Previous Abstract | Next Abstract
Printable Version
October 03, 2020
10/3/2020 12:00:00 PM - 10/3/2020 1:00:00 PM
Room Virtual
Postoperative Critical Care Improves Mortality - Causal Inference Analysis Of A 248 Hospital Cohort
Tharusan Thevathasan, M.D., Danny J. Wong, M.D., Steve K. Harris, M.D.,Ph.D., Professor Ramani S. Moonesinghe, M.D.
University College London Hospital, London, United Kingdom
Disclosures: T. Thevathasan: None. D.J. Wong: None. S.K. Harris: None. P.S. Moonesinghe: Funded Research; Self; Research costs were supported by the NIAA, RCA, UCL, UCLH NIHR and UCL Surgical Outcomes Research Centre.
Background: Without an absolute indication for organ support, there is equipoise over who may benefit from postoperative critical care. Utilization of critical care is correlated with critical care bed availability which varies stochastically. This sets up a natural experiment where we can compare outcomes for those treated when critical care bed capacity is under strain or not, and use this to infer the causal effect of postoperative critical care on postoperative outcomes.

Objective: To investigate the causal effects of direct postoperative critical care versus surgical ward admission on patient morbidity and mortality by controlling for measured and unmeasured confounding.

Methods: We conducted a prospective, international, multicenter cohort study in 248 hospitals in the United Kingdom, Australia and New Zealand, recruiting patients over seven consecutive days in 2017. We included adult patients undergoing inpatient surgery without an absolute indication for postoperative critical care admission.

We first performed a risk-adjusted analysis using multivariable logistic regression with 29 demographic, preoperative and intraoperative predictor variables to account for observed confounding. We analyzed the association between postoperative admission to critical care versus surgical ward on patient morbidity using the Postoperative Morbidity Survey (POMS) on day 7, as well as on 30-day and 60-day mortality.

To make causal inferences, we accounted for observed and unobserved confounding by repeating the aforementioned analysis using an instrumental variable method with instruments on critical care bed strain (i.e., number of free beds and discharge-ready patients at the time of surgery).

Results: 21,935 patients were included in this study, of which 1,960 (8.9%) were admitted directly to critical care after surgery. 156 (0.7%) and 176 (0.8%) patients died within 30 and 60 days.

Accounting for observed confounding, critical care compared to ward admitted patients had an 109% increased risk (95% Confidence Interval, 1.96-2.23, P<0.001) for developing postoperative morbidities on day 7, as well as 91% (95% CI, 1.49-2.32, P<0.001) and 77% (95% CI, 1.38-2.17, P<0.001) higher risks for 30-day and 60-day mortality, respectively.

Accounting for observed and unobserved confounding, critical care admitted patients had a 77% (95% CI, 1.43-2.19, P<0.001) increased causal risk for having POMS-defined morbidity on postoperative day 7. However, 30-day and 60-day hospital mortality risks were 9% (95% CI, 0.81-1.0, P=0.06) and 10% (95% CI, 0.8-1.0, P=0.04) lower in critical care patients compared to ward patients (see Figure 1).

Of note, mortality benefits increased incrementally with critical care admission of higher risk surgical patients (see Figure 2): Critical care patients with Surgical Outcome Risk Tool-predicted 30-day mortality >9% had 35% lower 30-day mortality risk (95% CI, 0.27-1.04).

Conclusions: Although critical care admission immediately after surgery places patients at higher risk of short-term morbidity (e.g. due to invasive monitoring, new ICU-acquired infections or delirium), it confers longer-term mortality benefits (at 30 and 60 days).
Figure 1
Figure 2

Copyright © 2020 American Society of Anesthesiologists