Researchers from Princeton (N.J.) University are using machine learning to design a system that could reduce the frequency of tests and improve the timing of critical treatments for intensive care unit patients.
To create the system, the researchers used data from more than 6,060 patients admitted to the ICU between 2001 and 2012. The research team presented its results Jan. 6 at the Pacific Symposium on Biocomputing in Hawaii.
The analysis looked at four blood tests measuring lactate, creatinine, blood urea nitrogen and white blood cells. These indicators help diagnose two serious problems for ICU patients: kidney failure or sepsis.
"Since one of our goals was to think about whether we could reduce the number of lab tests, we started looking at the [blood test] panels that are most ordered," said co-lead study author Li-Fang Cheng.
The team's algorithm uses a "reward function" that encourages a test order based on how informative the test is at a given time. In other words, there is greater reward in giving a patient a test if there is a higher probability that the patient's state is significantly different from the previous measurement.
To test the utility of the lab-testing policy they created, the researchers compared the reward function values that would have resulted from applying their system with the testing regimens that were actually used for the 6,060 patients in the study.
The researchers found the policy the machine learning algorithm generated would have yielded more information on the patient's condition than the actual testing their clinicians followed.
Additionally, when looking at white blood cell tests, the algorithm could have reduced the number of lab test orders by up to 44 percent.
They also found their approach would have helped alert clinicians to intervene sometimes hours sooner when a patient's condition started to deteriorate.
"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," said senior study author Barbara Engelhardt, PhD.