AI drafts for the 12 nursing problem areas — now with a recommended score

The 12 nursing problem areas are a statutory documentation requirement, and the requirement is stricter than it is often read: all twelve areas must be addressed — a problem does not have to be found in all twelve. In practice, that difference means the nurse has to work through the service user's entire record before she can even start writing.

That first pass is the one Validi now takes. On the service user's profile, under the tab "The 12 nursing problem areas", you will find the ✨ AI draft button. It gathers what the record already says about each of the twelve areas and proposes a wording — with references to the notes the proposal rests on. What is new is that the draft now also recommends a score on a 1-5 scale, which the nurse confirms or changes.

The principle is unchanged: the AI decides nothing. It gathers and recommends. The professional makes the call and saves.

What are the 12 nursing problem areas?

The twelve areas are the framework that nursing documentation has to cover wherever a service carries out nursing tasks: functional level, musculoskeletal system, nutrition, skin and mucous membranes, communication, psychosocial circumstances, respiration and circulation, sexuality, pain and sensory perception, sleep and rest, knowledge and development, and elimination of waste products.

The framework comes from the Danish Patient Safety Authority's guidance on nursing documentation (guidance no. 10239 of 1 December 2025), which elaborates on executive order no. 1361 of 24 November 2025 on the patient records of authorised healthcare professionals. The guidance puts it briefly, and it is worth holding onto: address each area, and note down what is relevant.

At inspection, it is rarely an individual score that causes trouble. It is the gaps — areas with neither an assessment nor a reason why the area is not relevant. A draft that says outright that an area is not described makes those gaps visible while there is still time to do something about them.

For a service with many people in treatment, the assessment itself is rarely what takes the time. It is the gathering — reading three months of notes to work out what the record actually says about sleep, nutrition and skin.

How the AI draft works, step by step

1. The nurse opens the service user's profile and the "The 12 nursing problem areas" tab.
2. She clicks ✨ AI draft.
3. Validi works through everything held on the service user in the system: journal notes, medication and prescriptions, urine samples, attendance, and uploaded Word and PDF documents. The most recent notes carry the most weight.
4. For each of the twelve areas, it proposes a description in objective, documentable language — and an action plan wherever the material clearly calls for one.
5. Every proposal comes with source references to the specific notes, with dates, for example (see Doctor's note 2026-06-14), so the evidence can be checked on the spot.
6. On top of that, it recommends a score from 1 to 5 (No problems → Very severe problems) with a one-sentence rationale.
7. The nurse works through each area, confirms or changes both the text and the score, and saves area by area.

A draft for Sleep and rest might look like this:

"The service user describes repeated night-time waking and difficulty falling asleep after 02:00 (see Journal note 2026-06-02). No sleep medication has been prescribed during the period (see Doctor's note 2026-06-14)."

Alongside the text sits the recommended score — here 3, Moderate problems — with the rationale: recurring sleep difficulties with no documented effect on attendance. The nurse can open both notes, confirm or change the score, and add her own assessment before saving the area.

"Expand all" opens all twelve areas at once, so the whole assessment can be worked through in a single pass.

The approach is the same one Validi uses for drafts of treatment plans and status reports: the system writes the first version, the professional writes the assessment.

A recommendation is not a decision

A number on a scale looks more finished than a piece of text does. That is why the recommendation is kept within three clear limits.

No score without evidence

Areas the record does not describe get the text "Not described in the record." — and no recommended score. Nothing is guessed. In the same way, "No problems" is only recommended where the material positively describes the area as unproblematic. An absence of mention is not the same as an absence of problems.

Nor is an undescribed area a failure. The requirement is that all twelve areas are addressed — not that findings turn up in all twelve.

Nothing is saved automatically

The draft only fills in empty fields, and only pre-selects a score in areas that have not yet been assessed. Assessments already written are left untouched. A recommended score does not become part of the record until the nurse has confirmed or changed it and saved that particular area.

The rationale is support, not record

The one-sentence rationale behind the recommendation appears on screen as an aid to the assessment, and it is not saved. It is there so that the professional can see what the recommendation rests on — and disagree with it.

AI-assisted content is visible as such

Any version of an assessment containing AI-assisted content is visibly marked 🤖 AI-assisted — including in the version history. At an inspection or in a complaint case, you can therefore document what started life as a draft and who addressed it.

The marking is not just good practice. Article 50 of the EU AI Act requires transparency about content generated with AI, and all AI-generated content in Validi is marked and logged.

GDPR and data processing

Civil registration numbers are never sent to the AI service. Names are pseudonymised before the call is made and reinserted in Validi afterwards.

The AI functions run on Azure OpenAI in the EU, in the Sweden Central region, under a data processing agreement. All personal data is stored and processed in the EU. This is the same line the rest of the platform follows — read more about GDPR in a journal system for addiction treatment.

Getting started

The feature is enabled per clinic. It is not active until you ask for it, and it does not change how you document today. What it removes is the gathering work that sits in front of the assessment.

If you already work with digital care plans in Validi, the twelve areas are the next place where the same principle pays off: the system gathers the material, you make the call.

To see the feature demonstrated on fictional data, you can book a no-obligation walkthrough of Validi.

Frequently asked questions about AI drafts for the 12 nursing problem areas

Does the AI replace the nurse's assessment?

No. The draft is a recommendation built on what is already in the service user's record. The nurse reads each area, confirms or changes both the text and the score, and saves area by area. Nothing saves itself. The professional judgement is — and remains — the professional's responsibility, exactly as the documentation rules require.

What happens if the record says nothing about an area?

The draft writes "Not described in the record." and recommends no score. That is deliberate: a score has to have evidence behind it, and an empty area must not look like an assessment. The nurse can then address the area another way — from a conversation with the service user, for instance — and write it in.

Can we see what is AI-assisted?

Yes. Any version of an assessment with AI-assisted content is visibly marked 🤖 AI-assisted, and the marking follows through into the version history. So you can always document what came from a draft and what the professional confirmed or changed. That makes the feature useful at inspection rather than a liability.

Are civil registration numbers sent to the AI?

Never. Civil registration numbers are not included, and names are pseudonymised before the call and reinserted in Validi afterwards. The AI functions run on Azure OpenAI in the EU under a data processing agreement, and all personal data is stored and processed in the EU.

Where can we see what a recommendation is based on?

In the draft itself. Every wording carries source references to the specific journal notes, with dates — for example (see Doctor's note 2026-06-14) — so you can open the note and judge for yourself whether the evidence holds. That is the quickest route to disagreeing with a recommendation: go to the source.