The national radiology triage criteria, by clinical scenario

The Triage Tool holds the Radiology National Clinical Network's prioritisation criteria as a library of clinical scenarios. Each scenario gives the recommended imaging and a priority category. Anyone can read it without signing in.

Health NZ · national triage criteria 17 clinical areas Open to read, no sign-in

Open the Triage Tool

What it is

A referral should get the same priority in Invercargill as in Whangārei. The Triage Tool puts one set of criteria in front of every triager, and the clinical section leads who write the criteria keep it up to date.

The scenario library

As of November 2026, 461 scenarios written by section leads, in 17 clinical areas. Each gives the recommended imaging, a priority and a comment on when the call changes.

Browse and search

A triager picks a clinical area, opens a subheading and reads its scenarios. A search box filters the scenarios in the section that is open.

The Priority Guide

The National Prioritisation Guide on one page: P and S categories with their turnaround and reporting targets, acute priorities and service disruption levels.

Triage Tool · Breast / Palpable breast masses
The Breast section open in the Triage Tool.
          The sidebar lists 17 clinical areas with a scenario count for each. The Palpable breast
          masses subheading is expanded, showing scenario cards with priority badges (P3 or P2,
          and P2), the recommended modality (Mammography or US), a note on the imaging and a
          comment on when to upgrade.
A section open at one subheading. Each card shows the priority, the recommended imaging and the comment. Captured on beta, which also shows a Triage Scout link that production does not have yet. Click a screenshot to enlarge it.

Scenarios and priorities

Section leads write the library and keep it current. Their names, and the date each section was last updated, show at the top of the section.

How the library is organised

Section17 clinical areas, for example Breast or Spine
Subheadingfor example "Palpable breast masses"
Scenarioone presentation and its recommendation

A link to a scenario opens the tool at that scenario and highlights it, so a recommendation can be shared in an email or a triage note.

What a scenario says

Tx The clinical scenario: the presentation, worded the way referrals put it.
M The modality: the recommended imaging, with any alternative (for example Mammography / US). Some scenarios say "HNZ Imaging not indicated" instead.
P The priority: a P or S category, sometimes a pair such as "P3 or P2".
C The comment: when to upgrade, and what changes the call.

Priority categories

P categories count from the day the referral is received. S categories count from a target date the referrer specifies.

P1 within 1 week, P2 within 2 weeks and P3 within 6* weeks. If capacity is constrained, P4 can wait up to 6 to 12* weeks and P5 more than 12 weeks. Reports are due within 24 hours for P1 and P2, and 48 hours for P3 to P5. The asterisks are the guide's own; its notes are in the Priority Guide.

Finding a scenario

In production, a triager finds a scenario by browsing. On beta, Triage Scout also suggests scenarios from the clinical details of a referral.

1

Browse

Pick a clinical area from the sidebar, then open the subheading that fits. Production opens on "Select a section" and starts here.

2

Search within a section

The box at the top of a section filters its scenarios by the words typed.

3

Triage Scout, on beta

Type or paste the clinical details and Scout suggests the closest scenarios, each with a confidence rating and a sentence on why it matched. Scout gathers likely scenarios from the library, then an AI model ranks them. A query of a single word gets a keyword search instead. If the text uses an abbreviation Scout does not know, it asks what it means before matching.

Triage Scout · beta
Triage Scout on beta with the query "62F
          palpable left breast lump, 3 weeks, no prior imaging". The best match is under
          Breast, Palpable breast masses: "Palpable breast mass. Female, 40 years of age or
          older, initial evaluation." It shows P2, Mammography / US, a high confidence tag,
          a line on why the AI matched it, and a View in section button.
An invented referral matched on beta. The result shows the scenario as the section leads wrote it, the confidence and the reason for the match.
Scout can only point to scenarios already in the library and does not set a priority of its own. When nothing fits it says "No matching scenario found". If the AI service fails, it offers a keyword search instead.
The text typed into Scout goes to an external AI service (Claude) to find matches. The page asks for no patient names, NHI numbers, dates of birth or other identifying details, and says the tool keeps no copy.

Features

Everyone can browse and read the Priority Guide. Section leads, editors and reviewers sign in to change the library.

For everyone
Browse

Sections and scenarios

Everyone, no sign-in

The sidebar lists the clinical areas with a scenario count for each. A section opens with its subheadings closed; opening one shows its scenario cards.

  • The top of each section shows when it was last updated, and a Section leads button lists who maintains it.
  • A link to a scenario opens the section at that scenario and highlights it.
Reference

Priority Guide

Everyone, no sign-in

The National Prioritisation Guide, opened from Guides in the header. It sets out the P and S categories with their turnaround and reporting targets, PET-CT, the acute priorities from under 30 minutes to under 2 days, the four service disruption levels and what can still be imaged at each, and the reset and restore rules.

Under the reset and restore rules, if services have to be paused, services for Māori and other at-risk populations are the last stopped in any priority band and the first restarted.
Priority Guide · P categories
The National Prioritisation Guide with
            the P categories open. P1 to P5 each have a definition, an overall turnaround (under
            1 week, under 2 weeks, under 6* weeks) and a report-within target (24 or 48 hours),
            with rows for PET-CT and overdue reporting. Closed panels below are for S
            categories, acute priority, service disruption levels and reset and restore.
The P categories. The S categories, acute priorities, disruption levels and reset and restore are in the panels below them.
For section leads, editors and reviewers
Access

Who can change what

Signed in

Sign-in uses the HNZ Rad Tools sign-in shared with the other tools, with a code sent by email. An administrator gives each person one of three roles.

ActionReviewerEditorAdmin
See AI drafts waiting for review Yes Yes Yes
Work the review queue Yes Yes Yes
Edit an existing scenario Yes Yes Yes
Add, archive or reorder scenarios No Yes Yes
Change sections, subheadings and section leads No Yes Yes
Send an approved scenario back to review No No Yes
Editing

Edit mode

Reviewers, editors, admins

A signed-in user switches to edit mode with Edit in the header and works on the page itself. Nothing is saved until they click Save and confirm.

  • Modality is picked from a standard list (CT, MRI, US, XR, Fluoro, DEXA, PET/CT, Mammography, SPECT, Angiography), never typed.
  • Priority is picked with P1 to P5 and S1 to S5 buttons. A pair such as P2/P3 is allowed.
  • "HNZ Imaging not indicated" marks a scenario where the right answer is no scan, and clears its modality and priority.
Review

Review queue

Reviewers, editors, admins

Only AI-written content needs sign-off. Review in the sidebar opens a queue that shows one scenario at a time with every field editable. Approving saves any edits first, so an approval covers what was on screen. Rejecting asks for a reason.

  • The queue covers AI drafts first, then original scenarios whose comment the AI rewrote. It can be limited to either kind (AI Generated or AI Enhanced).
  • If a colleague has already approved or rejected a scenario, the queue treats it as reviewed.
  • An administrator can send an approved scenario back to review.

New scenarios from referral data

The hand-written library covers part of what arrives in referral queues. To widen it, a year of real CT and MRI referrals was run through Claude models to propose new scenarios. Readers see none of them until a reviewer approves it.

1

Classify

Referrals went to the model in batches of 80, CT and MRI mixed, so that the same clinical question asked of both modalities became one scenario. Each scenario was placed under an existing section and subheading. This gave over 7,500 raw scenarios.

2

Merge and consolidate

Near-duplicates across batches were merged, leaving 5,502. A second, larger model then grouped clinically equivalent scenarios and rewrote them in the library's wording, leaving 2,118.

3

Sort by evidence

A scenario seen in 3 or more referrals became an AI draft for review. One seen less often went to a review pool that only editors and reviewers can see. The import checked for duplicates once more as scenarios went in.

4

Review

An AI draft carries an amber "AI Generated" badge. A reviewer approves it in the review queue, after which everyone sees it with a green "AI Generated, Radiologist Approved" badge, or rejects it with a reason. Rejected drafts are archived, not deleted.

As of November 2026 the production library holds 461 scenarios written by section leads, 699 AI drafts waiting for review and 1,418 scenarios in the review pool. Readers see the section leads' scenarios and any AI drafts that have been approved.

Where to find it

The Triage Tool is at triage.hnzradtools.nz. Anyone can browse it and read the Priority Guide without an account. Section leads, editors and reviewers sign in with the HNZ Rad Tools sign-in, and an administrator gives them their role. Triage Scout is on the beta site, beta.triage.hnzradtools.nz, while it is tested.

Open the Triage Tool

How it is built is in the architecture pack (Health NZ staff sign-in).