A worklist ordered by what needs attention

A clinician's list of patients, studies or tasks, ordered so that what most needs attention is at the top (by acuity, by risk, by a flag, or by an AI's confidence that something is wrong) rather than by arrival time or name. The pattern includes how the ordering is explained, how it can be overridden, and what happens to the items that are not urgent.

A worklist is where a clinician's attention is allocated, and the order of a list is an instruction about where to look first. When the order is time of arrival, the sickest patient can be tenth. When the order is a triage score or an AI flag, the list carries a claim (this one matters more) that the clinician has to be able to trust and to question.

Ordering well can shorten the time to a critical finding. Ordering opaquely can bury one, because an item pushed to the bottom is an item that may not be seen today.

Prioritisation moves risk rather than removing it. An AI that segregates unremarkable chest X-rays (which is how Qure.ai describes one use of qXR) shortens the queue for the reporter and lengthens the interval before a study the AI called normal is read by a person, so the false-negative rate of the ordering becomes a clinical property of the service.

The list should say why an item is where it is: a score, a flag, a rule, with its source. It should never hide items (deprioritised is not dismissed) and it should show age as well as rank, so that a low-priority item that has waited a long time rises. The FDA's clinical decision support guidance draws a line at time-critical decisions, where a clinician cannot realistically review the basis of a suggestion; a worklist that reorders in real time needs its reasoning visible at a glance for the same reason.

Worklists are tables, and tables are where screen-reader support most often breaks: rank, flag and status need to be real columns with headers, not colour and position. Sorting and filtering must be operable by keyboard with visible focus (WCAG 2.2, 2.4.7), and a change in order that happens automatically should be announced (Status Messages, 4.1.3), so that a clinician using assistive technology is not reading a list that has silently reshuffled. Priority conveyed by red or bold alone fails Use of Color (1.4.1).

Long lists need a way to move through them that does not depend on fine pointer control (Target Size Minimum, 2.5.8), and a session that times out mid-list should not lose the clinician's place (Timing Adjustable, 2.2.1).

Two products in the library describe a prioritised list on their public pages, from opposite ends of the pipeline. Qure.ai describes qXR as clearing backlogs by segregating unremarkable chest X-rays, and reports a negative predictive value for that step: the company's figure, from its own page. That is prioritisation by removal: the ordinary is set aside so the remainder is read sooner. Huma describes a unified clinician dashboard on which multiple teams review patient data, with identifying and prioritising at-risk patients made easier and automated flagging and triage available as a dashboard feature. That is prioritisation by promotion: the concerning is raised.

Neither page shows the list itself, how its order is explained to the clinician, or how it is overridden. Those are the parts Recent would most want to see, and the parts a screenshot would show; the library will add them as they can be sourced.