D notes that these constructs fluctuate. Without a minimal detectable change figure, you cannot distinguish genuine progress from measurement noise, leaving you unable to evaluate any commissioned intervention.
• ·Which instruments do you use at which ages, and where are the published norms? The taxonomy is age-flat, but your tools cannot be. Different developmental stages demand tailored instruments with published normative data at every step.
• ·What do you measure that highlights the child's strengths? The reforms repeatedly emphasise strengths and needs. This is not ornamental language. A tool that only logs deficits misses half the brief and reinforces the deficit-framed approach that neurodivergent advocates rightly challenge.
• ·Can the output be aggregated and benchmarked? Structured scores from validated instruments aggregate into a clear population-level needs assessment and can be tracked over time; free text and unverified rating scales cannot. Ask also how the five areas correlate to ensure you aren't double-counting data.
• ·Where is the independent evaluation conducted in UK schools with UK populations? This challenges every supplier in the market. Most validated instruments were normed elsewhere, making cross-cultural validity in British classrooms the weakest link in the field. Ask anyway. The transparency of the response will tell you who is being honest.

Setting the standard for the future

Appendix D is one of the most consequential pages in the consultation, yet it remains one of the least discussed. As a shared language for educators, it represents a genuine step forward.

However, as the foundational specification for a statutory digital record that will follow 1.7 million children and underpin every needs assessment in England, neither the taxonomy itself nor anything built upon it has been tested.

The reform is being shaped right now. The standards you demand during procurement will set the standard for the entire sector.