Hallucinations and disturbed behaviour in the critically ill: incidence, patient characteristics, associations, trajectory, and outcomes.

Purpose: To use natural language processing (NLP) to study the incidence, characteristics, trajectory, associations, and outcomes of hallucinations and disturbed behaviour in intensive care unit (ICU) patients.
Methods: We used NLP to scan clinical progress notes of a large cohort of ICU patients to detect words indicating that a patient had experienced hallucinations. We also used NLP to detected disturbed behaviour during ICU stay. Moreover, we studied the use of antipsychotic medications in a nested cohort. Finally, we obtained the demographics, trajectory, associations, and outcome of these patients.
Results: We conducted a non-interventional, observational study of 7525 patients. We found that 625 (8.31%) had experienced hallucinations. Among these, 623 (99.7%) also had NLP-diagnosed behavioural disturbance (NLP-Dx-BD). In contrast, in patients without hallucinations, only 3274 (47.4%) were NLP-Dx-BD positive. Among the 2904 nested cohort patients with electronic medications data, 252 (8.7%) experienced hallucinations. Of these, 60 (23.8%) received medications compared with 147 (5.5%) (p < 0.001) patients without hallucinations. There was no difference on outcomes in patients with or without hallucination.
Conclusions: Hallucinations affect one in 12 ICU patients and are strongly associated with disturbed behaviour, and the use of antipsychotic medications. Hallucinations may represent another phenotype of critical illness associated neurocognitive dysfunction and require a dedicated research program.
Competing Interests: Declarations. Ethical approval and consent to participate: This study was approved by the Austin Hospital Human Research Ethics Committee (LNR/19/Austin/38), which waived the requirement for informed consent. Consent for publication: The authors hereby provide consent for this manuscript titled "Hallucinations and Disturbed Behaviour in the Critically Ill: Incidence, Patient Characteristics, Associations, Trajectory, and Outcomes" to be published. The authors confirm that they have read and approved the final version of the manuscript. They further affirm that the work described in the manuscript is original and has not been published elsewhere. Competing interest: The authors declare no competing interests.
(© 2025. The Author(s).)