What Changed When We Standardised Handover

Starmer's I-PASS paper is in every published patient-safety syllabus by now. The headline: 30% reduction in preventable adverse events when nine pediatric residency programs implemented a standardised handoff bundle (NEJM, 2014).

What gets less attention is what changed between the pilot and the follow-up.

The bundle, in one paragraph

I-PASS is a mnemonic for the five things a handover has to carry: Illness severity (is this patient stable, watcher, or unstable?), Patient summary, Action list, Situation awareness and contingency planning (if X happens overnight, do Y), and Synthesis by the receiver — the read-back, where the incoming doctor says the plan out loud and the outgoing one corrects it.

Nothing in that list is a surprise to anyone who has run a ward. The contribution of the paper was not the content. It was the claim that writing the content down in a fixed order, every time, changes outcomes — and then the evidence to support it.

What decayed, and what didn't

The I-PASS follow-up (Starmer et al., JAMA Pediatrics, 2017) showed that the gains held at one year, with continued reduction in errors. But it also showed that adherence to the structure decayed — by year-end, only about 65% of handovers were following the full I-PASS structure, down from 84% at the rollout's peak.

That decay is not laziness, and it is worth being precise about where it comes from. A structured handover costs time per patient. On a list of eight, the cost is invisible. On a list of twenty-eight at the end of a fourteen-hour shift, it is the difference between finishing handover at 8:10 and finishing at 8:40 — and the outgoing resident has been awake for nineteen hours.

The elements go in a predictable order, too. Contingency planning is the first to be compressed into "call me if anything happens." The receiver's synthesis is the first to be dropped entirely, because it is the only element that costs the incoming doctor time and it feels redundant when you have just been told the plan. The patient summary survives longest, because it is the part that feels like handover.

So the bundle erodes from the ends inward, and what is left after a year is a fluent, fast, unstructured summary — which is exactly what the ward was doing before the rollout.

The part that should have surprised people

The patient-safety gains held despite the decay. The implication is that the structure trained the residents in how to think about handover, even when they stopped using the bundle to the letter. The structure became internalised.

That is a genuinely encouraging finding and also a slightly dangerous one, because it is easy to read it as "the checklist was training wheels, and we can take them off." Internalised structure survives in a resident who did the rollout. It does not transfer to the intern who arrives in year three, having never seen the bundle enforced, and who learns handover by imitating the fluent unstructured version their senior now performs.

The enforcement trap

This matters for tool design. If a tool tries to enforce I-PASS rigidly — with required fields, validation, refusing to submit incomplete handovers — it creates the same friction that the paper-based protocol created. Residents adapt by routing around the tool.

Routing around is not hypothetical and it is not subtle. It looks like a required "contingency planning" field containing the word nil. It looks like a patient summary that reads "as before." It looks, most often, like the real handover happening in the corridor in ninety seconds and the tool being filled in afterwards, from memory, as an administrative chore — at which point the tool has stopped being a clinical instrument and become documentation.

Every mandatory field you add buys you a completed field and costs you a little of the trust that makes the tool worth opening.

Suggest, don't enforce

If a tool suggests the structure — by composing a draft in I-PASS format, that the resident can edit — the structure becomes the default but the resident retains the agency to adapt it for the case at hand. The I-PASS framework becomes a starting point rather than a checkbox exercise.

The economics flip. Under enforcement, structure costs the resident time and they pay it grudgingly. Under suggestion, structure saves them time, because the draft is already written — and the fastest path to a finished handover happens to be the structured one. The resident is not being compliant. They are being efficient, and the compliance is a side effect.

This is the design choice we made in Medows. The auto-composed handover follows I-PASS structure ("Now: …" and "Pending: …" for the most pressing two elements, with optional fields for illness severity and synthesis). It is composed from what actually happened on the shift — the orders placed, the results still outstanding, the vitals that moved — rather than from what the outgoing resident can recall at 8 p.m. The resident edits before sending. The structure is preserved without being enforced.

The pending-result line is the one that earns its place most often. A DKA patient whose 11 p.m. ABG had not returned by handover is the canonical case: nothing about it requires a hero, only a field that survives the shift boundary.

If you're running the rollout

1. Measure decay, not adoption. Adherence at month one tells you the teaching worked. Adherence at month twelve tells you whether the workflow did.
2. Watch the contingency and synthesis elements specifically. They fail first, and they are the two doing most of the safety work.
3. Count the cost per patient, then multiply by the list size. A protocol designed on a list of eight behaves differently on a list of twenty-eight.
4. Do not respond to decay with enforcement. You will get completed fields and worse handovers.

If you're rolling out I-PASS in your programme and finding the year-two decay starting to bite, the right intervention is not more enforcement. It is making the structured handover the path of least resistance.