Johns Hopkins AI models predict risk of ICU delirium

According to a new study, two dynamic analytic models developed at Johns Hopkins University predicted patients prone to delirium when tested on two datasets drawn from 100,000 hospital stays in the primary care unit. particular of a hospital in Boston.


Delirium – sudden episodes of confusion, inattention, paranoia, agitation and hallucinations – can put patients at high risk for prolonged hospitalization, dementia and future death. By predicting delirium, clinicians are alerted to possible countermeasures that can minimize adverse outcomes, according to the premise of the artificial intelligence study published in the journal Neurology. Journal of Anesthesiology.

“For many of these physiological transitions, we think there are early warning signs that may not be obvious to the clinician but can be detected using various types of supportive model analysis. which we used here,” said Dr. Robert Stevens, associate professor of anesthesia and intensive care medicine at Johns Hopkins University School of Medicine and lead author of the study, as publish new findings.

According to the online summary, the primary goal is to predict ICU delirium by applying machine learning to data routinely collected in electronic health records.

EHR data contains signatures that are associated with risk of delirium, according to Kirby Gong, a graduate student in the Johns Hopkins Department of Biomedical Engineering and lead author of the study.

Using publicly available ICU data, the researchers developed two predictive models.

A static model takes a snapshot of patient data – such as age, disease severity, other diagnoses, physiological variables, and current medications – immediately after admission to predict risk. delirium at any point during the hospital stay.

When tested with a different ICU dataset from a Boston hospital, the single snapshot was able to predict which patients would experience delirium 78.5% of the time.

The second dynamic model tracks hourly and daily information, including repeated blood pressure, pulse and temperature readings, and continuously updates the risk of delirium over the next 12 hours. When tested with Boston ICU data, it predicted patients to be prone to delirium up to 90% of the time.

Stevens is currently testing models on Johns Hopkins Medicine ICU data and plans to design a clinical trial to test the use of algorithms and how they can shape clinical care in The new patient was admitted to intensive care, according to the announcement.


AI is used in precision medicine to accelerate translational and clinical research to improve disease prediction and treatment.

With machine learning, the researchers found it was possible to predict all sorts of patient outcomes, from the impact of drug dosages to how changes in the skin’s barrier can manifest in allergies and autoimmune conditions. immunity, etc.

According to healthcare leaders, organizations like JHU and UPMC Enterprises are looking at AI and precision medicine to improve healthcare outcomes and streamline costs.

Dr Matthias J, Kleinz, senior vice president and head of translational science at UPMC Enterprises, said: “The tangible benefits are streamlined clinical workflows, patient outcomes, and patient outcomes. improved staffing and the ability to optimize resource allocation and reduce long-term care costs.” Healthcare IT News last year.


“Being able to distinguish between patients at low and high risk of delirium is extremely important in the ICU because it allows us to devote more resources to interventions in high-risk populations,” Stevens said. know in the statement.

Andrea Fox is the senior editor of Healthcare IT News.
Email: [email protected]

Healthcare IT News is a publication of HIMSS.


News7g: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button