Bias in research on neuroprognostication

The question who will (not) recover from a coma is crucial in intensive care medicine. One of the challenges in attempts to answer this question for individual patients is the occurrence of so-called ‘self-fulfilling prophecies’ (SFPs): predictions of poor outcome leading to withdrawal of life-sustaining treatment, thus realizing the prediction. If this happens, one will never know for sure whether the prediction was indeed correct and the decision to withdraw justified. 

In clinical practice SFPs are considered unavoidable, because the alternative implies that some patients are kept alive in conditions they would never have accepted. In a paper published in the November 2022 issue of the Journal of Medical Ethics Mayli Mertens (University of Copenhagen & University of Twente), Marianne Boenink (Ethics, IQ Healthcare, Radboudumc) and colleagues point out how such SFPs may also distort research in neuroprognostication. When assessing the accuracy of novel prognostic tests, predicted outcomes are compared with actual outcomes. However, actual poor outcomes partly result from cases in which life-sustaining treatment was withdrawn after neuroprognostication with existing tools. This implies one can never be sure that death would have occurred in these cases anyway. Using these outcomes as a standard to assess novel prognostic tools thus may result in persistent bias towards poor outcome predictions, as well as amplification of the tendency to withdraw treatment. The paper’s suggestions for dealing with this risk of bias are highly relevant for research aiming to improve predictive capacities in any healthcare setting where predictions may irreversibly impact treatment outcomes.

Publication

Mertens M, King OC, van Putten MJAM, Boenink M. Can we learn from hidden mistakes? Self-fulfilling prophecy and responsible neuroprognostic innovation. J Med Ethics. 2022 Nov;48(11):922-928.