Many companies have evaluated the utility of deep learning, artificial intelligence and predictive analytics to identify unstable patients in the inpatient setting.  Nowhere is this need more useful that the intensive care unit where patient status can change minute to minute.  Physicians possessing advanced warning of decompensating patients can act quickly to prevent catastrophic events.

This clinical need has – as yet- been unfulfilled for a number of reasons.  First, as evidenced by the overwhelming lack of success of the “rapid response” teams to lower mortality, identifying instability does not necessarily translate into improved treatment.  In the case of rapid response teams, the only proven outcome has been bringing more (and less ill) patients to the ICU.  Moreover, as is the case of any data-analysis company, the data involved must be analyze-able – otherwise it’s GIGO (garbage in, garbage out).  Lastly, the predictive tools must be better than those currently available with our usual information.  A technology must prove utility above routine vital signs, laboratory examinations and observation.

Two new companies are attempting to break the paradigm for predicting decompensation in patients in the ICU.  Similarly, an Israeli company, Intensix, is conducting clinical trials attempting to prove its’ own predictive algorithms in ICU patients (https://www.intensix.com/).

Only time will tell if these algorithms will help care for patients and treat illness or merely provide more background noise for treating clinicians.

http://spectrum.ieee.org/the-human-os/biomedical/diagnostics/in-hospital-intensive-care-units-ai-could-predict-which-patients-are-likely-to-die