Accuracy of predictions of survival at admission to the intensive care unit

https://doi.org/10.1053/jcrc.2001.21794Get rights and content

Purpose: The purpose of this study was to compare the accuracy of outcome predictions made on the day of intensive care unit (ICU) admission by critical care physicians, critical care fellows, and primary team physicians.

Patients and Methods: Fifty-nine consecutive patients admitted to a Medical-Surgical ICU were included in the study. Two ICU attending physicians and two critical care fellows, not involved in medical management, evaluated each new ICU patient at admission and after 48 to 72 hours. Altogether six ICU attendings and six fellows were involved in the study.

Each investigator separately assigned probability to each patient of being discharged alive from the ICU and the hospital. On the day of admission the primary service was also asked to estimate the likelihood of successful outcome. All values are expressed in percentiles. Statistical analysis was performed by a logistic regression procedure with a binary outcome. Data are presented as mean±SD.

Results: Fifty-nine patients were surveyed. Twenty-six (44%) patients died in the ICU, 3 (5%) died in the hospital wards, and 30 (51%) were discharged alive from the hospital.

ICU attendings most reliably and accurately estimated patient outcome on admission compared with critical care fellows and primary team physicians. ICU attendings were more consistent than ICU fellows at predicting outcome at 48 and 72 hours. Clinical predictions were better for patients at the extremes of disease severity, and the accuracy of predictions in these patients was highest. Accuracy was diminished in patients with moderate compromise of clinical status.

Conclusion: ICU attendings predicted most accurately and consistently the final outcome of patients, and ICU fellows estimated outcome more reliably than the primary service. For the most part, the primary service tended to overestimate the probability of favorable outcome of patients for whom ICU admission had been requested. Additionally, clinical accuracy of survival or mortality was best for those patients at the extremes of clinical compromise: this point seems to confirm the validity of using clinical judgement as a guide to restricting ICU resources for those severely compromised or mildly compromised. This study also indicates that predictions of outcome in critically ill patients made within days of admission are statistically valid but not sufficiently reliable to justify irrevocable clinical decisions at present.

References (10)

  • LemeshowS et al.

    Modeling the severity of illness of ICU patients: A systems update

    JAMA

    (1994)
  • CastellaX et al.

    A comparison of severity of illness scoring systems for intensive care unit patients: Results of a multicenter, multinational study. The European/North American Severity Study Group

    Crit Care Med

    (1995)
  • MorenoR et al.

    Outcome prediction in intensive care: Results of a prospective, multicentre, Portuguese study

    Intensive Care Med

    (1997)
  • BeckDH et al.

    Prediction of outcome from intensive care: A prospective cohort study comparing Acute Physiology and Chronic Health Evaluation II and III prognostic systems in a United Kingdom intensive care unit

    Crit Care Med

    (1997)
  • KruseJA et al.

    Comparison of clinical assessment with APACHE II for predicting mortality risk in patients admitted to a medical intensive care unit

    JAMA

    (1988)
There are more references available in the full text version of this article.

Cited by (28)

  • Prognosis and futility in neurosurgical emergencies: A review

    2020, Clinical Neurology and Neurosurgery
    Citation Excerpt :

    Though prognostic information is useful in these situations, it often lacks reliability and tends to be overly pessimistic [3,4]. Furthermore, the validity of prognostic disclosure varies based on provider experience with more experienced providers generally providing more accurate prognostic predictions [5,6]. From this discussion, it is clear that neurosurgeons, housestaff, and physician extenders must have a strong understanding of evidence-based prognostic data and medical futility so that they may correlate a patient’s condition to a prognosis or functional outcome in a non-biased manner.

  • The ability of intensive care unit physicians to estimate long-term prognosis in survivors of critical illness

    2018, Journal of Critical Care
    Citation Excerpt :

    However, adding these short stay patients to the study population would not affect the predictive performance as measured by the discrimination for survival (c-index 0.68; 95%CI 0.67–0.70) or for poor outcome including survival and HRQoL (c-index 0.63; 0.62–0.66). Other studies into the prognostic ability of ICU physicians have focused on the different levels of physician experience [6,29]. In these studies, attending intensivists consistently outperformed ICU fellows with regard to the predictive value of their outcome predictions.

  • Assessing trauma care provider judgement in the prediction of need for life-saving interventions

    2015, Injury
    Citation Excerpt :

    How human factors research results can best be made accessible to math-based systems developers remains an open question. The confrontation between the classically expert-judgement-based “art of medicine” point of view and that of “evidence-based medicine” is neither the focus nor the concern of the present study; however, the need to merge the two effectively in computer-based support systems is of concern across many health care disciplines, most particularly in critical care, to allow for better triage and distribution of resources [23-26]. Within this general debate, the clearest role for math-based systems development—computer modelling and instrumentation based on this work—is in improved data-gathering and decision-assist tools [27,28].

  • Development of a daily mortality probability prediction model from Intensive Care Unit patients using a discrete-time event history analysis

    2013, Computer Methods and Programs in Biomedicine
    Citation Excerpt :

    The APACHE and MPM scoring systems are global measures of illness severity and outcome prediction; however, the Multiple Organ Dysfunction and Sequential Organ Failure Assessment scores were initially designed to describe organ dysfunction or failure in critically ill patients. As various articles [5,12,22,30,46,49,51] have pointed out, severity scores are imperfect at predicting individual clinical chance of survival. Several reasons may explain this deficit [12,33,46,49].

View all citing articles on Scopus
View full text