Artificial Intelligence in Delirium Care: Predicting, Detecting, and Preventing Delirium in Older Hospitalized Adults

Delirium is a common and serious complication of hospitalisation in older adults, but it is frequently missed because symptoms fluctuate, hypoactive presentations are subtle, and dementia complicates recognition of acute change. Artificial intelligence (AI), including machine learning, natural language processing (NLP), deep learning, and large language models, may support earlier risk stratification, enhanced surveillance, documentation review, and delivery of personalised prevention. This narrative review examines current evidence for AI in predicting, detecting, and preventing delirium in older hospitalised adults. It considers model development, external and temporal validation, calibration, clinical utility, workflow integration, frailty, dementia, multilingual documentation, explainability, fairness, privacy, and accountability. Existing studies show promising discrimination in retrospective datasets, and NLP can identify delirium-related information that is absent from diagnostic codes. However, performance often declines during external validation, calibration is inconsistently reported, and few systems have been tested prospectively in routine care. Established multicomponent non-pharmacological interventions remain the strongest evidence-based approach to prevention, while evidence that AI-triggered recommendations improve patient outcomes is limited. Future research should therefore move beyond technical accuracy toward prospective, multicentre, human-in-the-loop evaluations that measure alert burden, clinician response, intervention fidelity, patient-centred outcomes, and subgroup equity.