New Delirium Prediction Tool Uses AI Algorithms To Predict Patients in ICU Who Are Most Likely at Risk of the Disorder

Researchers at Johns Hopkins University have created artificial intelligence (AI) algorithms that can detect early warning symptoms of delirium and forecast which patients are at high risk of delirium at any moment throughout an intensive care unit (ICU) stay.

According to the press release, delirium affects more than one-third of hospitalized patients and 80% of ICU patients at some time during their stay. Delirium can produce agitation, hallucination, disorientation, and inattention in patients, all of which can have a negative impact on patient outcomes.

What Is Delirium?

MSD Manual describes delirium as an acute, transient, usually reversible, and fluctuating attention disorder. It may affect people of any age, but it is more prevalent among older individuals. Approximately 10% of admitted older patients suffer from delirium and about 15% to 50% experience it during hospitalization.

More so, delirium is common among patients who just had surgery, as well as in nursing home residents and ICU patients. In younger people, delirium is usually due to drug use or a life-threatening systemic disorder.

Sometimes, delirium is confused with dementia even if both are separate disorders. Delirium and dementia are cognitive disorders but they differ in many aspects. Whereas delirium affects attention, is typically an acute illness, and is often reversible, dementia affects memory, which changes the brain and is generally irreversible.

During the COVID-19 pandemic, delirium has also become common among older patients and often occurs without any other typical symptoms of the infection. Doctors noted that the disorder is most likely to result in poor hospital outcomes and even death.

AI for Detecting Delirium in Patients

Johns Hopkins Medicine in collaboration with Johns Hopkins University engineering students showed how AI algorithms have the potential to detect the warning signs of delirium and can predict delirium-prone patients in the ICU.

In the study, titled "Predicting Intensive Care Delirium with Machine Learning: Model Development and External Validation" published in the journal Anesthesiology, researchers wrote that the newly designed AI algorithms by undergraduate and master's level engineering students in a precision medicine class a dataset that spans over 200,000 ICU stay at 208 hospitals in the country.

Experts said that anti-delirium interventions, including medications, care bundles, and therapy are effective in treating delirium. However, there are limited time and resources to prevent delirium among ICU patients.

As per AZO Robotics, the team developed two computerized models using the data to predict the risk of delirium. The first model analyzes patient data following admission that has information on the severity of the illness, age, other diagnoses, physiologic variables, and present medication.

Meanwhile, the second model tracks information over the hours and days of admission. It takes into account the patient's blood pressure, pulse rate, and temperature readings to monitor the patient's risk of delirium over the next 12 hours.

The first model showed that it had the potential to forecast which patients would have delirium within 24 hours at 78.5% of the time. But the second model performed better and was able to forecast accurately up to 90% of the time. Based on the results, the team is testing the model on historical patient data and is looking at a possible clinical trial to test the AI.

Check out more news and information on Delirium in Science Times.