Batch 01 · ER triage

Priority queue collapse during flu season

  1. System failure identification

    During flu season the triage queue stops behaving as a priority queue. High-acuity arrivals are seen promptly, but mid-acuity patients wait indefinitely as each new arrival outranks them. The queue is still ordered correctly at every individual step, which is what makes it hard to see.

  2. Data pipeline diagnostics

    Dashboards report median wait, which the promptly seen high-acuity cases hold down. Starvation in the middle of the queue lives entirely in the upper tail, where no metric looks. A metric built around the centre of a distribution cannot describe its tail.

  3. Predictive computational model

    Track the 90th percentile wait per acuity band, and the age of the oldest unseen patient in each band. A rising oldest-waiting age is the signal that queue ordering has stopped being fair. Both figures are computable from data the department already records at triage.

  4. Workflow integration

    Bind the oldest-waiting age to the existing queue display rather than a new alert. Charge nurses already read that screen, and adding a column changes behaviour where a pager would not. The measure has to sit where the reordering decision is actually made.

Back to the case study database