Performance of a guideline-recommended algorithm for prognostication of poor neurological outcome after cardiac arrest

Abstract

Purpose: To assess the performance of a 4-step algorithm for neurological prognostication after cardiac arrest recommended by the European Resuscitation Council (ERC) and the European Society of Intensive Care Medicine (ESICM). Methods: Retrospective descriptive analysis with data from the Target Temperature Management (TTM) Trial. Associations between predicted and actual neurological outcome were investigated for each step of the algorithm with results from clinical neurological examinations, neuroradiology (CT or MRI), neurophysiology (EEG and SSEP) and serum neuron-specific enolase. Patients examined with Glasgow Coma Scale Motor Score (GCS-M) on day 4 (72-96 h) post-arrest and available 6-month outcome were included. Poor outcome was defined as Cerebral Performance Category 3-5. Variations of the ERC/ESICM algorithm were explored within the same cohort. Results: The ERC/ESICM algorithm identified poor outcome patients with 38.7% sensitivity (95% CI 33.1-44.7) and 100% specificity (95% CI 98.8-100) in a cohort of 585 patients. An alternative cutoff for serum neuron-specific enolase, an alternative EEG-classification and variations of the GCS-M had minor effects on the sensitivity without causing false positive predictions. The highest overall sensitivity, 42.5% (95% CI 36.7-48.5), was achieved when prognosticating patients irrespective of GCS-M score, with 100% specificity (95% CI 98.8-100) remaining. Conclusion: The ERC/ESICM algorithm and all exploratory multimodal variations thereof investigated in this study predicted poor outcome without false positive predictions and with sensitivities 34.6-42.5%. Our results should be validated prospectively, preferably in patients where withdrawal of life-sustaining therapy is uncommon to exclude any confounding from self-fulfilling prophecies.

Predicting neurological outcome after out-of-hospital cardiac arrest with cumulative information; development and internal validation of an artificial neural network algorithm

“…The latest treatment guidelines for comatose cardiac arrest survivors recommend multimodal prognostication to be performed at least 72 h after ROSC. Moseby-Knappe et al found that the current 4-step algorithm for neurological prognostication after cardiac arrest recommended by the European Resuscitation Council (ERC) and the European Society of Intensive Care Medicine (ESICM) identified patients with poor outcome with a 39% sensitivity and 100% specificity [34]. These guidelines are based on the consensus opinion of leading experts in the field.…”

Investigating Novel and Conventional Biomarkers for Post-resuscitation Prognosis: The Role of Cytokeratin-18, Neuron-specific Enolase, and Lactate

“…Prognostication of neurological outcome after cardiac arrest is based on clinical neurological examination, electrophysiological investigations, neuroimaging, and biomarkers. 9 However, in our study, we only focused on biomarkers in the peripheral blood and we did not evaluate the mentioned examinations or tools for prognostication. Most of our patients obtained the best CPC of 3, 4, or 5, which are mentioned as poor neurological outcomes in most recent studies contrary to our categorisation (CPC 4-5 for poor and 1-3 for good outcome).…”

“…Early predictors of outcome that would support clinical decision-making are required to avoid costly and intensive resources in cases of futility, to guide level-of-care decisions and goals-of-care conversations with family members 7 . The current guidelines and studies are mostly dealing with neurological prognostication in patients remaining comatose after CA 8,9 , while high proportion of patients (especially after IHCA) reach acceptable neurological function but suffer from multiple organ failure, which may lead to death independently of neurological status 10 .…”

On predictions in critical care: The individual prognostication fallacy in elderly patients

“…Depending on the impact of a specific outcome or the cost of prediction errors, the performance of a statistical model can be unsatisfactory or unacceptable for some people or subgroups [ 26 ]. For example, in a study on prognostication of poor neurological outcome after cardiac arrest that aimed at preventing false positives predictions, the false negative rate reached approximately 30% [ 27 ]. This means that a substantial number of patients will be treated in vain.…”