Could Machine Learning Make Intensive Care More Efficient?
Learn about how machine learning could make ICU care more efficient.
Join the DZone community and get the full member experience.Join For Free
As the very name suggests, intensive care units are one of the highest pressured environments in healthcare, where every decision can have drastic consequences. It's an environment where each new test can yield vital clinical information but equally adds to the already steep costs associated with keeping the patient alive.
A team from Princeton University has developed an AI-based system that they believe will help clinicians assess whether that extra lab test will deliver valuable information or not.
The work, which was documented in a recently published paper, collected data from around 6,000 patients to try and determine whether four blood tests would yield important information for diagnosing kidney failure or sepsis.
The algorithm utilized a reward function that uses data on the value of past tests to help determine the value of subsequent tests. So, if previous tests were clinically useful, the algorithm determines the probability to be higher than what the subsequent tests will be. It also took into account the financial cost involved in the test, plus the risk for the patient.
Suffice to say, in an intensive care unit, time is of the essence, so it's vital that any decision support tool can provide not only accurate but swift recommendations. For the research, the team used the computational grunt available at Princeton, but it would obviously need to be workable in a less powerful environment found in most hospitals.
Nonetheless, the research showed that their system was able to provide better recommendations than clinicians themselves achieved in the clinical setting. Indeed, had the doctors used the algorithm, it would have reduced the number of blood tests by around 44 percent, while also shaving hours from the decision-making process.
"With the lab test ordering policy that this method developed, we were able to order labs to determine that the patient's health had degraded enough to need treatment, on average, four hours before the clinician actually initiated treatment based on clinician ordered labs," the researchers explain.
The next step is for the team to try and deliver their decision support tool into the hospital. They're working with data scientists at Penn Medicine's Predictive Healthcare Team to try and develop a product that's ready for market and believe they can do so in the next few years.
"This is one of the first times we'll be able to take this machine learning approach and actually put it in the ICU or in an inpatient hospital setting and advise caregivers in a way that patients aren't going to be at risk," the team concludes. "That's really something novel."
Published at DZone with permission of Adi Gaskell, DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.