Related Experiment Videos
Clinician predictions of intensive care unit mortality
Graeme Rocker1, Deborah Cook, Peter Sjokvist
1Department of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada.
Insights
Physician predictions of less than 10% intensive care unit survival strongly influence life support decisions. These predictions are more powerful indicators of mortality than illness severity or organ dysfunction in critically ill patients.
Area of Science:
- Critical Care Medicine
- Prognostic Research
- Medical Decision-Making
Background:
- Accurate prediction of outcomes is crucial for end-of-life discussions in intensive care units (ICUs).
- Evaluating the accuracy and impact of clinical survival predictions is essential for patient care.
Purpose of the Study:
- To assess the accuracy of physician survival predictions in critically ill, mechanically ventilated patients.
- To determine the consequences of these predictions on life support decisions.
Main Methods:
- Prospective cohort study involving 851 mechanically ventilated adult patients across 15 tertiary care centers.
- Daily assessment of Multiple Organ Dysfunction Score (MODS), life support use, patient preferences, and physician/nurse survival probability estimates.
- Analysis of baseline characteristics (APACHE II score) and daily clinical factors as predictors of ICU mortality.
Main Results:
- Physician estimates of <10% ICU survival were recorded for 40.1% of patients.
- These low survival predictions were associated with increased withdrawal of mechanical ventilation, inotropes/vasopressors, and dialysis (p < .001).
- Physician predictions of <10% survival were more impactful than illness severity (APACHE II, MODS) or organ dysfunction in predicting mortality.
Conclusions:
- Physician estimates of <10% ICU survival are significant predictors of subsequent life support limitation.
- These predictions are more powerful than illness severity or evolving organ dysfunction in forecasting ICU mortality.
- The findings highlight the substantial influence of physician prognostic assessments on clinical management and patient outcomes.
Objective:
Predicting outcomes for critically ill patients is an important aspect of discussions with families in the intensive care unit. Our objective was to evaluate clinical intensive care unit survival predictions and their consequences for mechanically ventilated patients.
Design:
Prospective cohort study.
Setting:
Fifteen tertiary care centers.
Patients:
Consecutive mechanically ventilated patients > or = 18 yrs of age with expected intensive care unit stay > or = 72 hrs.
Interventions:
We recorded baseline characteristics at intensive care unit admission. Daily we measured multiple organ dysfunction score (MODS), use of advanced life support, patient preferences for life support, and intensivist and bedside intensive care unit nurse estimated probability of intensive care unit survival.
Measurements And Main Results:
The 851 patients were aged 61.2 (+/- 17.6, mean + SD) yrs with an Acute Physiology and Chronic Health Evaluation (APACHE) II score of 21.7 (+/- 8.6). Three hundred and four patients (35.7%) died in the intensive care unit, and 341 (40.1%) were assessed by a physician at least once to have a < 10% intensive care unit survival probability. Independent predictors of intensive care unit mortality were baseline APACHE II score (hazard ratio, 1.16; 95% confidence interval, 1.08-1.24, for a 5-point increase) and daily factors such as MODS (hazard ratio, 2.50; 95% confidence interval, 2.06-3.04, for a 5-point increase), use of inotropes or vasopressors (hazard ratio, 2.14; 95% confidence interval, 1.66-2.77), dialysis (hazard ratio, 0.51; 95% confidence interval, 0.35-0.75), patient preference to limit life support (hazard ratio, 10.22; 95% confidence interval, 7.38-14.16), and physician but not nurse prediction of < 10% survival. The impact of physician estimates of < 10% intensive care unit survival was greater for patients without vs. those with preferences to limit life support (p < .001) and for patients with less vs. more severe organ dysfunction (p < .001). Mechanical ventilation, inotropes or vasopressors, and dialysis were withdrawn more often when physicians predicted < 10% probability of intensive care unit survival (all ps < .001).
Conclusions:
Physician estimates of intensive care unit survival < 10% are associated with subsequent life support limitation and more powerfully predict intensive care unit mortality than illness severity, evolving or resolving organ dysfunction, and use of inotropes or vasopressors.