Can doctors predict how likely a patient with delirium is to die?

Delirium in the intensive care unit (ICU) is linked to higher death rates. Doctors can use prediction models to estimate how likely a delirious patient is to die, but these tools are not perfect. They give a general risk level, not a certain outcome.

What the research says

A 2025 study developed prediction models specifically for delirious ICU patients ⁶. These models used information available before admission, such as frailty, number of medications, and ability to perform daily activities. The models had fair accuracy, with an area under the curve (AUC) of 0.74 for predicting death at 90 days and 1 year ⁶. An AUC of 0.74 means the model correctly distinguishes between patients who will die and those who will survive about 74% of the time — better than chance, but not highly precise.

Other research confirms that delirium itself is a marker of higher mortality risk. A meta-analysis found that deep sedation in the ICU is linked to both more delirium and higher death rates ³. Another study showed that acute kidney injury increases the risk of delirium and death ². These factors can be combined into prediction tools.

However, prediction models are not yet accurate enough to make individual decisions. The same study noted that while the models performed well overall, they should be used to support clinical judgment, not replace it ⁶. Doctors typically consider multiple factors — age, underlying illness, severity of delirium — when discussing prognosis with families.

What to ask your doctor

• What factors do you consider when estimating my loved one's risk of dying from delirium?
• Are there any prediction tools or scores you use for delirious patients in the ICU?
• How does the depth of sedation or use of restraints affect the outlook?
• What is the range of possible outcomes, and how certain are you about the prognosis?
• Can you explain how pre-existing health conditions or frailty influence the risk?