Data-derived subtypes of delirium during critical illness

Area of Science:

• Critical Care Medicine
• Neuroscience
• Data Science

Background:

• Previous delirium research relied on psychomotor symptoms or risk factors for subtype identification.
• Machine learning has identified treatment-responsive subtypes in other acute illnesses but not delirium.
• Understanding delirium heterogeneity is crucial for targeted interventions.

Purpose of the Study:

• To identify distinct, data-derived subtypes of delirium in critically ill patients.
• To compare the clinical characteristics and outcomes of these identified subtypes.
• To explore the relationship between data-derived subtypes and previously defined delirium classifications.

Main Methods:

• Secondary analysis of a large, multicenter prospective cohort study of ICU patients with respiratory failure or shock.
• Unsupervised latent class modeling using pre-delirium data to identify subtypes.
• Comparison of demographics, clinical characteristics, and outcomes across identified delirium subtypes.

Main Results:

• A four-class model best fit the data, identifying distinct delirium subtypes.
• Subtype 2 exhibited higher mortality, while Subtype 4 had longer delirium and coma durations.
• All identified subtypes showed significant cognitive impairment at follow-up, irrespective of subtype.

Conclusions:

• Four data-derived delirium subtypes were identified, requiring validation in independent cohorts.
• These subtypes may represent differential treatment effects, informing future clinical trials.
• Findings can guide mechanistic research and the development of targeted delirium treatment strategies.