Cognitive screening Standardised instruments Accessible design

Custom neurodiversity assessment software

If you have standardised instruments administered digitally, you are not building a questionnaire. The administration rules are fixed, the norm tables come from the rights holder, the outcome is health-related, and the interface helps determine whether you measure what you intended to measure. Appfront builds software that lets a professional administer, score and record such instruments. The software supports this; interpreting the outcome remains with the professional who carries out the assessment.

What makes screening different from testing

A test measures whether someone knows the material. There are right and wrong answers, a pass mark that the programme itself may set, and the result is about the work delivered. With a standardised instrument for cognitive screening, it works differently. The raw score says nothing on its own: it only gains meaning when compared with a norm group of similar age and background, and that comparison only holds if the administration was carried out as it was when that norm group was compiled. The instructions, the order of the items, whether going back is allowed, the presence of practice items and any time limit are part of the instrument. They are not design choices you may improve during a design session.

That shifts the brief. You are not building a form that stores answers, but an administration environment that adheres to the instrument's rules, records the conditions under which it was administered, and presents the outcome in a way that nobody mistakes for a conclusion. The professional who carries out the assessment weighs the outcome against the history, the conversation and whatever else is known. The software calculates and records; that professional interprets.

The organisations that commission software for this sit at different points in the chain. A practice or youth care organisation uses instruments within a support or treatment pathway. A school or partnership of schools wants to know what support a pupil needs and works with an educational psychologist or psychologist to find out. An occupational health service or reintegration agency looks at what someone needs in order to work, with a client at arm's length who may only see a small part of the outcome. A training provider or employer wants to organise support for people who learn or work differently, without collecting data it is not allowed to have. What these situations have in common is an instrument that cannot be modified and an outcome that only a limited circle may read. What sets them apart is who the responsible party is, and that in turn determines almost everything in the permissions model and the retention period.

Are you in the right place? This page is about instruments with norms and about outcomes that are health-related. If you want to test knowledge, skills or suitability, with an item bank, your own cut-off score and a score you can define yourself, you belong with building assessment software. There, the question is how to test fairly and efficiently. Here, the question is how to respect an instrument that is not yours, and how to handle its results with care.

The score gets its meaning from outside

A raw score means nothing without the norm group it is compared against. That comparison is part of the instrument.

The instrument is not yours

Item texts, administration rules and norm tables are usually protected. What you may administer digitally is set out in the licence.

The outcome carries more weight

A finding about cognitive functioning is health-related and falls under special category personal data.

Administering an instrument without changing it

Moving from paper to screen looks like a matter of retyping, and that is exactly where the damage is done. A progress bar shows how much is left and thereby changes the pace. A button that allows going back changes an instrument that was normed on first instinct. An extra answer option such as "don't know" seems friendly, but it was not in the original administration and no norm group has ever used it. The same goes for an autosave that unintentionally closes an item, for an animation between two items and for an instruction that has been shortened because it did not fit on mobile.

An administration environment should therefore enforce the instrument's rules explicitly rather than leave them to the builder. What is the fixed item order, which practice items come first, is going back allowed, is there a discontinue rule after a number of consecutive errors, and does the time count per item or across a block. If the instrument has a time limit, the measurement must keep screen loading time and connection latency off the clock, otherwise you end up measuring the infrastructure. Administration on different devices calls for attention to the opposite problem: an item that can be taken in at a glance on a large monitor becomes a scrolling task on a phone, and that is a different task.

What is recorded alongside the answers is just as important. The circumstances determine how the outcome should be read, and they cannot be reconstructed afterwards. Think of the device and screen size, the input method, whether headphones or sound were used, the language of administration, whether someone was present, whether the instruction was read aloud, how often and how long pauses were taken, and whether the administration was interrupted. This data belongs with the administration, not in a log that gets cleaned up after a while.

Language deserves a paragraph of its own, because it is where most good intentions fail. An instrument that was normed in Dutch becomes a different instrument once translated: words differ in frequency and in difficulty, and the norm group never took that translated version. A translate button in the interface is therefore not an accessibility feature but an intervention in the instrument, and it should only be there if the rights holder supplies a version in another language with the norm tables that go with it. What you can do is make the surrounding environment multilingual: the explanation of what is going to happen, the consent question, the navigation and the feedback. If an interpreter is involved or the instructions were read aloud, that belongs in the administration conditions, not in the professional's loose notes.

Finally, administration does not always take place on a good connection. On location, in a classroom or at someone's home, an interruption must not mean lost data, so answers are stored securely on the device and synchronised later, with an explicit rule for the case where the same administration was continued on two devices. Equally important is what the professional sees when they return to an interrupted administration: where it stopped, how long the interruption lasted and whether the instrument allows it to be resumed later. Sometimes it does not, and then the environment has to say so rather than offering the button anyway.

Norm tables and scoring that you don't invent yourself

The step from raw score to derived score is calculation work with edge cases, and it is precisely the part you must not improvise. Norm tables are built per norm group, often in age bands and sometimes with a breakdown by education level or language background. The software has to know what happens when someone falls on the boundary between two bands, when the raw score lies outside the range of the table, when some of the items are missing and the administration can therefore not be scored, and whether interpolating between table values is allowed. These are not design questions: the answer comes from the instrument's manual, and failing that, from asking the rights holder.

In practice, we work this out as follows. The scoring rules and tables come in as source material and are not retyped by hand from a printout. The implementation is tested against the worked examples that come with the instrument, including the edge cases. Every administration records which version of the instrument and which norm table was used, so a report from last year can still be reproduced exactly. If a new norming comes out, it is placed alongside the old one and not in place of it: existing administrations stay scored against their own table, and a recalculation is an explicit action with a visible trail. An outcome that silently shifts after an update is unusable in this domain.

The software calculates, the professional interprets

How the outcome appears on screen determines how it is read. A number on its own suggests a precision that isn't there, so alongside the derived score we show the norm group it was compared with and the measurement error that goes with the instrument. The wording stays descriptive: a score lies below or above the norm group's average. What that means for this person is not stated, because that is the professional's judgement. If the instrument itself flags a cut-off value that calls for follow-up assessment, that is a feature of the instrument that we carry over, not a conclusion the software adds.

That includes what we don't build, and we'd rather say so openly. No feature that assembles a conclusion or a label from the answers, no automatically generated advice that could be read as a finding, and no home-made risk score placed alongside the instrument's own outcome. Nor do we build an overview that lines up people within an organisation by score, because a list like that invites uses the data was never collected for. If you do want to work with signals, they come from the instrument itself, in the instrument's own wording, with the norm group alongside. That isn't modesty about what software can do: it's the recognition that these outcomes take on their meaning in the conversation that follows, and that a system that fills in that meaning in advance gets in the professional's way.

The same principle is built into the way of working. An administration first gets the status draft and doesn't leave the system as a report until the qualified professional has signed it off, with room for their own commentary and for observations made during the session. The record captures who supervised the administration and who assessed the outcome, even when those are two different people. Automatically sending a result to the person assessed or to a supervisor is therefore ruled out: there is always a human between the score and the conversation.

What screening software needs to do

These components are interdependent. An administration that enforces the rules is worth little if the scoring can't be traced, and careful reporting doesn't help if everyone in the organisation can view the individual answers.

Instrument management with versions

Item order, practice items, discontinue rules and norm tables stored as fixed configuration, with several versions side by side and each administration tied to a single version.

Administration that enforces the rules

One task per screen, going back only if the instrument allows it, timing that keeps loading time off the clock, and pausing and resuming without loss.

Scoring from the instrument's own source

Raw score to derived score according to the rights holder's tables, with recorded rules for borderline cases and a calculation that remains reproducible.

Adjustments that are recorded

Font size, contrast, a calm screen, read-aloud instructions and extra breaks adjustable per person, with what was adjusted visible for each administration.

Two reports from one administration

A view for the professional with scores, norm group and administration conditions, and a plain-language view for the person assessed and their supervisor.

Access at record level

Roles that determine who may see individual answers and who sees only the outcome, with logged access and a record that isn't open organisation-wide.

Accessibility is a design requirement, not an afterthought

In this field, accessibility isn't an extra added at the end, because it affects the validity of what you measure. If navigating the environment demands attention, that attention doesn't go to the task. An interface that adds stimuli therefore partly measures itself. That's an uncomfortable thought for a design team used to atmosphere and motion: here, leaving out motion is the better choice. No animations between items, no video that starts by itself, no elements that jump around as soon as something loads, no decorative background behind a task, and no two instructions in the same sentence. What remains is calm, predictable and dull, and that is exactly right.

The hard side is simply work that has to happen early. WCAG 2.2 at level AA is the starting point, with EN 301 549 as the European standard for the digital accessibility of ICT that public organisations refer to; government bodies also publish an accessibility statement on the state of their services. In practice, that means a logical focus order, full keyboard operation, sufficient contrast, information that is not conveyed by colour alone, touch targets that are large enough, and error messages that say what is wrong. Add one requirement that is specific to test administration: a session must not time out while someone is thinking about a task. We describe our approach in more detail under building digitally accessible software.

Accessible is not the same as adapted, and that is exactly where the tension in this work lies. A larger font, less on screen at once, instructions read aloud or an extra break are reasonable adjustments that allow someone to show what they can do. At the same time, an adjustment can affect the conditions under which the norm group was built, particularly if the instrument has a time limit. We do not solve this by banning adjustments or applying them silently, but by making them possible, recording them per person and keeping them visible in the report. The professional then knows what they are looking at and can weigh up comparability.

Two practical points often only surface late. Assistive software and visual material do not always go together: if someone reads the screen with a screen reader, a task that relies on shape, colour or position simply does not work. The question then is whether a different version of the instrument exists for that situation, not whether we can invent an alternative ourselves. The second point is that the language of the environment is not the same as the language of the instrument. You may not simplify the instructions for a task, but you can simplify everything around it: what someone is about to start, how many parts it consists of, what happens with the result and who they can contact with a question. That surrounding layer determines whether someone starts with confidence or with the brakes on.

Settings should also belong to the person, not to a target group. Someone chooses their preferences once, or has them assigned at intake, and they then apply to every administration without having to be asked again. We test with people and not only with a scanner: automated checks find some of the problems, but whether a task feels calm and the instruction lands first time only becomes clear when someone actually uses it.

  • WCAG 2.2 at level AA as the starting point, not the final check
  • Full operation with keyboard and assistive software
  • Calm and predictability over visual embellishment
  • No session that times out during a task
  • Settings per person, not per target group
  • Every adjustment recorded and visible in the report
  • Tested with real users, not only with a scanner
Not sure about a large project yet?

Test your idea first: a working prototype in 1 day

With OneDayBuild, we make your idea tangible in a single day for €1,150, so you know whether further development is worth the investment. Decide to go ahead with the full build? Then we deduct the cost in full.

View OneDayBuild →

Special category personal data, purpose limitation and retention period

A result about cognitive functioning is health-related, and therefore falls under the special categories of personal data in the GDPR. This category is not subject to an ordinary balancing test: processing is in principle prohibited unless a specific exception applies, for example because the care is provided by a professional bound by confidentiality or because the data subject has given explicit consent. That is not a tick box in a processing register but a design constraint: it determines who may see the data, which fields may exist at all and how long they are retained. The role you work in matters more here than which instrument you administer.

Purpose limitation is the hardest principle in this domain, because the temptation is strong. Data collected to give someone appropriate support cannot simply be reused for selection, for assessing employees, for benchmarking between locations or for research. Each of those purposes needs its own legal basis and usually its own conversation. Data minimisation follows directly from this: the intake form asks only for what is used in the assessment or the reporting, and fields nobody can explain the purpose of are removed from the model. We would rather build a small data model with sharply defined roles than a broad model with many permissions. You can read how we set up a broader record of processing activities under building a GDPR compliance platform.

The surrounding set-up is standard work that you should have arranged in advance. Your organisation is the data controller, a builder or hosting provider is the processor, and this is recorded in a data processing agreement that also lists the sub-processors. For this category of data, a data protection impact assessment, in practice called a DPIA, is the starting point rather than the exception. Beyond that: encryption in transit and at rest, hosting in an environment and a jurisdiction you can justify, identifying data kept separate from answers so that an export for administration or research carries no names, access that is logged, and access that is locked down per role instead of being open across the organisation. For the broader requirements of a healthcare environment, see building healthcare software.

One situation deserves a separate warning, because this is where things go wrong in practice. If an employer is in control, there is very little room to process employees' health data; the Dutch Data Protection Authority's position is that this should go through the occupational health service or the company doctor, and that the employer is told only what someone needs in order to do their job. In software, that means a strict separation: the professional works in the record, the employer sees no answers and no scores, and what is sent to them is the advice on workplace adjustments. If you do not build that distinction into the roles, you build a system that gets its own user into trouble.

What is often overlooked is the administration side. Administrators and developers in practice have access to a production environment, and with this category of data you need to set that up explicitly rather than leave it to habit. That means working with test data that contains no real people, granting access to production only temporarily and with a recorded reason, and ensuring that error messages and log files carry no answers or scores. The same applies to exports, because a spreadsheet printout is the quickest way for special category personal data to leave your environment: exporting is a right that only a few roles should have, and it is logged. And an incident should follow a rehearsed path, with someone who knows when a report to the regulator and to the data subjects is required.

Retention periods that differ per role, and reporting in two directions

There is no single retention period that is right for this kind of software, because the period follows from the role in which the data was recorded. A healthcare record has a statutory retention obligation that runs for a long time, whereas a school, an occupational health service or a reintegration programme should keep to a short period. That is why we attach the clock not to the whole record but to data categories: individual answers and time measurements may disappear sooner than the established outcome that belongs in the record. The system then genuinely deletes, leaving a trail that proves it happened, and it knows what to do with an abandoned assessment instead of leaving it as a draft forever.

The same choices come back on the reporting side. A single assessment produces two views: one for the professional, with raw scores, norm group, measurement error and administration conditions, and one in plain language for the person assessed and, if they wish, for their support worker. That second version describes what was done, what came out of it and which agreements follow, without a label and without numbers that get passed around with no explanation. A support worker gets what they need to act, not the whole file. And because people have the right to access their data and to have inaccuracies corrected, this version is not an appendix produced after the fact but a full output of the system, with room for the person's own view.

Special category personal data Purpose limitation Data minimisation Retention period per role Pseudonymisation Roles and access logging Data processing agreement DPIA as a starting point Reproducible scoring

When an existing platform is the better choice

This is not an empty corner of the market. Test publishers have their own digital administration environments, and for some readers of this page those are simply the right answer. If you want to administer and score licensed instruments, and nothing else, you buy that off the shelf from the party that is also responsible for the norms. New norms come with it, the scoring is the rights holder's scoring, and you do not have to demonstrate that your own implementation is correct. We would rather say so now than halfway through a quote process.

There is also a limit that money or ambition cannot move. If the licence states that the instrument may only be administered within the publisher's environment, then building your own administration module is no longer a design question but off the table. What remains is the process around it, and that is often exactly where the pain is. So start any project with the licence terms on the table: what is allowed digitally, which norm groups may be used, which output you may show, and whether the result may enter your own system. The answers to those questions almost automatically tell you what there is to build.

Custom development therefore usually pays off not on the administration itself, but on everything around it. Think of intake and triage with your own questions, a waiting list, referral and feedback, working across multiple locations with different professionals, your own observation or interview forms that need to end up in the same file alongside a licensed instrument, reporting in your organisation's own words, and a role model that separates more strictly than a general platform can. In an occupational health service or an employer context, that last point is not a luxury but the core of the assignment. Accessibility is another reason: if the platform does not reach the level your target group needs, you are stuck with someone else's roadmap.

So what we look at in a first conversation is not the feature list, but the three constraints that determine everything else: what the licence permits, in which role you process the data, and who may see which output. Until those three are settled, every functional wish is a sketch, because they can rule each other out. A result you want to show a client may be an output you are not allowed to share. An instrument you want to administer digitally may be tied to a single environment. Only once that is clear does the conversation about intake, file, reporting and integrations become concrete, and usually it then turns out that the gain is not in the administration but in the process around it.

In practice, we therefore often end up with a middle route. The assessment takes place where it belongs, while intake, record, permissions and reporting sit in your own environment, with an integration or a controlled transfer of the result and without needlessly copying individual answers. You can read what such a record layer looks like under developing a case management system. The flip side comes with the territory: if you build it yourself, keeping up with norms, accessibility requirements and privacy frameworks falls to your own organisation, and that work is never finished.

  • Publisher's platform if you only administer licensed instruments
  • Platform if the licence does not allow administration outside it
  • Custom software for intake, triage, record and feedback
  • Custom software if you have your own lists alongside a licensed instrument
  • Custom software if roles need to be separated more strictly
  • Middle route: assessment with the publisher, record and reporting with you

Frequently asked questions about screening software

No. The software supports a professional in administering an instrument, in calculating scores according to the established norm tables and in recording the assessment in a record. Interpreting the result is up to that professional, who weighs up the history, the context and the conversation. A screening result is a signal that may prompt further assessment, not a conclusion about a person, and the software presents that result accordingly.

Two things. The data falls under a stricter regime, because a result about cognitive functioning is health-related and therefore counts as special category personal data. And scoring is not free-form: for a standardised instrument, the administration rules and norm tables are held by the rights holder, and you may not devise or approximate them yourself. If you only want to test knowledge or skills, without those two requirements, a generic assessment environment is a better fit.

Only if the rights holder permits it. Item texts, administration rules and norm tables are generally protected and supplied under licence, sometimes only for use within the publisher's own platform. That is one of the first things we look into: may this instrument be administered digitally, which norm groups may be used and which output may you show. Without that agreement, you would be building an assessment environment for instruments you are not allowed to put in it.

By making them possible and recording them. A larger font, fewer stimuli on screen, instructions read aloud or a break between sections are reasonable accommodations. At the same time, an accommodation can affect the conditions under which the norm group was built, for example if the instrument has a time limit. The software therefore records for each assessment what was adjusted, so the professional can take that into account and it remains visible in the report.

Where you want it, and no longer than you can justify. The retention period depends on the role you work in: a healthcare record carries a statutory retention obligation that runs for a long time, whereas a school, an occupational health service or an employer should keep to a short period. In the data model, we separate the raw answers from the outcome, so the detail can be deleted sooner than the conclusion that should remain in the record.

Yes, and it's more than an export button. People have the right to access their own data, and a printout full of scores without explanation helps no one. That's why we build two views of the same assessment: one for the professional with raw scores, norm group and administration conditions, and one in plain language that describes what was done, what the outcome was and which agreements follow from it.

Related services

Assessment software for testing and evaluation

If it's about knowledge, skills or suitability, with your own item bank and a cut-off score you set yourself, then building assessment software is the right starting point. No third-party norm tables, no health-related outcome.

Software for mental healthcare

If the question arises within a treatment context, with records, care pathways and professional confidentiality, then building mental healthcare software fits better with the process in which the assessment is only one step.

Intake before the conversation begins

The work before the assessment, from registration and triage to gathering what is already known, belongs to building intake software. Data minimisation is just as strict there.

Compare your assessment process with the licence

Tell us which instruments you administer and what the licence says about them, what role you work in and who ultimately sees the result. These three answers usually show whether you need a platform, custom software around the administration, or a combination of both. And whatever gets built, the professional remains the one who interprets the result.

Edit content