Skip to main content

Archived Comments for: Delirium and mortality risk prediction: a story in evolution

Back to article

  1. Appropriateness of Cox models for Assessing Predictors

    Eduard Vasilevskis, VA - Tennessee Valley, Geriatric Research Education and Clincal Center and Vanderbilt University Medical Center, Nashivlle, TN

    4 November 2010

    Our previous letter to the editor(1) created some unanticipated errors in communication that prompted us to write this second letter with the hope of leaving the readership with factual knowledge of this confusing topic regarding the best methods by which to assess predictor relationships from cohort studies. Specifically, we want to address the statement by van den Boogaard and colleagues that Cox regression methods are not valid for assessing delirium as a predictor of mortality.(1) The authors support this opinion by inappropriately citing work by Steyerberg(2) and Cook.(3) We could not find any statement in these references that claimed that Cox models are inappropriate for assessing predictors. In fact, Steyerberg stated that Cox models are more than appropriate for modeling survival and considered them a “natural extension of the logistic model to the survival setting.”(2) Similarly, the manuscript by Cook highlighted the importance of using log likelihood approach based on regression models such as Cox regression as a more sensitive alternative method to c statistics.(3) Furthermore, the use of multivariable analysis has been repeatedly emphasized by field experts as appropriate and robust.(4,5)

    In delirium research, Cox models have been rigorously applied in assessing its independent effect on mortality, methods which were actually suggested and required by the statistical reviewers from multiple high impact journals.(6,7) Rather than solely incorporating delirium at baseline as a prognostic marker, we recommend evaluating the cumulative effects of delirium throughout the critical illness on long-term outcomes. Our empirical research have recently documented the appropriateness methodologically of using Cox regression modeling with delirium as a time-varying exposure variable to achieve scientifically valid results.(8)


    Eduard E. Vasilevskis, MD
    Jin H. Han, MD, MSc
    Ayumi Shintani, PhD, MPH
    Timothy D. Girard, MD, MSCI
    E. Wesley Ely, MD, MPH


    Reference List
    (1) Vasilevskis E, Han J, Shintani A, Girard T, Ely EW. Delirium and mortality risk prediction: A story in evolution. Crit Care 2010;14:449.
    (2) Steyerberg EW. Clinical prediction models: A practical approach to development, validation, and updating. Rotterdam: Springer Science+Business Media, LCC, 2009.
    (3) Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 2007;115:928-935.
    (4) Hlatky MA, Greenland P, Arnett DK et al. Criteria for evaluation of novel markers of cardiovascular risk: A scientific statement from the American Heart Association. Circulation 2009;119:2408-2416.
    (5) Moons KGM. Criteria for scientific evaluation of novel markers: A perspective. Clin Chem 2010;56:537-541.
    (6) Ely EW, Shintani A, Truman B et al. Delirium as a predictor of mortality in mechanically ventilated patients in the intensive care unit. JAMA 2004;291:1753-1762.
    (7) Pisani MA, Kong SYJ, Kasl SV, Murphy TE, Araujo KLB, Van Ness PH. Days of delirium are associated with 1-year mortality in an older intensive care unit population. Am J Respir Crit Care Med 2009;180:1092-1097.
    (8) Shintani AK, Girard TD, Eden SK, Arbogast PG, Moons KGM, Ely E. Immortal time bias in critical care research: Application of time-varying Cox regression for observational cohort studies. Crit Care Med 2009;37:2939-2945.

    Competing interests

    The authors declare that they have no competing interests.

Advertisement