From detection to decision: a conceptual framework of uncertainty migration in disorders of consciousness

Accurate assessment of consciousness in patients with disorders of consciousness (DoC) remains a longstanding challenge in neurocritical care and neurorehabilitation. Although standardized behavioral assessments continue to serve as the clinical foundation for diagnosis, their reliance on overt behavioral responses may underestimate residual consciousness in a subset of patients. Recent advances in multimodal neuroimaging and neurophysiological techniques have substantially expanded the capacity to detect covert consciousness, reducing some of the inherent limitations of behavioral assessment and facilitating the identification of patients with preserved cognitive processing despite an absence of observable responses. However, these technological advances have not eliminated uncertainty in DoC assessment. Instead, uncertainty may be reconfigured as assessment progresses from behavioral observation to neural measurement, the inference of conscious awareness, prognostic evaluation, and clinical decision-making. Whether neural activity can reliably reflect subjective conscious experience, how neurophysiological evidence should inform prognostic evaluation and therapeutic decisions, and the emerging risks introduced by artificial intelligence-assisted assessment all indicate that uncertainty in DoC has evolved beyond a purely technical challenge into a broader problem of interpretation and clinical decision-making. Artificial intelligence (AI) further intersects with these uncertainties across multiple stages by contributing to signal processing, multimodal integration, consciousness inference, prognostic prediction, and clinical decision support. In this review, we propose the conceptual framework of "uncertainty migration" to characterize how uncertainty may change in form, persist across epistemic transitions, be redistributed between stages or actors, and accumulate through interactions between residual and newly introduced uncertainties across five interconnected domains of DoC assessment: behavioral assessment, neural measurement, consciousness inference, prognostic evaluation and clinical decision-making. These processes of transformation, propagation, redistribution, and accumulation are neither mutually exclusive nor necessarily linear, as discordant multimodal findings and longitudinal changes may prompt the reassessment of earlier evidence. We argue that multimodal approaches fundamentally change how consciousness-related evidence is acquired, and may reduce particular sources of uncertainty without eliminating the inherently inferential nature of consciousness assessment. Consequently, future advances in the DoC field should extend beyond improving diagnostic accuracy toward developing comprehensive strategies for uncertainty management that emphasize rigorous interpretation of evidence, longitudinal reassessment, multimodal evidence integration, transparent AI governance,and shared decision-making. Such an approach may better support cautious, transparent, and patient-centered clinical decisions in the face of persistent uncertainty.