Batch 01 · ER triage

Triage acuity drift during understaffed shifts

  1. System failure identification

    Under load, triage assignments drift. The same presentation is scored a level lower when the queue is long, compressing the acuity distribution and pushing sicker patients into slower pathways. The effect is strongest exactly when accurate prioritisation matters most.

  2. Data pipeline diagnostics

    Acuity is treated as an objective property of the patient, so nothing models it as an output of the assessing nurse under conditions. Drift is invisible because no metric compares scores against load. Without a load covariate, drift is indistinguishable from a genuinely less sick population.

  3. Predictive computational model

    Track acuity distribution per nurse per shift against that nurse’s own baseline, controlling for queue depth. A distribution that flattens as load rises is the measurable form of drift. Comparing a nurse against their own history controls for individual scoring style.

  4. Workflow integration

    Report drift to the nurse educator monthly as a calibration signal, never to the individual in real time. A live accuracy alert during a busy shift would add pressure to the condition causing it. Measurement aimed at the individual would be read as surveillance and resisted.

Back to the case study database