In the latest application of machine learning in healthcare, researchers at the UCLA Medical Center developed an algorithm to predict hypotension during surgery with strong accuracy (92% at 5 minutes before the onset of hypotension). The significance of this development is that it allows surgeons to proactively anticipate hypotension rather than react to it. This can help physicians avoid fatal post-op complications.
While critics question the usefulness of such algorithm since hypotension is preventable and experienced providers will likely have phenylephrine and ephedrine ready in such cases, it's always fun to think about what advancements such as this mean for the future of healthcare and what tools and automation will be available to physicians in coming years. Additionally, the data was collected from arterial waveform recordings, though in routine surgeries, arterial lines are not always utilized.
Despite limitations in usefulness over an experienced anesthesiologist, these developments are still worth investigation. As the technology is still in its infancy, there will undoubtedly be further advancements to real-world applications that will arise from additional research.
The methodology was built atop nearly 550,000 minutes of surgical arterial waveform recordings from 1,334 patients’ records (which included more than 25,000 instances of hypotension). It required high-fidelity recordings, but from them it could extract more than 3,000 unique features per heartbeat, leading to millions of data points to base the algorithm on, according to the study.