In a recently published study, Impact of individualized oxygenation targets on mortality in critically ill adults: a machine learning analysisresearchers in Aotearoa, New Zealand, have teamed up with scientists in the United States to use advanced machine learning techniques to pioneer new frontiers in medicine.
Traditionally, randomized clinical trials determine the average effect of a treatment on patient outcomes. Such clinical trials have revolutionized medicine by providing high-quality evidence about the effects of treatments on patient outcomes. However, one of the key problems with these trials is that they assume that all patients will respond to treatment in the same way.
In a recent study, a collaborative research team used machine learning to generate personalized predictions about the impact of increasing or decreasing oxygen levels on mortality in critically ill adults receiving life support in the intensive care unit (ICU). .
Specifically, we used machine learning to generate a model using data from one study, the Pragmatic Investigation of Optimal Oxygen Targets (PILOT) trial, and then used data from another study, the Intensive Care Unit Randomized Trial. We tested the model using data from a trial comparing two. Approaches to the Oxygen Therapy (ICU-ROX) Trial.
Their results were amazing. Professor Paul Young, one of the study's senior investigators and Deputy Director and Head of Intensive Care Medicine at the Medical Research Institute of New Zealand (MRINZ), explains: Lower oxygen levels changed dramatically. At one extreme, the model predicts a 27.2% absolute reduction in mortality risk by using a lower oxygen target, and at the other extreme, the model predicts a 27.2% absolute reduction in mortality risk by using a higher oxygen target. , predicted a 34.4% absolute reduction in mortality risk. ”
“Overall, if patients in the ICU-ROX trial had received the oxygen therapy recommended by the machine learning model, their mortality rate would have been 6.4 percentage points lower,” Professor Young said.
Dr Alex Psirides, chair of Te Whatu Ora's Critical Care Advisory Group, said: “About 40% of the 24,000 patients admitted to New Zealand's ICUs each year are on life support. These findings are confirmed. It is expected that approximately 600 New Zealand lives will be saved each year.”
“If the results are true and generalizable, they would be surprising,” Dr. Derek Angus, associate editor of the Journal of the American Medical Association, commented in an accompanying editorial. “If we could instantly assign every patient to the appropriate predicted group,
If we determine benefit versus harm and assign oxygen targets accordingly, this intervention would theoretically result in the largest improvement in lives saved from critical illness in the history of the field. ” said Dr. Angus.
MRINZ Director Professor Richard Beasley pointed out that the implications of this research go far beyond oxygen therapy for patients in the ICU, saying: “We are using machine learning to use data from clinical trials to personalize patient treatment. “This technique for predicting reactions is probably the best we've ever had.” Advances in medical evidence generation occur over the decades. ”
Professor Beasley said: “It is difficult to overstate the extent to which this type of research has the potential to change medicine.”
“For decades, medical practices have experienced the tension between choosing between individualized care that is not evidence-based and care that is evidence-based but not individualized. Paul Young and his colleagues have shown that machine learning methods can predict individual treatment outcomes, enabling evidence-based personalized care.” says Professor Beasley.
This study represents an important advance in medicine with its innovative use of machine learning in determining personalized oxygenation goals. It has the potential to bridge the gap between evidence-based and personalized care, potentially significantly improving patient outcomes. This approach represents a transformative change in medical research and practice.
