Spreading a sepsis screening pathway across four emergency departments

Not yet, and probably not as designed. A hard stop is a control on the record, not on the clinical judgement that is currently failing.

Read the two behaviours separately. Nurses are not failing to screen because they disagree with it; the field is three clicks in and triage queues are real. A hard stop will not create time at triage - it will create a queue, which is exactly nursing leadership's concern, and queues in emergency departments cause their own harm. Doctors are overriding because they believe they can identify sepsis themselves. A hard stop would force them to click through, not to think differently; it converts a clinical judgement problem into a data-entry problem.

My recommendation:

1. Remove the friction now, without waiting for the release: a single-click screening shortcut reachable from the triage view, tested in one department within two weeks.
2. Make overrides visible and coded - reason required, reviewed weekly with clinical leads, fed back by specialty. That changes behaviour through scrutiny rather than blocking.
3. Reserve the hard stop for the highest-risk step, pilot it in one department with peak-hour queuing measured, and set a decision date and a rollback threshold in advance.
4. Put a safety net outside the field: an observations-triggered alert that fires whether or not screening was completed.

The medical director is right that the current position is unsafe. The instrument is wrong.

The gap is too large and too consistent to be explained by training volume, and 'it worked in the pilot' is not evidence that it will work elsewhere. What spread is implementation fidelity, not the pathway.

Compare pilot and spread sites on the things that actually drive compliance: who owned it locally, whether compliance was fed back visibly, whether a champion was on each shift, whether the workflow was adjusted for local layout and staffing, and case mix. My working hypothesis - to be tested, not assumed - is that the pilot benefited from proximity to the design team and weekly feedback, both of which disappeared at spread.

I would run a fidelity check at each site: is the screening step reachable, is feedback posted, is there a named clinical lead, is there a safety huddle. Then correlate fidelity with compliance. Transfer the mechanisms, not the training slides: local clinical ownership, visible weekly data by department, peer feedback on overrides, and a fast route for staff to report a workflow problem that actually gets fixed.

Two cautions. Do not average the three sites; 47% and 66% may have different causes. And do not set a target before you know the ceiling - 81% is the best observed, not the achievable maximum.

Safety response first, change programme second. In the next ten days I would not push the rollout harder, because it would look like the organisation is protecting the pathway rather than the patients.

Days 1–3:

• Review both incidents and establish whether screening was omitted, overridden, or unreachable at the time. These are three different failures with three different fixes.
• Check whether the pathway was usable on those shifts - staffing, system access, downtime.
• Escalate to the patient safety committee and, through the medical director, be explicit about interim risk.

Days 4–10:

• Put an interim safety net in place that does not depend on the buried field - for example an observations-triggered prompt at triage, or a nurse-led check before transfer out of triage.
• Brief the departments honestly: what happened, what is changing, what to do meanwhile. Vague reassurance damages credibility more than bad news.
• Agree a single named owner for the pathway with authority to change the workflow.

Then return to spread. The incidents are also data: they show where the process fails under pressure, which is more useful than another compliance report.