The Stroke Time Window Under Heavy Load

The 73-year-old man arrives in casualty at 9:48 p.m. with right-sided weakness and mild dysarthria. The family says he was fine at 8:15 p.m. when they left for dinner. They found him at 9:30 sitting in the chair, unable to lift his right arm.

Last-known-well: 8:15 p.m. Current time: 9:48 p.m. The window is 93 minutes from onset. The 4.5-hour thrombolysis window is wide open. CT is available. The medicine team is on the floor.

Everything about this case is favourable. That is what makes it worth writing about.

What happened

The patient is registered, triaged. The resident orders a CT. The CT is done at 10:32 p.m. The radiologist reads it at 10:51 p.m. — no haemorrhage. The resident calls neurology at 11:04 p.m. Neurology arrives at 11:32 p.m. The clinical exam is repeated. Thrombolysis is given at 12:08 a.m.

Total: 140 minutes from arrival to needle. The door-to-needle target is 60 minutes (AHA/ASA, multiple updates). The published Indian DTN times average 60–90 minutes in stroke centres with established pathways, and 120+ minutes in centres without (Pandian et al., Int J Stroke, 2017).

The difference between a 60-minute DTN and a 140-minute one is, for the average ischaemic stroke patient, roughly two million neurons. Time is brain. The published number is approximately 1.9 million neurons lost per minute in a typical large-vessel occlusion (Saver, Stroke, 2006).

Where the time went

• 12 minutes in registration / triage queue
• 44 minutes waiting for CT (single scanner, three competing requests)
• 19 minutes for radiology read
• 13 minutes for neurology to arrive
• 36 minutes for neurology decision and thrombolysis prep

No single step was bad medicine. Every step was reasonable. The cumulative effect was bad.

This is the characteristic shape of a systems failure in a busy hospital. Nobody was slow. Nobody made a wrong call. There is no name to put in an incident report, which is exactly why the 140 minutes will not be reviewed by anyone.

Serial where it could be parallel

Read the list again and notice that almost every interval waits on the one before it. Neurology was called after the radiology read, which came after the scan, which came after the queue. Thrombolysis preparation began after neurology arrived and examined.

Very little of that ordering is clinically required. Neurology could have been called at 9:50, on the clinical picture and the last-known-well time alone, and been in the department before the patient came off the scanner. The thrombolysis checklist — weight, anticoagulant history, recent surgery, BP control — could have been worked through during the 44 minutes of scanner queue, when the resident was doing nothing but waiting. Consent could have been discussed with the family in the same window, since the family was present and the conversation takes ten minutes.

None of that requires a resource the hospital does not have. It requires knowing, at 9:50 p.m., that this patient is on a clock — and everyone downstream knowing it too.

The number that keeps getting lost

The most consequential piece of information in this case is not on any monitor. It is "8:15 p.m." — a time reported once, by a family member, in the first two minutes.

It goes into the clerking note, usually in the third paragraph of a free-text history. From there it must survive to the radiologist, to the neurologist, to whoever finally makes the call. In practice it does not survive; it gets re-derived. Each new clinician asks the family again, and the family — now four hours into the worst evening of their year — answers slightly differently each time. "Around eight." "Just before dinner." "Maybe half seven?"

A time window is not a fact you can afford to re-establish three times. And the patients who lose most from this are the ones where it is genuinely ambiguous: unwitnessed onset, wake-up stroke, the patient found on the floor. Those are precisely the cases where a firmly recorded, first-told last-known-well is the difference between treating and not treating.

What a workspace can and cannot do

A workspace cannot fix the CT queue. It cannot make a second scanner appear or bring the neurologist closer. It should not pretend otherwise, and it is not a substitute for a code-stroke pathway with a named activation number — if your hospital has one, that pathway beats any of this.

What it can do is remove the delays that come from information not travelling:

• Pin last-known-well to the patient card, entered once, visible to everyone who opens the patient, with the elapsed time counting up. Not a number to be recalled — a clock on the screen.
• Attach the CT to the card the moment it is reported, and push it to the person waiting rather than waiting for them to check.
• Pre-stage the thrombolysis order so the consultant arrives to a pre-filled order pending sign-off, with the weight and the contraindication checklist already worked through.
• Make the clock visible on the list, not just in the patient. The resident holding twenty-eight patients needs to see, from the list view, that Bed 4 has a window closing.

Small things. Each shaves three minutes. Three minutes is six hundred thousand neurons.

The uncomfortable arithmetic

Fifteen minutes of the 140 in this case were pure information latency — time in which no scan was running, no clinician was examining, and nothing was happening except one person not yet knowing what another person already knew. That is the fraction software can actually claim.

Fifteen minutes is not a headline. It is also, on the published arithmetic, around twenty-eight million neurons for this one man, and it costs nothing to recover. Under heavy load, with a single scanner and a neurologist two floors away, the gains are not going to come from anywhere dramatic. They come from the queue not being the only thing that is serial.