When to stop: De-resuscitation in the intensive care unit era – A critical analysis of decision-making frameworks, prognostic tools and ethical considerations

End-of-life decision-making in intensive care units (ICUs) represents one of the most challenging aspects of critical care medicine, involving complex interactions between medical prognostication, ethical considerations and family dynamics. As technology advances and survival rates improve, the question of when to transition from curative to palliative care becomes increasingly complex. This narrative review synthesises current evidence on decision-making frameworks for withdrawal of life-sustaining treatments, evaluates prognostic tools and their limitations and examines ethical considerations guiding de-resuscitation decisions in the modern ICU era. Recent evidence demonstrates significant global variability in end-of-life practices, with withdrawal of life-sustaining treatment prevalence ranging from 0% to 84% amongst ICU deaths across different regions. Approximately 20% of deaths in developed countries occur in or shortly after an ICU stay, with most deaths preceded by decisions to limit or withdraw therapy. While prognostic tools, including biomarkers, neurophysiological testing and clinical assessments, have improved, accurate individual prognostication remains challenging, with approximately 50% of patients remaining prognostically indeterminate. Neurofilament light chain demonstrates excellent prognostic accuracy with area under the curve (AUC) values exceeding 0.93–0.98 at 24 h post-cardiac arrest. Artificial intelligence-based models show higher discriminating abilities than conventional severity scores with AUC values of 0.86–0.97. Emerging evidence highlights the risk of self-fulfilling prophecy bias in prognostication studies, where early withdrawal decisions may preclude potential recovery. Structured multimodal prognostication approaches, combined with enhanced family communication strategies, offer pathways to improved decision-making. However, significant implementation gaps remain, particularly in the resource-limited settings, necessitating culturally adapted frameworks that balance evidence-based practice with local healthcare realities.