Decision-making for ICU admission: is there a place for AI? Exploring and understanding meaning, experiences, and perspectives

Abstract

Intensive care unit (ICU) admission (or refusal) decisions are intricate, relying on the assessment of patient benefit by the intensivist, sole decision-maker, without clear guidelines. As artificial intelligence (AI) rapidly develops predictive applications in healthcare, it could serve as a valuable forecasting tool to aid intensivist decision-making. This study aimed to explore whether intensivists are in demand of an “intelligent” decision-support tool for admissions and, if so, their specific expectations. A cross-sectional qualitative study was conducted. Ten French intensivists were interviewed between February and May 2024 and their contributions were analyzed using grounded theory method. Participants varied in experience (1–35 years), region of practice (hyper-urbanized, n = 5, or not) and familiarity with AI (n = 4). Nine and a half hours of interviews revealed that the ICU admission process is a complex undertaking, depending on many criteria and often involving unconventional decision pathways. Intensivists’ decision-making was inherently subjective, deeply tied to moral values aiming to preserve their medical integrity, though sometimes manifesting as biases that result in unfairness. Participants frequently voiced discomfort with the uncertainty in “grey areas”, that is to say poorly documented and non-extreme situations. While intensivists expected AI to counteract their biases and augment their knowledge, they also feared a loss of the humanity in decision-making to the expense of the technical element. Exploring experiences of intensivists regarding ICU admission revealed that the decision is less a matter of protocol than of nuanced, subjective assessment. AI likely has a role in supporting ICU admission or refusal decisions, provided its limitations as a purely informative statistical tool are acknowledged.

Ähnliche Arbeiten