Transfer time to intensive care: no single threshold survives the data

Design

secondary analysis of a retrospective multicentre cohort, 2009-2020, logistic regression by transfer-time quintile adjusted for hospital and APACHE III, stratified by hospital type

Population

18,798 adults admitted directly from emergency departments to intensive care in twelve Dutch hospitals, across seven diagnostic groups

Primary outcome

hospital mortality by emergency department to intensive care transfer time

Effect

median transfer 1.9 hours (IQR 1.2-3.1); in cardiac arrest, academic hospitals OR 1.48 to 2.94 with longer time, non-academic hospitals OR under 1.0 throughout; no association in five of seven diagnostic groups

Time from emergency department arrival to intensive care admission is a favourite quality metric, on the assumption that shorter is better. This secondary analysis of a Dutch cohort covering 2009 to 2020 across twelve hospitals tested it in 18,798 adults admitted directly from the department, across seven diagnostic groups each with more than 1,500 patients, adjusting for hospital and APACHE III score.

Median transfer time was 1.9 hours. In out-of-hospital cardiac arrest the association depended entirely on the type of hospital: in academic centres longer transfer time tracked with higher mortality (OR 1.48 at 1.1-1.6 hours rising to 2.94 beyond 3.4 hours), while in non-academic teaching hospitals every quintile of longer time was associated with lower mortality. In non-operative trauma a positive association appeared in the pooled cohort (OR 1.90 beyond 3.4 hours) but vanished once stratified by hospital. Overdose, sepsis, pneumonia, respiratory failure and intracranial haemorrhage showed nothing at all.

The opposing directions in cardiac arrest are the point. Transfer time is not an exposure so much as a summary of everything else happening — a patient who dies quickly in the department never accrues a long transfer time, and a patient stabilised on the floor may legitimately wait. Departments should keep working on avoidable delay, but a blanket transfer-time target applied across diagnoses will measure case mix and organisation rather than quality of care.

• Do not adopt a single ED-to-ICU transfer time target across all diagnoses
• Immortal time bias explains part of this: the sickest patients cannot accrue long waits
• Focus audit on identifiable avoidable delay — bed availability, handover, transport
• No association was seen in sepsis, pneumonia, overdose, respiratory failure or intracranial haemorrhage
• Findings come from Dutch hospitals with short median waits of 1.9 hours; they do not describe a department boarding for many hours

The statistics, in plain English

When an association reverses direction between two subgroups of the same cohort, as it does here between academic and non-academic hospitals in cardiac arrest, the most likely explanation is confounding rather than a real biological difference. The trauma finding illustrates the same thing: an odds ratio of 1.90 in the pooled data that disappears on stratification by hospital was a between-hospital difference masquerading as a timing effect. Adjustment for APACHE III captures severity at admission to intensive care, not deterioration in the department, which is precisely the pathway of interest.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.