Elsevier

Resuscitation

Volume 84, Issue 3, March 2013, Pages 280-285
Resuscitation

Clinical paper
Common laboratory tests predict imminent death in ward patients

https://doi.org/10.1016/j.resuscitation.2012.07.025Get rights and content

Abstract

Objective

To estimate the ability of commonly measured laboratory variables to predict an imminent (within the same or next calendar day) death in ward patients.

Design

Retrospective observational study.

Setting

Two university affiliated hospitals.

Patients

Cohort of 42,701 patients admitted for more than 24 hours and external validation cohort of 13,137 patients admitted for more than 24 hours.

Intervention

We linked commonly measured laboratory tests with event databases and assessed the ability of each laboratory variable or combination of variables together with patient age to predict imminent death.

Measurements and main results

In the inception teaching hospital, we studied 418,897 batches of tests in 42,701 patients (males 55%; average age 65.8 ± 17.6 years), for a total of >2.5 million individual measurements. Among these patients, there were 1596 deaths. Multivariable logistic modelling achieved an AUC–ROC of 0.87 (95% CI: 0.85–0.89) for the prediction of imminent death. Using an additional 105,074 batches from a cohort of 13,137 patients from a second teaching hospital, the multivariate model achieved an AUC–ROC of 0.88 (95% CI: 0.85–0.90).

Conclusions

Commonly performed laboratory tests can help predict imminent death in ward patients. Prospective investigations of the clinical utility of such predictions appear justified.

Introduction

Among hospital patients, serious adverse events, including death, are relatively common.1, 2, 3 Many such events and deaths appear preventable1, 2, 3, 4, 5, 6 because they are preceded by physiological and clinical deterioration. Multiple attempts have been made to avoid such deaths5, 6, 7, 8, 9, 10 including the introduction of rapid response teams (RRTs) systems to respond to acute physiological deterioration.9 Such systems, however, are problematic because the identification of patients at risk is subject to the accuracy of observations,9 judgment about the patient's condition,8 diligence in the measurements of vital signs,8, 9, 10, 11 vigilance during the entire 24 hour period,11 and, finally, willingness to call for help in a timely fashion.12, 13, 14, 15 These shortcomings contribute to incorrect non-activation or delayed activation of an appropriate response to patient deterioration.14, 15, 16 Non-activation and delayed activation are, in turn, associated with increased mortality.14, 15, 17, 18 These recurrent observations suggest the need for a better approach so that an appropriate response can occur, or where necessary, earlier end of life discussion can take place and unnecessary and unwanted chest compression can be avoided.

A system based on objective data electronically collected as part of standard care might assist in the identification of high risk patients. Such data already exist in essentially all hospitals of developed countries in the form of common laboratory tests (e.g. biochemistry, hematology, arterial blood gases). In association with clinical information, they have already been found helpful in estimating risk of the death after ICU admission19, 20 and in cohorts of ward patients.21, 22 It seems, therefore, physiologically plausible and, by analogy, logical, that laboratory data might similarly help identify other hospital patients at risk of imminent death.

Accordingly, we performed a study to determine whether common laboratory variables might serve as useful predictors of imminent death in ward patients. In particular, we hypothesised that commonly performed laboratory tests, when used in combination, might have a fair to good ability to predict the patient's death either on the same day or during the following calendar day (imminent death).

Section snippets

Methods

This study of laboratory data and their link with deaths is part of a continuing audit of emergency activity and mortality approved by the Austin and Alfred Hospital Human Research Ethics Committees, which waived the need for informed consent for this specific project.

Information about the date of all deaths at the Austin and Alfred Hospital is collected in a specific dedicated administrative electronic database. Similarly, the central laboratory of the hospitals electronically stores all

Results

Having selected nine chosen variables, we studied their values across 418,897 batches of tests in 42,701 ward patients (males: 55%; average age: 65.8 ± 17.6 years) for a total of >2.5 million individual measurements.

Among study patients, there were 1596 patients who died with 3064 batches taken on the day prior to or the day of death. The test results measured during the day of or the day before a death were compared with the same test results measured during other periods as shown in Table 1 and

Statement of key findings

We conducted a study of more than 40,000 ward patients from a tertiary hospital admitted for >24 hours to test whether commonly measured laboratory variables would help predict imminent death. We found combinations of such tests had fair to good predictive values. We further found that, using combinations of tests and specific thresholds, we could define potentially clinically useful levels of specificity and/or sensitivity for such predictions. Finally, we confirmed the potential external

Conclusions

We conducted a study of more than 2.5 million single measurements and more than 400,000 batches of laboratory tests in more than 40,000 hospital ward patients and found that several individual laboratory tests as well as combinations of tests had fair to good predictive value in identifying patients at risk of death within the same or next day. We confirmed these findings in a separate cohort from another teaching hospital. These findings provide proof-of-concept evidence that laboratory tests

Financial support

This project was partially supported by the Cooperative Research Centres Programme for Smart Services funded by the Australian Government.

Conflict of interest statement

We further warrant that all authors have read and approved the manuscript and that there are no conflicts of interest in relation to this paper.

Acknowledgments

The authors wish to acknowledge the work of Mr. Lawrence Hudson and Mr. Christopher MacManus of the Health Informatics Department at Alfred Health in assisting with obtaining ethics approval for the project, and with data extraction and management.

References (56)

  • S.W. Waldo et al.

    Pro-B-type natriuretic peptide levels in acute decompensated heart failure

    J Am Coll Cardiol

    (2008)
  • M. Haase et al.

    Accuracy of neutrophil gelatinase-associated lipocalin (NGAL) in diagnosis and prognosis in acute kidney injury: a systematic review and meta-analysis

    Am J Kidney Dis

    (2009)
  • C.M. Walshe et al.

    Urine glutathione S-transferase as an early marker for renal dysfunction in patients admitted to intensive care with sepsis

    Crit Care Resusc

    (2009)
  • S.M. Bagshaw

    Subclinical acute kidney injury: a novel biomarker-defined syndrome

    Crit Care Resusc

    (2011)
  • J. Vetro et al.

    Antecedents to cardiac arrests in a hospital equipped with a medical emergency team

    Crit Care Resusc

    (2011)
  • R.M. Trinkle et al.

    Medical reviews before cardiac arrest, medical emergency call or unanticipated intensive care unit admission: their nature and impact on patient outcome

    Crit Care Resusc

    (2011)
  • C.I. Knott et al.

    A retrospective cohort study of the effect of medical emergency teams on documentation of advance care directives

    Crit Care Resusc

    (2011)
  • S. Jones et al.

    Bedside electronic capture of clinical observations and automated clinical alerts to improve compliance with an early warning score protocol

    Crit Care Resusc

    (2011)
  • S.J. Warrillow et al.

    Major surgery in Victoria and the United States: a comparison of hospital mortality in older patients

    Crit Care Resusc

    (2008)
  • T.A. Brennan et al.

    Incidence of adverse events and negligence in hospitalized patients: results of the Harvard medical practice study I

    Qual Saf Health Care

    (2004)
  • G.R. Baker et al.

    The Canadian adverse events study: the incidence of adverse events among hospital patients in Canada

    CMAJ

    (2004)
  • C. Vincent et al.

    Adverse events in British hospitals: preliminary retrospective record review

    BMJ

    (2001)
  • M.D. Buist et al.

    Recognising clinical instability in hospital patients before cardiac arrest or unplanned admission to intensive care. A pilot study in a tertiary-care hospital

    MJA

    (1999)
  • R. Bellomo et al.

    A prospective before-and-after trial of a medical emergency team

    MJA

    (2003)
  • M. Buist et al.

    Recognizing clinical six year audit of cardiac arrests and medical emergency team calls in an Australian outer metropolitan teaching hospital

    BMJ

    (2007)
  • A. Lee et al.

    The medical emergency team

    Anaesth Intensive Care

    (1995)
  • D.A. Jones et al.

    Current concepts: rapid response teams

    N Engl J Med

    (2011)
  • M.A. DeVita et al.

    Findings of the first consensus conference on medical emergency teams

    Crit Care Med

    (2006)
  • Cited by (30)

    • A new simplified model for predicting 30-day mortality in older medical emergency department patients: The rise up score

      2020, European Journal of Internal Medicine
      Citation Excerpt :

      However, one ICU study showed that a combination of more than one abnormal vital signs predicts mortality better than separate vital signs [32]. Age and the four laboratory tests in our model were identified as predictors of mortality in other studies as well [7,8,12,33–35]. Albumin, BUN, LDH and bilirubin are markers of severe diseases, such as sepsis, malignancy, renal failure, upper gastrointestinal bleeding, ischemia or shock, haemolysis and liver- and/or bile duct disease.

    • Laboratory alerts to guide early intensive care team review in surgical patients: A feasibility, safety, and efficacy pilot randomized controlled trial

      2018, Resuscitation
      Citation Excerpt :

      Together with clinical information, they have already been found to assist in estimating mortality risk after ICU admission [12,13]. It seems, therefore, physiologically reasonable and, by analogy, clinically logical, that such laboratory-based information might similarly help identify hospital ward patients at higher risk of complications [14,15]. Accordingly, we performed a pilot randomized controlled trial of patients in surgical wards.

    • Comparison of the Between the Flags calling criteria to the MEWS, NEWS and the electronic Cardiac Arrest Risk Triage (eCART) score for the identification of deteriorating ward patients

      2018, Resuscitation
      Citation Excerpt :

      This suggests that the use of eCART could improve the stratification of patients enabling more immediate action for those known to be at higher risk to prevent further deterioration and the reduction of ‘unnecessary’ Rapid Responses calls as compared to BTF. The combination of multiple laboratory tests and vital sign observations to develop predictive models has been investigated in a number of previous studies dealing with ward and emergency department patients [28–30]. We believe that moving towards the electronic medical record will overcome previously described issues with aggregated scoring systems [31] and allow an electronic algorithm to increase the positive predictive value of the afferent arm and assist in the risk stratification of patients so that resources can be used appropriately.

    • Development and validation of a decision tree early warning score based on routine laboratory test results for the discrimination of hospital mortality in emergency medical admissions

      2013, Resuscitation
      Citation Excerpt :

      Whilst the curves are very similar in shape, the position of LDT-EWS scores differs for males and females – a LDT-EWS score = 4 for a female (35.2% calls triggered; 75.3% deaths visited) is in an almost identical position on its curve to a LDT-EWS score = 5 for a male (36.7% calls triggered; 75.8% deaths visited). There is increasingly interest in the role of laboratory tests as predictors of patient outcome in recent years.7–28 Many of these authors have described the development of complex prediction models based on logistic regression techniques, but all use one or more variables, such as combinations of age, symptoms and physiology, in addition to routinely measured laboratory test results.7–28

    View all citing articles on Scopus

    A Spanish translated version of the summary of this article appears as Appendix in the final online version at http://dx.doi.org/10.1016/j.resuscitation.2012.07.025

    View full text