An Algorithm Can Predict Cardiac Arrest 24 Hours Before it Happens

We may earn a commission from links on this page.

Predictive Medical Technologies has developed a system that can mine the medical data of a patient—lab reports, monitors, nurse notes, etc.—and predict whether that patient will suffer from cardiac arrest or respiratory failure within 24 hours.

It's a system that can be integrated into hospitals that are "at a certain technological level" without any new hardware, sampling or extra time. That technological level is rare though, with only 100 US hospitals properly equipped. Bryan Hughes, CEO of Predictive Medical Technologies, explains how it works:

Without giving away too much of our secret sauce, we use non-hypothesis machine learning techniques, which have proven very promising so far. This approach allows us to eliminate any human "expert" bias from the models.


The current model has been tested and proven retrospectively (looking at old data and determining outcome) and can be used to predict cardiac arrest and respiratory failure. The next version should be able to detect sepsis, renal failure and re-intubation risk, as well. PMT is going to start a validation trial to see how well it works in real times but a formal FDA trial is still a year away.

This sounds deliciously futuristic (even though it's just plain math) and puts us one step closer to the precogs of Minority Report. Well, sorta. [Predictive Medical Technologies via Forbes]