Background: The optimal perioperative use of intensive care unit (ICU) resources is not yet defined. We sought to determine the effect of ICU admission on perioperative (30 day) and long-term mortality.
Methods: This was an observational study of all surgical patients in Scotland during 2005-7 followed up until 2012. Patient, operative, and care process factors were extracted. The primary outcome was perioperative mortality; secondary outcomes were 1 and 4 yr mortality. Multivariable regression was used to construct a risk prediction model to allow standard-risk and high-risk groups to be defined based on deciles of predicted perioperative mortality risk, and to determine the effect of ICU admission (direct from theatre; indirect after initial care on ward; no ICU admission) on outcome adjusted for confounders.
Results: There were 572 598 patients included. The risk model performed well (c-index 0.92). Perioperative mortality occurred in 1125 (0.2%) in the standard-risk group (n=510 979) and in 3636 (6.4%) in the high-risk group (n=56 785). Patients with no ICU admission within 7 days of surgery had the lowest perioperative mortality (whole cohort 0.7%; high-risk cohort 5.3%). Indirect ICU admission was associated with a higher risk of perioperative mortality when compared with direct admission for the whole cohort (20.9 vs 12.1%; adjusted odds ratio 2.39, 95% confidence interval 2.01-2.84; P<0.01) and for high-risk patients (26.2 vs 17.8%; adjusted odds ratio 1.64, 95% confidence interval 1.37-1.96; P<0.01). Compared with direct ICU admission, indirectly admitted patients had higher severity of illness on admission, required more organ support, and had an increased duration of ICU stay.
Conclusions: Indirect ICU admission was associated with increased mortality and increased requirement for organ support.
Trial registration: UKCRN registry no. 15761.
Keywords: epidemiology; intensive care; surgery.
© The Author 2016. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: journals.permissions@oup.com.